...

Non-Invasive Assessment of Liver Fibrosis with P-Magnetic Resonance Spectroscopy and Dynamic Contrast

by user

on
3

views

Report

Comments

Transcript

Non-Invasive Assessment of Liver Fibrosis with P-Magnetic Resonance Spectroscopy and Dynamic Contrast
Linköping University Medical Dissertations No. 1351
Non-Invasive Assessment of Liver
Fibrosis with 31P-Magnetic Resonance
Spectroscopy and Dynamic Contrast
Enhanced Magnetic Resonance Imaging
Bengt Norén
Faculty of Health Sciences
Division of Radiological Sciences
Department of Medicine and Health Sciences
and Center for Medical Image Science and Visualization
Linköping University, Sweden
Linköping 2013
Non-Invasive Assessment of Liver Fibrosis with 31P-Magnetic Resonance
Spectroscopy and Dynamic Contrast Enhanced Magnetic Resonance Imaging
Linköping University Medical Dissertations No. 1351
© Bengt Norén, 2013
Published articles have been reprinted with the permission of the copyright holder
ISBN 978-91-7519-705-0N
ISSN 0345-0082
Printed by LiU-Tryck, Linköping, Sweden, 2013
CONTENTS
Background
1
Diffuse Liver Disease
4
Liver Fibrosis, Cirrhosis, Complications and HCC
4
Evaluation of Liver Function and Prognostic Scores
7
Assessment of Fibrosis: Present ‘Gold Standard’
8
Assessment of Fibrosis: Non-Invasive Techniques
9
Nuclear Magnetic Resonance
12
Basic Physics
12
The Spectroscopy Technique – MRS
13
Chemical Shift
13
Spin-Spin Coupling
14
Localization Methods
14
Liver Metabolites of Interest in 31P –MRS
14
Dynamic Contrast Enhanced MRI – DCE-MRI
16
Gd- EOB- DTPA (Primovist®)
16
Hepatocyte Uptake and Excretion Mechanisms
17
Quantification Procedures
31P
18
–MRS
18
DCE-MRI
19
Aims of the Study
21
Materials and Methods
22
Localized In Vivo 31P NMR Spectroscopy
22
Data Acquisition
22
External Referencing
23
Processing
23
Absolute Quantification of In Vivo
Liver Metabolite Concentrations
Dynamic Contrast Enhanced MRI – DCE-MRI
25
26
Data Acquisition
26
Image Analysis
27
Quantitative Measurements of Gd-EOB-DTPA Uptake
27
Visual Assessment of Gd-EOB-DTPA Excretion
30
Subjects, Paper I –IV
32
Clinical Data
33
Laboratory Analysis
33
Liver Biopsy and Histopathological Grading
34
Statistical Analysis
Results
36
38
Localized In Vivo 31P NMR Spectroscopy, Paper I-II
38
Concentrations Determined Using MRS
38
MRS Concentrations Expressed as Anabolic Charge, AC
40
MRS vs Laboratory Data
42
Dynamic Contrast Enhanced MRI, Paper III-IV
43
Final Diagnosis and Fibrosis Scoring
43
Pharmacokinetic Uptake Parameters vs Fibrosis Stage
43
Visually Assessed Contrast Excretion vs Contrast Uptake
Parameters, Histopathology and Blood Tests
Discussion
31P
– MR Spectroscopy
46
48
48
Dynamic Contrast Enhanced MRI
52
Clinical Significance
56
Conclusions
62
Acknowledgements
63
References
64
Original Papers
73
ABSTRACT
The present study aims at demonstrating phosphorus metabolite concentration
changes and alterations in uptake and excretion of a hepatocyte specific contrast
agent in patients with diffuse - or suspected diffuse - liver disease by applying two
non-invasive quantitative MR techniques and to compare the results with histopathological findings, with focus on liver fibrosis.
In the first study phosphorus-31 MR spectroscopy using slice selection (DRESS) was
implemented. Patients with histopathologically proven diffuse liver disease (n = 9)
and healthy individuals (n = 12) were examined. The patients had significantly lower
concentrations of phosphodiesters (PDE) and ATP compared with controls. Constructing an ‘anabolic charge’ (AC) based on absolute concentrations, [PME] / ([PME]
+ [PDE]), the patients had a significant larger AC than the control subjects.
The MRS technique was then, in a second study, applied on two distinct groups of
patients, one group with steatosis and none-to-moderate inflammation (n = 13) and
one group with severe fibrosis or cirrhosis (n = 16). A control group (n = 13) was also
included. Lower concentrations of PDE and a higher AC were found in the cirrhosis
group compared to the control group. Also compared to the steatosis group, the cirrhosis group had lower concentrations of PDE and a higher AC. A significant correlation between fibrosis stage and PDE and fibrosis stage and AC was found. Using
an AC cut-off value of 0.27 to discriminate between mild (stage 0-2) and advanced
(stage 3-4) fibrosis yielded an AUROC value of 0.78, similar as for discriminating between F0-1 vs. F2-4.
Dynamic contrast enhanced MRI (DCE-MRI) was performed prospectively in a third
study on 38 patients referred for evaluation of elevated serum alanine aminotransferase (ALT) and/or alkaline phosphatase (ALP) levels. Data were acquired from rei
gions of interest in the liver and spleen by using single-breath-hold symmetrically
sampled two-point Dixon 3D images time-series (non-enhanced, arterial and venous
portal phase; 3, 10, 20 and 30 min) following a bolus injection of Gd-EOB-DTPA
(0.025 mmol/kg). A new quantification procedure for calculation of the ‘hepatocyte
specific uptake rate’, KHep, was applied on a two-compartment pharmacokinetic
model. Liver-to-spleen contrast ratios (LSC_N) were also calculated. AUROC values
of 0.71, 0.80 and 0.78, respectively, were found for KHep, LSC_N10 and LSC_N20
with regard to severe versus mild fibrosis. Significant group differences were found
for KHep (borderline), LSC_N10 and LSC_N20.
In study four, no significant correlation was found between visual assessments of
bile ducts excretion of Gd-EOB-DTPA and histo-pathological grading of fibrosis or
the quantified uptake of Gd-EOB-DTPA defined as KHep and LSC_N.
In conclusion 31P-MRS and DCE-MRI show promising results for achieving a noninvasive approach in discriminating different levels of fibrosis from each other.
ii
LIST OF PAPERS
I. Absolute quantification of human liver metabolite concentrations by localized
in vivo 31P NMR spectroscopy in diffuse liver disease.
Noren, B., Lundberg, P., Ressner, M., Wirell, S., Almer, S., and Smedby, Ö.
(2005) Eur Radiol 15(1), 148-157
II. Separation of advanced from mild fibrosis in diffuse liver disease using 31P
magnetic resonance spectroscopy
Noren B, Dahlqvist O, Lundberg P, Almer S, Kechagias S, Ekstedt M, Franzén L,
Wirell S and Smedby Ö. (2008) European Journal of Radiology 66 (2), 313-320,
III. Separation of Advanced from mild fibrosis by quantification of the
hepatobiliary uptake of Gd-EOB-DTPA. Noren B, Forsgren MF, Dahlqvist Leinhard
O, Dahlström N, Kihlberg J, Romu T, Kechagias S, Almer S, Smedby Ö, Lundberg P
(2012) Eur Radiol. 23(1), 174-181.
IV. Visual assessment of biliary excretion of Gd-EOB-DTPA in patients with
suspected diffuse liver disease – a biopsy-controlled prospective study. Norén B,
Dahlström N, Forsgren M F, Dahlqvist Leinhard O, Kechagias S, Almer S, Wirell S,
Smedby Ö, Lundberg P. (2012) Manuscript.
iii
AUTHOR CONTRIBUTIONS
Paper I
I participated in the planning of the study together with radiation physicists and
hepatologists. I coordinated the pilot study, assisted in the data analysis and was
partly responsible for writing, editing and submitting the manuscript.
Paper II
I participated in the planning of the study design and was responsible for the management, coordination and partly executing the MRS examinations. I assisted in the
data analysis and interpretation, wrote the first and final draft of the manuscript and
managed the correspondence with the journal.
Paper III
I participated in the initial planning. I was responsible for the clinical part of the data
processing and assisted in the analysis and interpretation. I wrote the first and final
draft of the manuscript and managed the correspondence with the journal
Paper IV
I participated in the planning of the study design and the image review process. I
performed the image review, assisted in the analysis and wrote the first and final
draft of the manuscript
iv
LIST OF ABBREVIATIONS
2,3-DPG
2,3-diphosphoglycerate
99m
Tc
AC
99m
ADC
Apparent diffusion coefficient
AIH
Autoimmune hepatitis
ALP
Alkaline phosphatase
ALT
AMARES
Alanine aminotransferase
AMP
Adenosin monophosphate
AST
ATP
Aspartate aminotransferase
AUROC
Area under receiver-operating characteristic curve
CBD
Common bile duct
CHC
Chronic hepatitis C
CLD
CSI
Chronic liver disease
CT
Computed tomography
DCE-MRI
Dynamic contrast enhanced MRI
DRESS
DWI
Depth- Resolved Surface coil Spectroscopy
ECM
Extracellular matrix molecules
EES
ER
Extracellular extravascular space
ERETIC
Electronic Reference To access In vivo Concentrations
FA
Flip angle
FID
Free Induction Decay
FT
Fourier transformation
Gd
Gd-EOB-DTPA
Gadolinium
GGT
Gamma glutamic transpeptidase
GPC
Glycerophosphocholine
GPE
Glycerophosphoethanolamine
GSA
Galactosyl human serum albumin
HbsAg
Hepatitis B surface antigen
HBV
Hepatitis B virus
HCC
Hepatocellular carcinoma
HCV
Hepatitis C virus
HSC
Hepatic stellate cells
ICG(R)-15
Indocyanine green retention at 15 min
Technetium
Anabolic Charge
Advanced Method for Accurate, Robust and Efficient Spectral fitting
Adenosine triphosphate
Chemical Shift Imaging
Diffusion-weighted imaging
Endoplasmic reticulum
Gadolinium ethoxybenzyl diethylenetriaminepentaacetic acid
v
ISIS
Image Selective In vitro Spectroscopy
LSC_N
Liver-to-spleen contrast ratio, normalized
MANA
Multi scale adaptive normalizing averaging
MELD
Model for end-stage liver disease
MeP
Methyl Phosphonate
MMP
Matrix metalloproteinase
MP
Mobile phospholipids
MRE
MRI
Magnetic Resonance Elastograhpy
MRP
Multidrug resistance protein
MRUI
Magnetic resonance user interface
NADH
Nicotinamide adenine dinucleotide
NAFLD
Non-alcoholic fatty liver disease
NASH
NMR
Non alcoholic steatohepatitis
NTCP
Na(+)-taurocholate-cotransporting polypeptide
OATP
Organic anion transporting polypeptide
PBC
Primary biliary cirrhosis
PC
Phosphocholine
PCr
PDE
Phosphocreatine
PE
Phosphoethanolamine
Pi
Inorganic phosphate
PK (INR)
Protrombinkomplex International Normalized Ratio
PME
Phosphomonoesters
ppm
Parts per million
ROI
Region of interest
SI
Signal intensity
SNR
Signal to noise ratio
TE
TGF-β
TIMPS
Transient elastography
TIPS
UDPG
VARPRO
Magnetic Resonance Imaging
Nuclear Magnetic Resonance
Phosphodiesters
Transforming growth factor-beta
Tissue inhibitors of metalloproteinases
Transjugular intrahepatic portosystemic shunt
Uridinediphospho-glucose
Variable Projection
vi
BACKGROUND
The importance of the liver was clear already to medical practitioners in antiquity
since it produced one of the four body fluids – the yellow bile. Sickness was a result
of an imbalance between blood (heart), black bile (spleen), phlegma (brain) and yellow bile (liver). The idea of the four fluids remained in medicine until the 19th century
1
. The liver is extremely versatile and can be regarded as both an endocrine and exo-
crine gland. A multitude of vital molecules and substances such as proteins, cholesterol, triglycerides and bile acids are synthesized in the liver. As an exocrine gland
the liver produces bile. Other important functions concern glucose and fat metabolism, biotransformation of foreign substances, e.g. drugs, and storage of metabolites
and vitamins.
Diffuse liver disease includes a wide spectrum of different etiologies. They all have
Fig. 1. Etiology of diffuse liver disease. NAFLD= Non-alcoholic fatty liver disease.
NASH= Non-alcoholic steato hepatitis
1
the potential of causing chronic liver disease (CLD) and development of fibrosis possibly culminating in cirrhosis with an increased risk for hepatocellular carcinoma,
HCC. Some of the most common are summarized in Fig. 1.
A liver biopsy may be needed to help establish the diagnosis, histological inflammatory grade and fibrosis stage. The prognosis of CLD largely depends on the extent
and progression of liver fibrosis. There are, however, well-known drawbacks such as
the risk of complications, inter- and intra-observer variability, inaccurate staging due
to sampling error and the fact that heterogeneous distribution of fibrosis in the liver
parenchyma may not be reflected in a single biopsy
2 3 4 5 6 7.
Further, the speed of
fibrosis progression, commonly not linear over time 8, cannot be answered by a single
biopsy and serial biopsies are not an attractive solution for a number of reasons.
Animal models as well as data in human liver disease demonstrate that fibrosis, and
even cirrhosis, may be reversible
9 10.
These observations have strengthened the ef-
forts to find non-invasive alternatives allowing a close monitoring of patients and
facilitating clinical decision-making.
The challenge is to provide hepatologists with equal or better information compared
to a liver biopsy. Serum biomarkers, ultrasound elastography and a number of MR
applications have been proposed to replace liver biopsy either as single methods or
in combinations.
In order to compete with liver biopsy, non-invasive techniques should accurately
provide information concerning grade of inflammation, intracellular lipid content
and the presence of hepatic iron overload - apart from a correct fibrosis staging.
Up to date, no non-invasive technique, either alone or in combinations, completely
fulfils these demands. On the other hand, recent reports using MR techniques show
promising results in providing additional functional information, something a conventional biopsy will fail to reveal
11 12.
Although the full capacity of non-invasive
techniques has not yet been clarified, a shift towards a routine use of such measure2
ments is already here. In France, for example, non-invasive methods such as Fibrotest
(serum biomarkers) and Fibroscan (ultrasound elastography) have been approved by
the public health care system as first line estimates of fibrosis in patients with CHC. 13
The present study aims at demonstrating results from the application of two quantitative MR techniques in patients with verified diffuse - or suspected - liver disease, to
compare the results with histopathological staging of fibrosis, to compare with reports from other non-invasive techniques, and to discuss the potential of a multimodal MR approach as a clinical tool.
3
Diffuse Liver Disease
Liver Fibrosis, Cirrhosis, Complications and HCC
Liver fibrosis is due to repeated and/or longstanding liver injury of various etiologies
resulting in an abnormal continuance of the wound healing process causing excessive accumulation of extracellular matrix molecules (ECM) including collagen. The
extent and progression of liver fibrosis is crucial for the prognosis and management
of patients with liver disease. The natural history of liver fibrosis is influenced by
both genetic and environmental factors and it is a dynamic process – not static.
The space of Disse, located between the hepatic sinusoids and the hepatocytes, is
filled with fibrous scar tissue and the forming of fibrous scaring destroys the hepatic
architecture (Fig. 2).
Fig. 2. Anatomical relationship between hepatocytes, space of Disse and
Sinusoidal lumen
4
The main responsible cell for ECM production and fibrosis formation is the
myelofibroblast derived either from activated hepatic stellate cells (HSC), normally
storing vitamin A, or perivascular fibroblasts. In healthy liver the turn-over and homeostasis of ECM is regulated by matrix metalloproteinases (MMPs) and their specific inhibitors, TIMPS (tissue inhibitors of metalloproteinases). In chronically damaged
liver fibrogenic cytokines, e.g. TGF-β, and growth factors are released from macrophages, inflammatory cells and bile duct epithelia and responsible for the
myofibroblast activation. Initially the fibrogenesis is counterbalanced by removal of
excess ECM by proteolytic enzymes (MMPs). In activated HSC especially the expression of TIMP-1 is up regulated leading to inhibition of MMP and in the long run
fibrogenesis is favored over fibrolysis. Depending on origin of the hepatic injury the
fibrous tissue is initially differently distributed; fibrotic tissue is located around portal tracts in chronic viral hepatitis and cholestatic disorders while pericentral and
persinusoidal regions are affected in alcohol-induced liver disease. As the fibrosis
progresses bridging fibrosis and finally cirrhosis - an advanced stage of fibrosis - develop and form nodules of regenerating hepatocytes.
In cirrhosis the hepatic vascular structure is distorted leading to shunting of portal
and arterial blood directly into the hepatic veins, thus with impaired exchange between hepatic sinusoids and the hepatocytes 14 15 16 (Fig 3).
5
Fig. 3. Liver biopsy. van Gieson staining. Cirrhosis in a patient with NAFLD
The term ‘compensated cirrhosis’ describes the asymptomatic phase. As the disease
gradually progresses, a ‘decompensated’ stage may occur. This is characterized by
portal hypertension due to an increased intrahepatic vascular resistance and hepatocellular dysfunction. The transition from compensated to a decompensated stage
dramatically changes the median survival from 12 to 2 years and it occurs at a rate of
approximately 5-7% per year. Portal hypertension may cause severe and even lethal
complications such as bleeding from gastroesophageal varices, ascites, spontaneous
bacterial peritonitis, hepatorenal syndrome and hepatic encephalopathy
17
.
A typical feature in the cirrhotic liver is the development of regenerating nodules i.e.
hepatocytes with a high mitosis frequency. In some of these regenerating nodules
mutations occur causing low-grade dysplasia which may transform to high grade
dysplastic nodules and ultimately HCC – a process that usually takes 20-30 years. 18
The vast majority of HCC develop in cirrhotic livers, only about 10% in unaffected
liver parenchyma. Globally, HCC is among the ten most common malignancies ac6
counting for about 300,000 deaths each year. Worldwide, chronic HBV and HCV infections account for more than 80% of HCC with an increasing incidence. In Sweden
non viral liver disease is the most common cause of liver cirrhosis and the incidence
of HCC is about 4 -500 cases/year 18.
Evaluation of Liver Function and Prognostic Scores
Evaluation of liver function comprises blood liver tests reflecting the status of the
hepatocytes as well as the biliary tract, clearance/retention tests such as indocyanine
green (ICG-15) and scintigraphy techniques such as galactosyl human serum albumin (GSA) labelled with
Technetium (99mTc) for estimation of liver function. De-
99m
spite some similarities between clearance/retention tests and scintigraphy, the
scintigraphic techniques are nowadays utilized to a lesser extent 19 20 21 22 .
In patients with cirrhosis, clinical scoring systems are in use for predicting prognosis
and the timing for liver surgery, including transplantation. The Child-Turcotte score,
later modified as Child-Pugh score includes five empirically selected variables, two
based on clinical examination (encephalopathy and ascites) and three on blood tests
(bilirubin, albumin and prothrombin) 23, see Table 1.
Table 1. Child-Pugh score
Points
1
2
3
Encephalopathy
None
Minimal
Advanced(coma)
Ascites
Absent
Controlled
Refractory
S-Bilirubin(µmol/l)
< 34
34 - 51
> 51
S-Albumin(g/l)
> 35
28 - 35
< 28
PK (INR) *
< 1.7
1.7 - 2.3
> 2.3
* Limits from American Association of the Study of Liver Diseases.
7
By adding the individual points patients can be categorized into three groups of increasing severity; A (5-6 points), B (7-9 points) and C (10-15 points). Several limitations of this scoring procedure have been pointed out, such as empirically selected
variables, arbitrary use of cut-off values for the quantitative parameters, the fact that
each variable is given the same weight, the influence of renal function in the course
of cirrhosis is not included and the cause of cirrhosis is missing. Despite these shortcomings the Child-Pugh score is the most frequently used scoring system in predicting severity of the liver disease and it is easy to use in clinical routine.
The model for end-stage liver disease (MELD), originally created to predict the survival after transjugular intrahepatic portosystemic shunt (TIPS), was in 2002 adopted
in the US as the reference scoring system to rank and select patients for liver transplantation. The MELD score, based on serum creatinine, serum total bilirubin and
PK(INR), predicts short term mortality and the ideal timing for operation 17 18 24 25 26
Assessment of Fibrosis: Present ‘Gold Standard’
Liver biopsy is the ‘gold standard’ for the assessment of liver fibrosis. One commonly
used scoring system was developed by Batts and Ludwig
27
, see Table 2.
Table 2. Batts – Ludwig fibrosis scoring system
Stage
Description
Criteria
0
No fibrosis
Normal connective tissue
1
Portal fibrosis
Fibrous portal expansion
2
Periportal fibrosis
Periportal or rare portal-portal septa
3
Septal fibrosis
Fibrous septa with architectural distortion; no obvious cirrhosis
4
Cirrhosis
Cirrhosis
8
Other scoring systems, e.g. the Ishak and Metavir scores, are specifically designed for
patients with CHC. Both grade and stage are scored – the grade indicating inflammatory activity and the stage the amount of fibrosis present 28 29 .
Assessment of Fibrosis; Non-Invasive Techniques
Serum biomarkers, ultrasound elastography and a number of MR-applications have
been proposed in the assessment of liver fibrosis either as single methods or as combinations.
Serum Markers: There are multiple serum biomarkers available and they can be divided into two main types; direct – and indirect – markers. Direct markers such as
hyaluronan, procollagen III, type IV collagen30 can be measured individually or in
combination with other markers of liver fibrogenesis. Indirect markers reflect
metabolization or synthesis in the liver.
Fibrotest® is one commonly used fibrosis panel. It combines measurements of five
variables; alpha-2-macroglobulin, haptoglobin, GGT, apolipoprotein A1, and total bilirubin30.
Transient Elastography (TE): Sonographic elastography (FibroScan®) measures liver stiffness by emitting low-frequency waves (50 Hz) into the liver. The velocity of
the wave propagation is proportional to the tissue stiffness (the stiffer the tissue the
faster the wave propagation) and measured in kilopascals (kPa). In normal liver results are reported to be 4 – 6 kPa and in cirrhotic livers > 12 – 14 kPa. TE appears to
be able to differentiate healthy and cirrhotic livers but is less accurate to separate
normal liver from stage F1, stage F1 from F2 and even F2 from F3
31.
In a large pro-
spective study comprising about 13 000 examinations liver stiffness measurements
were uninterpretable in nearly one of five examinations mainly due to obesity and
limited operator experience 32 .
9
Contrast-Enhanced Ultrasound: A decrease in ultrasound contrast agent transit time
throughout the liver in cases of cirrhosis has been demonstrated
33 34
. Searle et al
demonstrated significant differences between F1 and F3 and F1 and F4 when measuring the difference between the hepatic vein and hepatic artery contrast arrival times
35.
Staub et al found that a transit time of < 13 s can separate advanced fibrosis with
an estimated AUROC of 0.85 using the latest contrast generation not being taken up
by the liver parenchyma 36.
Combination of Non-Invasive Methods: There are reports demonstrating that the
combination of various non-invasive methods appears promising. The combination
of TE and Fibrotest increases the diagnostic accuracy and the AUROC for significant
fibrosis (≥F2) and cirrhosis (F4), being 0.88 and 0.95 respectively 37.
Perfusion CT: When performing perfusion studies, microcirculatory changes in cirrhosis have been described
38 39
. In a one study by Ronot et al a mean transit time of
13.4 s allowed discrimination between fibrosis stage F1 and F2-3 with a sensitivity
and specificity of 71% and 65% respectively 40.
Functional MRI Methods: In addition to the MR applications presented in this
study additional techniques such as double-contrast enhanced MRI, diffusionweighted MRI (DWI) and MRE (elasticoviscous properties) have been proposed as
tools in the non-invasive approach to assess liver disease.
Double-Contrast Enhanced MRI: In double contrast enhanced MRI both SPIOs and extracellular gadolinium-based contrast is administrated. SPIOs will cause signal loss
in normal liver parenchyma while extracellular gadolinium-based contrast on delayed images will cause signal enhancement of fibrotic tissue. The internal liver architecture can be assessed by quantitative texture analysis. Bahl et al evaluated a
model based on texture features and found sensitivity and specificity of 91.9% and
83.9% respectively for classifying fibrosis F≤2 vs F≥3 41. In a study by Aguirre et al,
sensitivity, specificity, and accuracy values higher than 90% were achieved, with re10
gard to fibrosis, although the diagnostic performance depended on the specific sequence and scoring system used 42.
Diffusion-Weighted MRI (DWI): In order to assess whether hepatic fibrosis, or cirrhosis, is associated with a restriction in the diffusion of water in the liver, diffusionweighted MR imaging (DWI) has been proposed. It is recommended that the ‘apparent diffusion coefficient’, ADC, is based on at least 3 b-values and calculated within
multiple areas of the liver. Several studies have shown that ADC values are reduced
in cirrhotic livers 8 43.
MR-Elastography (MRE): MRE is a phase-contrast-based MRI imaging technique. Mechanical waves can be visualized and quantitatively measured in the liver parenchyma and waves in the range of 60-150 Hz are generated by an external force, i.e. an
acoustic driver placed over the right anterior chest wall. By estimating the wavelength of the strain waves from the acquired MRE images, the elasticity/stiffness is
calculated and measured (in kPa). Sensitivity and specificity figures of 98% and 99 %
respectively for detecting liver fibrosis have been reported. Discriminating between
mild fibrosis (F0-1) and moderate-severe (F2-4) shows higher accuracy than TE with
sensitivity and specificity figures each in the 80-85% range. The presence of ascites or
obesity has little effect on MRE and the examination covers a bigger volume of the
liver with the potential of a global assessment 8 43.
11
Nuclear Magnetic Resonance
The ‘Nuclear Magnetic Resonance’ phenomenon (NMR) was independently described in the 1940s by the groups of Purcell and Bloch
44 45.
They managed to meas-
ure the magnetic resonance in bulk material, liquids and solids. When it was later
demonstrated that the NMR frequency for the same nucleus in different chemical
compounds was different, it became a widely applied technique in chemistry for analyzing and characterizing the structure of molecules in solution, including biological
macromolecules. During the 1970s, several technical improvements made the NMR
phenomenon applicable in a clinical setting producing the first images in the late 70s
(cross section through a finger) and in the mid-1980s clinical MR-scanners were
commercially available.
Basic Physics
When atomic nuclei with magnetic properties (nuclei with an odd number of neutrons or protons) are placed in a magnetic field, they possess a basic property called
spin. This causes the nuclei to behave like a small magnet aligning itself with the external magnetic field. There is a high and a low energy state with the majority of the
spins in the lower state. The spins precess (or ‘rotation’) with a certain frequency depending on the strength of the external magnetic field and different nuclei have different frequencies (this is the characteristic Larmor frequency). Adding a radio frequent pulse, with the same frequency as the Larmor frequency, the nuclei will absorb
energy. The low energy state spins will be transformed to the higher state, and when
the pulse is turned off the nuclei return to equilibrium, due to the influence of T1 and
T2 relaxation, and energy is lost to the surroundings. This energy can be detected as
a NMR signal and transformed either to an image or a spectrum.
12
The Spectroscopy Technique
The basic condition for MRS is the presence of isotopes with a magnetic moment (i.e.
‘magnetic nuclear spins’). Relatively few molecules can be identified and only freely
mobile molecules (such as small metabolites in solution) provide enough signal for
detection, provided that the concentration is not too low.
MRS offers a non-invasive range of methods to study many metabolic conditions
such as evaluating the energy metabolism and synthesis/degradation of cell membranes as well as the intracellular pH and free Mg2+ concentration (31P -MRS), triglyceride content (1H-MRS) and the hepatic glucose metabolism (13C-MRS).
A magnetic field strength of 1.5 T, or greater, is typically necessary to obtain a sufficient signal-to-noise-ratio (SNR) and surface coils tuned to the nucleus of interest are
frequently used.
Chemical Shift
Due to differences in the intramolecular chemical environment and bonds a 1H nucleus in for example water and one in fat have different resonance frequencies.
Chemical shift is the relative difference in frequency measured in Hz and the difference in resonance frequency is usually in the range of 10 – 1200 Hz. The numbers are
relatively small and therefore multiplied with one million and expressed in ppm.
Thus the chemical shift difference between water and fat is 3.5 ppm at all field
strengths, corresponding to 220 Hz at 1.5 T.
The precise position of the resonance within a spectrum determines which chemical
compound has contributed to the signal and the area under the resonance waveform
is proportional to the number of molecules in the tissue. In this way MRS can be used
for measurements of both relative and absolute metabolite concentration.
The chemical shifts are field-independent unique frequencies of the resonances in the
spectrum. The unique shifts of resonances enable an easy identification of different
13
molecules. It can also provide unique information about molecular structure. The
frequency shift depends on the B0 field and is therefore often expressed in relation to
a reference compound. In 31P MRS the most commonly used reference is 85% H3PO4
(phosphoric acid at 0.00 ppm). Using this shift reference phosphocreatine (PCr) appears at -2.35 ppm.
Spin-Spin Coupling
Several spins in the same molecule can affect each other provided that the chemical
bond distance is equal to or shorter than about four bonds. The result is a splitting of
the resonances into doublets, triplets or quartets. If no splitting occurs the result is a
so-called singlet (e.g. PCr). For a spin system the splitting (J-value measured in Hz) is
constant and independent of the field strength.
Localization Methods
An important step in MRS is the accurate selection of the volume of interest. A
common technique is to combine a surface coil (usually a flat circular single turn coil
often used as both transmitter and receiver coil) with a volume selection sequence.
The most often used sequences concerning liver spectroscopy are DRESS, ISIS and
CSI.
Liver metabolites of interest in 31P – MRS
Due to its anatomic and metabolic features the liver is well suited for spectroscopy
studies. A 31P -MRS spectrum with the metabolites of interest is provided in Fig 4.
The 31P-MRS technique allows for detection and quantification of several phosphorus
compounds involved in energy metabolism (ATP and Pi) and membrane phospholipid metabolism (PME and PDE) 46. The PME resonance in a liver spectrum is mainly composed of resonances from phosphoethanolamine (PE) and phosphocholine
14
(PC), intermediates in the phospholipid synthesis, but also from AMP, coenzyme-A,
2,3-DPG and intermediates of carbohydrate metabolism (glycolysis) 47.
Fig. 4. 31P-MRS spectrum with an external reference MeP which appears at about 35 ppm.
Previous studies have shown elevated levels in the liver of PE and PC in rapidly proliferating non-malignant cells. The PDE resonance has two main contributors:
Glycerophosphoethanolamine (GPE) and glycerophosphocholine (GPC), but also
signals from endoplasmic reticulum are believed to contribute to the PDE resonance.
Concentration rates of these phosphodiester metabolites are decreased when there is
an increase in cell turn over or in rapidly proliferating cells.
In summary, changes in the concentrations represented by these two resonances
probably reflect the phospholipid membrane synthesis and the increase in cell turnover of the hepatocytes
48 49.
A broad resonance of mobile phospholipids (MP) is also present slightly up-field
from PDE. Other metabolites of particular interest are inorganic phosphate (Pi) and
the -, -, - groups of ATP (or rather NTP) reflecting the energy state, as well as
NAD(H) representing the redox conditions in the cells.
15
Pi is sensitive to changes in pH, while the chemical shifts of PCr and α-ATP are stable at normal physiological pH values. Therefore the chemical shift difference between Pi and PCr, or α-ATP, can be used to determine intracellular pH in the liver.
Dynamic Contrast Enhanced MRI – DCE-MRI
Gd-EOB-DTPA (Primovist®)
Gd-EOB-DTPA was launched in 2004 as Primovist® in Europe and Asia and as
Eovist® in the U.S. in 2008. It is a hepatocyte-specific contrast medium administered
intravenously and excreted in roughly equal amounts by the kidneys and hepatobiliary system, 43-53% and 41-51% respectively 50.
Primovist® is a paramagnetic contrast agent combining the properties of a typical extracellular agent with those of a hepatocyte specific one. Due to high protein binding
capacity, the T1 relaxivity increases resulting in an increase in signal intensities originating from blood and liver parenchyma. After a bolus injection of 0.025 mmol/kg,
corresponding to 0.1 mL/kg, the contrast agent behaves similarly to non-specific Gdchelates during the dynamic phases and improves the detection and characterization
of liver lesions in the hepatobiliary phase
51.
The maximum liver-specific enhance-
ment is reached after about 20 minutes in healthy livers, and in late phases, the biliary excretion allows for T1-weighted magnetic resonance cholangiography (MRC)
for assessment of bile ducts including detection of leakage from the bile ducts.
Since the injection volume of Gd-EOB-DTPA is smaller than that of non-specific gadolinium agents, this may cause timing problems and truncation artifacts in the arterial phase. Fluoroscopic triggering in combination with a low injection rate (1 mL/s), or
dilution with saline to allow for rapid injection of 2 mL/s have been suggested solutions. Gd-EOB-DTPA is well tolerated by humans, and no case of nephrogenic systemic fibrosis has been reported 51.
16
Hepatocyte uptake and excretion mechanisms
Gd-EOB-DTPA is transported over the sinusoidal membrane via the organic anion
transporting polypeptides OATP1B1 and OATP1B3. Bidirectional transport is seen
within the OATPs. Gd-EOB-DTPA is not metabolized within the hepatocytes and it is
excreted into the bile through the ‘multidrug resistance protein 2’, MRP2. The MRP2
excretion is a unidirectional ATP-dependent active transporter; moreover the
transport rate through MRP2 is limited, which results in a retention of Gd-EOBDTPA in the hepatocytes (Fig 5). Located at the sinusoidal membrane there are two
other members of the MRP family, MRP3 and MRP4, which may be expressed and
up-regulated under cholestatic conditions returning bile salt to portal circulation 51 52.
Fig. 5 Uptake of liver specific contrast agent, Gd-EOB-DTPA (‘Gd’), bilirubin (‘BR’) and bile salts (‘BS’)
into hepatocytes. The coloured arrows correspond to direction of transport as the transported molecules and
MRP3 and MRP2 represent transporting proteins other than diffusion; MR2 into the bile canaliculus. The
sizes of the symbols reflect their relative importance. There is a pronounced competition between the contrast agent and naturally occurring molecules. Appreciate the many different possible routes for flux contrast agent into and out of the hepatocytes. Omitted in this picture is the transporting protein OATP2,
which facilitates bile salts and bilirubin transfer into the hepatocyte.
17
Quantification Procedures
It is highly desirable to avoid semi-quantitative methods (in which SI is simply defined using arbitrary intensity scaling) since comparison between repeated examinations of the same patient, between patients, and between examinations performed on
different MRIs, is likely to be inconsistent, resulting in variable and inconclusive results.
31
P-MRS
The majority of previous 31P-MRS reports express metabolite concentrations as resonance integral ratios. If, e.g., PME/PDE is increased, it is then not clear whether PME
is increased or PDE is decreased. Absolute quantification is better suited to detect
which component(s) is involved in the observed changes. By using an external reference with known concentration of a phosphorus compound the concentration of previously mentioned phosphorus metabolites in the liver may be calculated. Furthermore, we propose a ratio based on absolute concentrations called anabolic charge,
AC, somewhat similar to the ‘energy charge’ (EC) based on the concentrations of nucleotides
53.
Because high concentration of PME is usually vaguely referred to as in-
dicative of anabolic activities, and correspondingly those of PDE as catabolic the
simplest such definition of this dimensionless parameter would be
[PME] / ([PME] + [PDE]).
One advantage of such a parameter is that it obviates the need of separately discussing the PME and the PDE concentrations, and from a metabolic point of view in a
more consistent manner than the widely used uncorrected spectral resonance integral ratio.
The calculation of absolute metabolite concentrations and AC has been applied both
in Papers I and II.
18
DCE-MRI
3D imaging using rapid 3D gradient echo acquisition allowing complete liver coverage has become a commonly used technique for dynamic contrast enhanced (DCE)
MRI 54. Data from such examinations contain T1-weighted signal-time curves for each
voxel within the FoV, and it is commonly collected either as a dynamic multiphase
study, or as a perfusion-weighted study.
In the former usually non-enhanced, arterial, portal-venous and equilibrium phases
are acquired, but also late time phases may be included depending on type of contrast agent used. In order to retrieve as much relevant information as possible in the
initial phases after contrast administration perfusion studies are needed. They require high temporal resolution, typically in the range of 40 coronal images acquired
every 3-4 seconds.
Perfusion weighted studies apply tracer-kinetic modelling (derived from quantitative
nuclear medicine) and to calculate relevant parameters three signal- time curves are
needed: the signal-time curve in the tissue, the arterial input function (hepatic artery)
and the venous input function (portal vein) 54 55. For a standard extracellular contrast
agent a dual-input one-compartment model is sufficient and parameters such as absolute arterial blood flow(Fa in mL/min), absolute portal venous blood flow (Fp in
mL/min), arterial fraction (ART in percent= 100 x Fa /( Fa + Fp ), portal venous fraction
(PV in percent = 100-ART), distribution volume (in percent) and the mean transit
time, MTT, in seconds ( the average time it takes for a gadolinium molecule to pass
from the arterial or portal venous input to the venous output) may be calculated. 55 56
When introducing an intracellular contrast agent such as Gd-EOB-DTPA, an additional compartment is preferably added to the model 11 54 . After the first pass of contrast agent the concentration changes in the liver parenchyma occur more slowly,
which allows for a different imaging approach i.e. a dynamic multiphase study with
high-resolution 3D isotropic acquisitions of 20-30 seconds.
19
Data from DCE-MRI studies contain quantitative parameters as mentioned above. In
order to calculate concentrations of contrast agent in different compartments the signal time curves have to be converted to R1 relaxation time curves that can be interpreted as contrast concentrations via the relaxivity of the contrast agent in the imaged tissue, before a tracer-kinetic model is applied.
In Paper III, DCE-MRI data were extracted and the uptake rate of the contrast agent
in the hepatocytes was calculated using a simplified pharmacokinetic twocompartment model of the liver and spleen. From late time series (10 and 20 min
post-contrast) normalized liver-to-spleen contrast rations (LSC_N) where calculated.
In Paper IV, the excretion of contrast in the bile ducts was visually assessed utilizing
late time series (10-20-30 min) from the same data set as in Paper III.
20
AIMS OF THE STUDY
The overall aim of the study was to investigate two very different MR applications,
31
P Magnetic Resonance Spectroscopy,
P MRS, and Dynamic Contrast Enhanced
31
Magnetic Resonance Imaging, DCE-MRI, as non-invasive tools in the assessment of
histologically verified fibrosis in diffuse liver disease by using a quantitative approach.
P MRS
31
1. To implement a quantification method using a slice-selective pulse sequence
(DRESS) for absolute 31P liver metabolite concentrations, and to apply it in a
limited number of patients with histologically proven diffuse liver disease for
evaluation and comparison with other techniques and healthy control subjects.
2. To compare quantitative
P-MRS results in two distinct groups of patients
31
with histologically proven diffuse liver disorders including a control group
and to evaluate quantitative 31P-MRS as a potential diagnostic tool.
DCE-MRI
3. To use DCE-MRI to characterize hepatocyte function using late dynamic
phases in patients presenting with elevated liver enzymes, but without any
clinical signs of hepatic decompensation, and to prospectively and quantitatively compare the hepatocyte-specific uptake of Gd-EOB-DTPA with
histopathological fibrosis stage.
4. To correlate, in a prospective study, 1) the quantified uptake of Gd-EOBDTPA defined as KHep and LSC_N (10 and 20 min), 2) the histo-pathological
fibrosis scoring, and 3) results from the liver and renal blood tests with a visual assessment of the contrast elimination via the bile ducts.
21
MATERIALS AND METHODS
Localized In Vivo 31P NMR-Spectroscopy
The development and implementation of the technique is described in detail in In
vivo Quantification of absolute Liver Metabolite Concentrations by
31P
NMR Spectroscopy
57.
Data Acquisition
The examinations were performed on a 1.5 T MR-scanner (Signa LX Echospeed plus,
version 9.1, General Electric Medical Systems, Inc., Milwaukee, WI, U.S.A.). Both patients and control subjects were examined after 4 h of fasting in the right recumbent
position, with the liver close to the centre of the surface coil, in order to reduce respiration-related smearing artifacts. A portable ultrasound scanner (SonoSite 180 plus,
SonoSite Inc., Bothell, WA USA) was used in Paper II in order to facilitate positioning
of the spectroscopy coil. The examination time varied between 45 and 60 minutes.
Fig. 6 shows a typical set up of the coil and detection volume.
Fig. 6. Patient positioned on the right side. Location of the transmit and receive surface coils, the marker ring,
the external reference, and the selected in vivo slice at a depth of approximately 40 mm above the surface coil.
22
The body coil was used to obtain 1H-localizer images, 20 axial 10-mm slices across
the appropriate section of the subject.
31P-MRS
was then acquired using a flat, non-flexible (flat geometry) circular single
tuned surface coil (transmit 8 in., receive 5 in.; General Electric, Waukesha, Milwaukee, WI). Non-localized spectra of an external reference were obtained using a
(‘hard’) pulse-acquire sequence (FIDCSI; TR 2 s, 128 transients, 1024 data points were
used, 2500 Hz spectral width, dead-time 455 µs, total acquisition time was less than 5
minutes). Then a depth-resolved spectrum (DRESS) of liver tissue was acquired using a slice thickness of 30 mm (Paper I). (TR 2 s, 1024 transients, 1024 data points
were used, 2048 Hz spectral width, dead-time 2.76 ms, total acquisition time was c.
35 min). The transmitter frequency was placed just downfield of the in vivo resonances to avoid any possible acquisition artifacts within the range of interest. For more
details see Paper I.
External Referencing
The external reference (for quantification) consisted of a small sphere with a diameter
of 20.0 mm. The reference solution used contained 219.5 mM MeP in distilled water,
which resulted in a resonance around 33 ppm (using the conventional chemical shift
standard 85% H3PO4 as a reference assigned to 0.00 ppm; corresponding to –2.35
ppm for phosphocreatine, and about –9.84 ppm for Mg-ATP). The reference was
placed underneath the surface coil (at a fixed location c. 40 mm below the centre of
the upper coil surface), at a similar distance from the coil as the in vivo detection volume, but on the opposite side of the coil. The saturation corrective factors were determined as described in Paper I.
Processing
In the past the most common approach for interpretation of spectra has been to perform the data analysis in the frequency domain after a Fourier transformation of the
23
FID. However, distortions of the FT spectrum due to imperfections in the measured
FID can cause some problems and several steps such as baseline correction and phasing often have to be done manually thereby risking an operator dependent influence
of the results.
The difficulties regarding frequency domain analysis can be handled with less approximations and assumptions using time domain analysis. As the time domain
methods carry out all the operations directly on the measured FID, the raw data can
be analyzed without processing steps. An additional feature to time domain analysis
algorithms as VARPRO (Paper I) and AMARES (Paper II) is the possibility to incorporate biochemical prior knowledge to the fitting routine, see Fig 7.
Fig 7. Screen dump of the result from a VARPRO fit of 31P MRS data obtained in human liver (FIDCSI, GE
Signa Horizon, 1.5 T, MR-unit Linköping University Hospital). The display shows the FT spectra of, A) the
original data, B) the reconstructed best fit, C) the individual fitted Lorentzians, and D) the residual spectrum.
MRUI for Java, jMRUI, ('Magnetic Resonance User Interface' MRUI, EC Human Capital and Mobility Networks, France) 58 was used for processing of the in vivo MR data
24
in the time domain incorporating prior knowledge using the AMARES algorithm
59
and the VARPRO 60.
The following metabolite assignments were used 61 62 : The phosphomonoester (PME)
resonance was assigned to phosphoethanolamine (PEth) and phosphocholine (PCho);
inorganic phosphate (Pi) was defined as a single resonance. The phosphodiester resonance
(PDE)
was
assigned
to
glycerophosphoethanolamine
(GPEth),
glycerophosphocholine (GPCho), and unspecified 'membrane phospholipids' (MP)
around 0.0 ppm
63 64 65.
MP was included in the PDE resonance due to insufficient
spectral resolution, in the absence of proton decoupling. The ATP resonances were
assigned and interpreted as previously (Paper I).
In addition, resonances for NAD(H) and in some cases also a couple of additional
resonances corresponding to UDPG (UI and UII) were used to improve the accuracy
of the fits. For more details about spectral analysis, see (Paper I).
The pH values were determined as previously (Paper I); the particular parameters
included in the pH calculation were: pKa = 6.75; c1 = 10.84 ppm; c2= 13.20 ppm (these
shifts refer to the chemical shift difference between Pi and ATP).
Absolute Quantification of In Vivo Liver Metabolite Concentrations
The absolute quantification was performed as described in Paper I, except that in
Paper II a constant averaged value of reference signal amplitude was used in the
analysis. The reason for this modification in the procedure was the observation of a
slow degradation of the reference compound during the course of the present study.
25
Dynamic Contrast Enhanced MRI – DCE-MRI
Data Acquisition
A 1.5 T Achieva MRI (Philips Healthcare, Best, The Netherlands) was used together
with a phase-array body coil. Single breath-hold symmetrically sampled two-point
Dixon 3D images 66 were acquired with sensitivity encoding (SENSE).
All subjects received a bolus injection of Gd-EOB-DTPA (0.025 mmol/kg) administered intravenously at a rate of 1 mL/s using a power injector (Medrad Spectris Solaris, Pittsburgh, PA, USA) followed by a 30 mL saline flush. Image time-series were
acquired pre- and post-contrast agent injection (non-enhanced, arterial and venous
portal phase, 3, 10, 20 and 30 min post-injection).
The FOV and acquisition matrix were adjusted, if necessary, according to subject
size. During the initial contrast agent wash-in phase, the arterial portal phase, a higher temporal resolution was employed. The delayed and non-enhanced images were
captured using the following parameters: repetition time = 6.5 ms, echo time = 2.3
and 4.6 ms, flip angle = 13°, typical acquisition matrix = 168/168, typical FOV = 261 x
200 x 342 mm3, slice thickness = 4 mm, slice gap = –2 mm, typical imaging time = 20.2
s.
The acquired in- and opposite-phase images were reconstructed into water and fat
images using the inverse gradient method 67 68 69. In order to correct for intensity heterogeneity in water/fat Dixon images, and to gain reference scaling throughout the
time-series, the intensity of voxels identified as containing pure adipose tissue was
used as an internal reference. This correction was performed using the multi scale
adaptive normalizing averaging (MANA) method 70 71 72.
26
Image Analysis
Data were acquired from regions of interest (ROIs) placed in the liver (n = 7) and the
spleen (n = 3), see Fig 8. The ROIs were placed at the same anatomical location
throughout the time series by an experienced radiologist avoiding focal lesions, large
vessels and bile ducts without the intention of strictly following the segmental division as introduced by Couinaud
73.
The radiologist was blinded to the results of the
histopathological findings.
Fig. 8. Example of placement of 7 ROIs within the liver (3 in the left liver lobe, and 4 in the right liver lobe)
and 3 ROIs in the spleen, in descending order. The images were captured at 20 min post-contrast injection.
The dashed outlines of ROIs found in a few of the images show the location of other ROIs in nearby slices,
e.g. in the image showing the placement of “Liver R3”, the outline of the placement of “Spleen 3” can be
seen, and vice versa.
Quantitative Measurements of Gd-EOB-DTPA Uptake
Region of interest signal intensities were normalized and recalculated to relaxation
rate values, R1, according to
74.
Based on the normalized SI, normalized liver-to27
spleen contrast rations (LSC_N) where calculated at 10 (LSC_N10) and 20 (LSC_N20)
minutes post-contrast, as described in 74.
The contrast agent uptake rate (KHep) was calculated by fitting a simplified pharmacokinetic two-compartment model of the liver and spleen (Fig 9), recently described
by Dahlqvist Leinhard et al
74
, to the R1 time-series using a least-squares regression
algorithm.
LIV ER
SPLEEN
EES
Vascular
plasma
space
C
C
EES
C Contras t agent
C
C
C
C
C
C
Hepat oc yte
C
C
C
C
C
C
C
C
C
Bile ductles
C
C
C
C
C
C
C
Intracellul ar
space
C
C
C
C
C
C
C
C
C
C
C
Cv
C
C
C
C
Fig. 9. Example of a two compartment pharmco-kinetic model using Gd-EOB-DTPA. The contrast moves freely
between the vascular plasma space and the extracellular extravascular space, EES. Parameters such as blood flow,
permeability across the fenestrated sinusoidal membranes and space of Disse and access to transport proteins is
responsible for the contrast uptake into the hepatocytes.
The splenic tissue was represented by one compartment consisting of splenic blood
and extracellular extravascular (EES) being exposed to the contrast agent and one
intracellular compartment inaccessible to the contrast agent. The compartments in
the liver consisted of one blood and EES compartment similar to the spleen, and one
hepatobiliary (including small bile ductless) accessible to the contrast agent 74. By relating the hepatic contrast agent concentration to the blood and EES concentration as
28
observed in the spleen, quantitative measurements of the late hepatic uptake was
achieved.
For comparison, the liver spleen contrast (LSC) ratio was determined according to
Motosugi et al
75
at 10 and 20 min post-contrast agent administration (LSC10 and
LSC20, respectively).
29
Visual Assessment of Gd-EOB-DTPA Excretion
In Paper IV, the Gd-EOB-DTPA excretion was visually assessed and reformatted axial images were evaluated for the three time series per patient (10-20-30 min) as illustrated in Fig 10. The 10, 20 and 30 min post-contrast image time-series for all 29 patients were reviewed randomly and the visual assessment of bile duct excretion of
Gd-EOB-DTPA was based on consensus reading performed by two experienced radiologists.
Fig 10. Example of Gd-EOB-DTPA excretion in the bile ducts in the three time series in one F0 patient with the
diagnosis of AIH, panel a-c, and two F4 patients with the diagnosis of AIH and NAFLD, panels d-f and g-i respectively. In the F0 patient, intrahepatic contrast agent is observed at 10 min, indicated with an arrow in image
a. A reduced amount of contrast in central intra hepatic bile ducts was observed between 10 and 20 min. Contrast
in the CBD, as indicated with an arrow at 30 min in image c, was observed for all time series (not shown). The
difference in contrast excretion between the two F4 patients may be noted – only poorly visible contrast at 30 min
in central intra hepatic bile ducts in patient d-f indicated with arrows in image f.
abc = AIH, def= AIH, ghi = NAFLD
30
Five anatomical regions (peripheral bile ducts in the left and right liver lobe, right
and left intra hepatic main branch and the extra hepatic bile ducts) for each timeseries were assessed and the presence of contrast agent was graded as 1 (“yes”) or 0
(“no”). The presence of Gd-EOB-DTPA in each anatomical region was after the reviewing process summarized on a four-grade scale; 3 = contrast visible at 10 min, 2 =
contrast visible at 20 min but not at 10 min, 1 = contrast visible at 30 min but not at
10–20 min, 0 = no contrast visible at 10–30 min.
Finally the five scores, one for each anatomical region, as well as a total visual score
obtained by adding the five separate scores, were related to the histo-pathological
findings, the quantitative contrast agent uptake parameters and blood tests.
Computer software
Matlab R2009b (The Math works Inc., MA, USA) was used for data analysis and model fitting. MeVisLab 2.1 (MeVis Medical Solutions AG, Bremen, Germany) was used
for ROI placement.
31
Subjects, Paper I-IV
In total (I-IV), 76 patients and 25 control subjects were examined (Fig. 11).
Fig 11. Distribution of patients and control subjects in study paper I – IV.
In Paper I, two groups were studied. The patient group included 9 subjects with disorders of a predominantly parenchymal origin, as well as disease of biliary origin
representing six different diagnoses. The control group consisted of 12 healthy individuals without evidence of liver disease, malignancy, alcohol abuse, or possible liver-toxic medication.
In Paper II three groups were studied, two patient groups and one group of 13
healthy individuals. The ‘cirrhosis group’ consisted of 16 patients with advanced fibrosis (stage > 3) and/or established cirrhosis. The second group, ‘NAFLD group’,
consisted of 13 patients who had no-to-moderate inflammatory changes and fibrosis
stages of 0-2. None of the controls had any history of acute or chronic liver disease
32
In Paper III 38 patients were studied prospectively. They were referred for evaluation of elevated serum alanine aminotransferase (ALT) and/or alkaline phosphatase
(ALP) levels. Physical examination and laboratory tests revealed no signs of liver cirrhosis. Five patients were symptomatic (fatigue n=1, episodes of cholangitis n=1,
jaundice n=3) while the remaining 33 were asymptomatic.
In Paper IV, 29 patients from study III who had complete late DCE-MRI time series
were included. Three patients were symptomatic (episodes of cholangitis n =1, jaundice n=2) while the remaining 26 were asymptomatic.
Patients and control subjects participated after their informed consent had been obtained. The studies were approved by the regional ethics committee in Linköping,
Sweden, registration numbers 98-070, M74-05 and M72-07 T5-08.
Clinical Data
Laboratory Analysis
In Paper I blood samples for measurement of liver function were performed on the
patients.
In Papers II, III and IV, extensive laboratory tests were performed on the patients
including liver function tests and tests for viral, autoimmune and metabolic liver disease as well as a broad laboratory panel including glucose, insuline, lipids, iron, autoimmune antibodies, anti-HCV and HbsAg. Child–Pugh grading was calculated for
all patients in Paper II.
Routine blood laboratory parameters including liver function tests were performed
on the control group in Paper II, while no liver function tests were performed on the
control group in Paper I.
33
Liver Biopsy and Histopathological Grading
Liver biopsies were obtained on all patients. All biopsies were performed on an outpatient basis using a 1.6 mm Biopince needle (Medical Device Technologies Inc., FL,
USA). The histopathologists were blinded to the results of the 31P-MRS and DCE-MRI
data.
In Paper I liver biopsies had been obtained at a median time of 117 months (interquartile range, IQR, 50 – 157) prior to the MR examination apart from one patient
where the biopsy was performed two months after the MR examination. In Paper II
liver biopsies were obtained at a median time of 31 (IQR 5 - 89) months in cirrhotic
patients and 4 (2- 5) months in the NAFLD groups. In Papers III and IV liver biopsies were performed on the same day and immediately after the MR-examination.
In Paper I the results were graded semi-quantitatively from 0 to 3 (0 meaning normal) for steatosis, inflammation and fibrosis by an experienced hepatologist unaware
of the results from the MRS study. Presence of cirrhosis was dichotomized as ‘yes’ or
‘no’.
In Paper II the degree of steatosis was graded 0–3 based on a representative area of
liver tissue that was occupied by fat vacuoles 76. All liver biopsies were re-evaluated
by an experienced histopathologist (L.F.) with respect to inflammatory activity and
fibrosis stage, and classified according to the Batts and Ludwig system 27. Inflammatory activity was semi quantitatively staged: none (0), minimal (1), slight (2), moderate (3) and severe (4).
In Papers III and IV, biopsies were obtained in order to assess the histological severity of an underlying liver disease, to confirm the plausible diagnosis or to elucidate
the reason for the elevated liver enzymes if prior diagnostic work-up was negative.
Biopsies were graded and classified according to the Batts and Ludwig system
read by the same liver pathologist as part of the clinical routine.
34
27
and
An overview of demographic variables, histopathology and lab tests are provided in
Table 3. For details see original articles.
Table 3. Overview of demographic variables, histopathology and laboratory tests
Paper I
Paper II
Paper III and IV
Patients
Controls
Cirrhosis
NAFLD
Controls
Patients
9
12
16
13
13
38
M:3.W: 6
M:9 W:3
M:8 W:8
M:9W.4
M:6W:7
M: 21W:17
Age
M:58.0
51.1
M:57.8
M:61.7
M:42.5
M: median 45
(average or median)
W:58.8
W:66.8
W: 71.8
W:50.7
W: median 55
Biopsy in relation to
Median
Median 31
Median 4
_
Same day
months
months
prior to
prior to
MR
MR
Batts and
Batts and
Ludwig
Ludwig
F0 - 4
F0 - 4
Yes
Yes
Number and gender
MR examination
_
117
months
prior to
after MR
MR
Histopathological
fibrosis scoring
Semi-
_
quantit..
0-3
Blood laboratory
Yes
No
tests at the time of
MR examination
35
_
Batts and Ludwig
F0 - 4
Yes
Yes
Statistical Analysis
Data are generally presented as means ± standard deviation. Overall, a p value less
than 0.05 was considered significant.
In Paper I, comparisons between groups were made using a non-parametric test,
(Mann-Whitney U test). Spectroscopic data, histopathological findings and liver
function tests were correlated with Spearman rank correlation, rho.
In Paper II, Kruskall-Wallis test to evaluate differences between the three groups and
Mann-Whitney U test for comparisons between two groups in order to localize the
significant differences were used. In addition, spectroscopic data, histopathological
findings and liver function tests were correlated with Spearman rank order correlation coefficient (rho). Logistic regression and Fischer´s exact test were used when
comparing histopathological grading of fibrosis with MRS data. For calculations,
Statistica (StatSoft Inc, Tulsa, OK, U.S.A.) and JMP software (SAS Institute, Cary, NC,
U.S.A.) were used.
In Paper III groups were compared using the unpaired Wilcoxon test, and receiveroperating characteristic (ROC) analysis was performed with Stata 12.0 (StataCorp,
College Station, TX, USA), using non-parametric calculation of the area under the
ROC curve (AUROC) with its 95% confidence limits.
In Paper IV statistical analysis was performed using Stata 12.0 (StataCorp, College
Station, TX, USA). Spearman rank correlation was used to compare visual assessment of bile duct excretion of Gd-EOB-DTPA vs. histo-pathological grading of fibrosis, described by the fibrosis score (F0–F4) as well as by a binary variable indicating
the presence or absence of advanced fibrosis or cirrhosis (F3–4), dynamic contrast
enhancement parameters and liver and renal blood tests. The diagnostic ability of
visually assessed bile duct excretion of Gd-EOB-DTPA with respect to the presence
36
or absence of advanced fibrosis (F3-4) was determined by calculating the area under
the receiver operating characteristic curve (AUROC) with a 95% confidence interval.
37
RESULTS
Localized In Vivo 31P NMR Spectroscopy, Paper I-II
Concentrations Determined Using MRS
As demonstrated in Paper I, the patients had significantly lower concentrations of
PDE (p< 0.05) and ATP- (p< 0.05) compared with the control group. They also had
higher concentrations of PME, although this difference was not significant. The absolute concentrations are somewhat higher than typically reported in the literature using the often assumed liver ATP concentration of 2.5 mM.
In Paper II we also found a significantly lower concentration of PDE when comparing the control and cirrhosis groups (p = 0.025) and the intermediate, ‘NAFLD’, and
cirrhosis group (p < 0.01). There was also a tendency towards increase in PME and
decrease in ATP-, although this was not significant. (Fig. 12).
Fig 12. Absolute concentrations (mM) of PME, Pi, ATP- and PDE in paper I and II. Error bars indicate
S.D. Asterisks denote significant difference (p< 0.05).
38
For comparison the concentrations calculated in both studies are presented in Table 4
together with results from previous studies (see lower section in Table 4).
Table 4. 31P MRS concentrations (in mMSD) Papers I and II
PRESENT STUDIES
n
PME
Pi
PDE
ßATP
AC
pH
P-MRS I
31
Patients
9
2.24  0.86 2.04  0.66
6.31  3.91
3.55  1.05
0.29
7.370.10
Controls
12
1.7  0.65 2.25  0.41
10.014.21
4.19  0.32
0.16
7.450.12
Patients
9
1.340.51
1.220.40
3.772.33
2.120.62
Controls
12
1.010.39
1.350.25
5.972.51
2.500.19
NAFLD
13
4.031.10
2.350.52
10.971.66
4.360.76
0.27
7.440.09
Adv. fibrosis/cirrhosis
16
3.971.05
2.230.56
9.082.39
3.850.75
0.31
7.450.07
Controls
13
3.770.68
2.350.53
10.851.71
4.290.64
0.26
7.480.09
9
2.8
1.7
9.9
3.6
1.62-1.46
8.44-13.89
2.5
2.4
2.8
7.6
1.6
1.8
5.0
12
3.3
1.9
8.4
NTP-2.9
5
0.7
1.8
3.5
1.6
16
1.1
1.8
4.5
2.5
21
0.8
2.2
5.3
-
Scaling to ATP- ß 2.5 mM
P-MRS II
31
PREVIOUSLY
REPORTED
CONCENTRATIONS
Tosner et al
(2001)77Controls
Sijens et al
(1998)78Controls
17
1.7-2.59
Li et al (1996)79
Controls
3.8
12
Conc scal -ATP 2.5mM
2.5
Buchli et al
(1994)80Controls
Rajanayagam et al (1992)81
Controls
Oberhaensli et al
(1990)82Controls
Meyerhof et al (1989)
Controls
83
39
MRS Concentrations Expressed as Anabolic Charge, AC
In Paper I, patients were characterized by larger AC values than control subjects,
more specifically 0.29 vs 0.16 (p = 0.006). In Paper II, findings were similar, with significantly higher AC in the cirrhosis group compared to both the control group (p =
0.001) and NAFLD group (p = 0.009).
Applying an alternative classification of fibrosis in Paper II, non-significant, F0-1,
versus significant fibrosis, F2-4, showed significant differences in PDE between stage
F0-1 and stage F4 and in AC between stage F0-1 and F2-3 (Figs. 13 A and B).
Fig 13. A Absolute mean concentrations (mM) of PME, Pi, PDE and ATP-ß in relation to biopsy
findings. Error bars indicate SD. Asterisks denote significant difference (p < 0.05) B Anabolic charge
in relation to biopsy findings. Error bars indicate SD. Asterisks denote significant difference (p < 0.05).
Furthermore, the relationship between fibrosis stage (F0-4) and 31P-MRS parameters
(PME, Pi, PDE, ATP-ß and pH) was tested with ordinal logistic regression. Significant
relationships with the stage of fibrosis were found for the concentrations of PDE (p =
0.003) and Pi (p = 0.016).
40
Using a PDE concentration of 10.5 mM as a cut-off value between fibrosis stage F0- 2
and F3-4 results in sensitivity of 81% and specificity of 69% for advanced fibrosis
(stage 3-4).
A PDE concentration of 10.5 mM also turned out to be the best cut-off value between
fibrosis stage F0-1 and F2-4. However, less favorable sensitivity and specificity figures were obtained, 72% and 63%, respectively (Table 5).
Table 5. Degree of fibrosis in relation to PDE concentrations
F0
F1
F2
F3
F4
Sum
PDE  10.5 mM
1
3
0
9
4
17
PDE 10.5 mM
4
3
2
3
0
12
Total
5
6
2
12
4
29
P = 0.0067 (Wilcoxon unpaired test) Sensitivity for F3-4: 13/16 = 81% (confidence interval 54 -96%). Sensitivity for F2-4: 13/18 = 72% (confidence interval 47-90%). Specificity for F3-4: 9/13 = 69% (confidence interval
39-91%) Specificity for F2-4: 7/11 = 63% (confidence interval 31-89%).
Including AC in the logistic regression (while excluding PME and PDE) gave a clearly significant result (p = 0.003). An AC cut-off value of 0.27 between fibrosis stage F02 and F3-4 showed a sensitivity of 93% and a specificity of 54%. The same cut-off
value turned out to be the best for separating between fibrosis stage F0-1 and F2-4.
The sensitivity and specificity figures were 89% and 55%, respectively (Table 6).
No significant relationship between the MRS data and the degree of steatosis or inflammation was found.
41
Table 6. Degree of fibrosis in relation to AC
F0
F1
F2
F3
F4
Sum
AC  0.27
2
3
1
11
4
21
AC  0.27
3
3
1
1
0
8
Total
5
6
2
12
4
29
P = 0.0068 (Wilcoxon unpaired test). Sensitivity for F3-4: 15/16 = 93 % (confidence interval 70-100%). Sensitivity for F2-4: 16/18 = 89 % (confidence interval 65-99%). Specificity for F3-4: 7/13 = 54 % (confidence interval 25-81%). Specificity for F2-4: 6/11 = 55 % (confidence interval 23-83%).
MRS Versus Laboratory Data
Correlation coefficients (rho) between the laboratory parameters and
31P-MRS
pa-
rameters were computed for all subjects in Paper I as well as Paper II (pooling the
three groups – controls, NAFLD and cirrhosis – together), and several significant correlations between laboratory- and spectroscopic parameters were found. Those findings that were significant within the patient groups are listed in Table 7.
Table 7. Correlation between MRS data and liver function tests in paper I and paper II
PME
Paper I
PDE
Paper II
Paper I
Pi
Paper II
Paper I
ATP-
Paper II
Paper I
Paper II
SUBJ
NAFLD CIRR
n=9
n = 13
-0.46
0.31
-0.30
-0.66*
-0.19
-0.59*
-0.53
0.17
0.70*
-0.20
0.19
-0.56*
-
0.08
-0.39
-
0.06
-0.59*
-
-0.58*
-0.46
-
-0.41
-0.57*
ASAT
-0.01
0.44
-0.29
0.1
-0.06
-0.22
0.07
0.11
-0.28
0.1
0.28
-0.19
ALAT
0.3
0.12
0.01
0.31
-0.08
0.26
0.23
0.33
0.33
0.08
0.33
0.20
ALP
-
0.17
-0.49*
-
0.07
-0.08
-
0.45
0.07
-
0.03
-0.64*
ALB
0.13
-0.16
-0.15
0.39
0.06
0.25
0.03
-0.44
0.59*
-0.14
-0.13
0.09
BIL. TOT
PK-INR
SUBJ
n= 16 n = 9
NAFLD
CIRR SUBJ
NAFLD
CIRR
n = 13
n= 16 n = 9
n = 13
n = 16 n = 9 n = 13
*p < 0.05, Spearman rank correlation
42
SUBJ NAFLD CIRR
n= 16
Dynamic Contrast Enhanced MRI, Paper III-IV
Final Diagnosis and Fibrosis Scoring
In Paper III histopathological evaluation confirmed the clinical diagnoses that had
been set before liver biopsy in 35 of 38 patients. The final diagnosis and fibrosis scoring are demonstrated in Table 8. In the remaining 3 subjects liver biopsy was normal
in 2 subjects and showed a preliminary stage of autoimmune hepatitis in 1 subject.
In the patient with suspected thioguanine-induced liver injury, there were no
histopathological signs of such injury; the reason for elevated ALT was NAFLD.
Table 8. Final diagnosis and fibrosis scoring
Final diagnosis
F0
Normal
2
NAFLD
5
F1
F2
F3
3
2
PSC
2
2
4
HCV
4
2
1
3
2
1
AIH
1
PBC
1
AAT-deficiency
Total no of subjects
F4
1
1
1
8
9
10
7
4
Data are presented as number of subjects. NAFLD, Non alcoholic fatty liver disease; PSC, Primary
sclerosing cholangitis; HCV, Hepatitis C virusinfection; AIH, Autoimmune hepatitis; PBC, Primary Biliary
Cirrhosis; AAT-deficiency, α1-antitrypsin deficiency.
Pharmacokinetic Uptake Parameters versus Fibrosis Stage
Mean and standard deviation of normalized SI in the liver and spleenic ROIs using
Gd-EOB-DTPA, separated according to fibrosis score, are shown in Fig. 14.
43
Fig 14. Normalised signal intensity (SI) for liver and spleenic ROIs
according to fibrosis stage. The error bars represent one standard
deviation.
KHep, LSC_N10, LSC_N20 LSC10, and LSC20 separated according to fibrosis stage
are shown in Table 9. The 20 min post-contrast acquisition was missing in 2 patients.
44
Table 9. Pharmacokinetic parameters and fibrosis stage,
Pharmacokinetic
No/moderate fibrosis
Advanced fibrosis
parameter
F0
F1
F2
F3
F4
KHep
0.39±0.21
0.42±0.26
0.48±0.30
0.35±0.29
0.19±0.15
LSC_N10
1.40±0.10
1.42±0.14
1.40±0.13
1.30±0.12
1.15±0.14
LSC_N20
1.60±0.13
1.54±0.13
1.54±0.21
1.44±0.12
1.25±0.19
LSC10
1.69±0.18
1.71±0.16
1.57±0.23
1.63±0.18
1.28±0.20
LSC20
1.93±0.28
1.85±0.17
1.73±0.31
1.81±0.17
1.32±0.19
Biopsies were grouped into two groups: no and moderate fibrosis (F0-2, n=27) and
advanced fibrosis (F3-4, n=11). After calculating the AUROC values to discriminate
between fibrosis grade F3-4 and F0-2, the highest values were found for LSC_N10,
LSC_N20 and KHep. Both normalized ratios, LSC_N10 and LSC_N20, showed a clear
significance in discriminating between the two fibrosis groups while KHep demonstrated a borderline significance. No significant group differences were found for
LSC10 and LSC20 (Table 10).
Table 10. Discrimination between fibrosis grade F0-2 and F3-4
F0- 2
F3-4
p-value
AUROC
95 % CI
KHep
0.44±0.26a
0.29±0.25c
0.050*
0.71
0.503-0.912
LSC_N10
1.41±0.12 a
1.24±0.14 c
0.004**
0.80
0.640-0.969
LSC_N20
1.56±0.16b
1.38±0.16d
0.010*
0.78
0.621-0.940
LSC10
1.65±0.20 a
1.50±0.25 c
0.099
0.68
0.483-0.877
LSC20
1.82±0.27 b
1.66±0.29 d
0.223
0.64
0.424-0.845
Mean and standard deviation for the estimated hepatocyte uptake rate and ratios grouped by fibrosis stage (F0-2
and F3-4). Calculated AUROC values and p values for group differences. an = 27, bn = 26, cn = 11, dn = 10.
Unpaired Wilcoxon test with significance level *p<0.05, ** p<0.01 was used.
45
Calculating the AUROC values for KHep, LSC_N10 and LSC_N20 to discriminate
between fibrosis grade F0-1 and F2-4 yielded 0.63, 0.65 and 0.66 respectively. No significant group differences were found.
Visually Assessed Contrast Excretion versus Contrast Uptake Parameters, Histopathology and Blood Tests
In Paper IV no significant correlation between visual assessments of bile ducts excretion of Gd-EOB-DTPA and histo-pathological grading of fibrosis or the quantified
uptake of Gd-EOB-DTPA defined as KHep and LSC_N was found. The diagnostic
ability of visually assessed bile duct excretion of contrast is shown in Table 11.
Table 11. Diagnostic ability of visually assessed bile duct excretion of Gd-EOB-DTPA with
respect to the presence or absence of advanced fibrosis (F3-4).
AUROC
95% confidence interval
Extra-hepatic bile duct
0.67
0.44-0.90
Main duct right liver lobe
0.56
0.32-0.80
Segmental branches right liver lobe
0.51
0.30-0.74
Main duct left liver
0.54
0.29-0.79
Segmental branches left liver lobe
0.55
0.31-0.79
Total visual score
0.54
0.23-0.86
Area under the receiver operating characteristic curve (AUROC) with 95% confidence interval
As regards correlation between visual assessment and liver and renal blood tests,
significant negative correlations were found between ALP and the visual assessment
of bile duct excretion in several anatomic regions, (Table 12).
46
Table 12. Visual assessment of bile duct excretion of Gd-EOB-DTPA vs. liver and renal
blood tests
Extra he-
Main duct
Segmental
patic bile
right liver
branches
duct
lobe
right liver
Main duct
Segmental
left liver lobe branches left
Total visual score
liver lobe
lobe
Bilirubin
0.19
0.21
0.18
0.20
-0.20
0.13
Albumin
0.07
0.09
0.04
0.06
0.31
0.14
AST
-0.06
-0.02
-0.16
-0.08
-0.20
-0.12
ALT
0.05
0.12
-0.10
0.13
-0.24
-0.06
ALP
-0.42*
-0.47*
-0.36
-0.42*
-0.20
-0.43*
GGT
-0.18
-0.20
-0.24
-0.22
-0.08
-0.19
Kreatinin
0.18
0.21
0.03
0.22
0.01
0.14
Spearman correlation (ρ).
47
DISCUSSION
Due to its anatomical position, and diverse biochemical and physiological properties,
the liver is well suited in the efforts to develop robust and clinical applicable methods based on a multimodal MR approach. The present studies focus on the quantitative assessment of fibrosis from a biochemical and functional point of view, i.e. 31PMRS and DCE-MRI.
31
P-MR Spectroscopy
The absolute quantification technique of individual metabolite concentrations applied in the pilot study (I) demonstrated that patients had significantly lower concentrations of PDE and ATP- compared with the control group and also higher concentrations of PME, although not significant. A metabolic ratio for ex PME/PDE may fail
to detect which component(s) is involved in the observed changes. The introduced
parameter anabolic charge, AC, based on absolute concentrations and taking concentration changes in PME as well as PDE into consideration, was significantly higher in
the patient group.
The main finding when applying the technique in two distinct groups of patients (II),
“NAFLD” and “cirrhosis”, was that both PDE concentration and AC correlated to the
degree of fibrosis and that a decrease in PDE concentration parallels the presence of
fibrosis. There was also a tendency, although not statistically significant, towards
decrease in ATP (in the cirrhosis group.
Corbin et al
46
applied a similar absolute quantification technique on patients repre-
senting compensated or decompensated cirrhosis. Their results indicated that ATP
and PDE levels are reduced in patients with decompensated cirrhosis, and they also
a found significantly higher PME/PDE ratio (based on absolute concentrations) in the
48
decompensated group. A study by Dezortova et al
84
, comparing 31P-MRS data with
clinical but not with histological data, support the idea that changes in PDE concentrations reflects the degree of liver dysfunction.
In relation to clinical assessment;
An interpreted elevation of PME (deduced from various spectral ratios), related to
functional impairment based on Child-Pugh’s score, has been described
85 86 87 88.
However, we were not able to confirm a significant elevation of PME concentrations.
When applying the Child-Pugh scores system on the ‘cirrhosis‘group we found a
negative correlation between PDE concentrations’ (rho = -0.58, p = 0.02). Results from
other studies are divergent, Dezortova et al
Corbin et al
46
84
agrees with this observation, while
found significantly lower levels of ATP and higher PME/PDE ratio in
patients with decompensated cirrhosis without any obvious correlation between metabolite concentrations and the Child Pugh score.
In relation to histopathological findings;
A limited number of reports have correlated metabolite concentrations or ratios with
histopathological findings. Kiyono et al
89
found no correlation between spectra and
histopathological grading (categorizing subjects into chronic hepatitis or cirrhosis) in
patients with chronic viral hepatitis or autoimmune hepatitis. Van Wassener et al
90
showed a correlation between PME/P (‘P’ was defined as the sum of the concentrations of the measured phosphagens) and histological intralobular degeneration/focal
necrosis, portal inflammation and piecemeal necrosis in various diffuse liver diseases. Angus et al 91 studied patients with alcoholic liver and found correlation between
PME/Pi and PME/ATP and a histologically defined hepatitis group.
In a study by Lim et al 92 of patients with hepatitis C it was shown that 31P-MRS was
able to separate mild from moderate disease and these two groups from cirrhosis (the
subdivision into mild and moderate disease based on Ishak fibrosis and
necroinflammatory scoring system). There was an increase in the PME/PDE ratio cor49
responding to the degree of hepatitis. A ratio of 0.2 denoted mild hepatitis and a ratio
equal to or greater than 0.3 denoted cirrhosis with a sensitivity and specificity of 80%.
In a recent study Godfrey et al
93
report no significant correlation between fibrosis
stage and 31P-MRS. However, this study used the conventional PME/PDE metabolic
ratio preventing a closer analysis of individual components.
In order for the method to be used in a clinical context, threshold values separating
groups from each other need to be defined. Examining decompensated patients adds
important information with regard to disease mechanisms and biochemical alterations but is less valuable in clinical routine.
In Paper II we applied a simplified classification with regard to fibrosis, i.e. none –
mild (F0-2) and advanced fibrosis – cirrhosis (F3-4). According to Poynard et al 94 the
progression of fibrosis outside the portal tract is a threshold for progressive disease
corresponding to stage ≥ F2. Using this cut-off fibrosis stage a PDE concentration of
10.5 mM and an AC cut-off value of 0.27 turned out to be the best cut-off value for
separating between F2 and F3 as well as F1 and F2, however, slightly less favorable
sensitivity and specificity figures were obtained in the latter comparison. Calculating
the AUROC for AC resulted in 0.78 for the discrimination F1 vs F2 as well as for F2
vs F3.
The findings indicate that a decreased concentration of PDE is a marker of degree of
fibrosis and that AC is a potentially clinically useful parameter in the differentiation
of significant from non-significant fibrosis.
Our results indicate that the major changes are linked to PDE and Pi as well as ATP
concentration changes, in addition to a minor elevation of PME, but the biochemical
mechanisms behind these changes need further clarification.
It is reasonable to assume that reduced ATP levels are linked to the formation of fibrosis and a reduced viable hepatocyte mass, which has been demonstrated in ani50
mal models 47. The PME and PDE resonances are generally regarded as indicators of
cell membrane turn over, but it appears likely that a signal arising from endoplasmic
reticulum (ER) is a considerable contributor to the PDE peak. A reduced amount of
ER in hepatocytes from decompensated compared to compensated patients has been
shown 46
87
.
A number of methodological improvements may lead to more accurate estimates of
absolute concentrations of phosphorylated liver metabolites. As pointed out in Paper
II, a slow degradation of the external reference compound during the course of the
study was noted and a modification using a constant averaged value of reference
signal amplitude was applied in the analysis. Therefore avoiding external chemical
compounds would be favorable. One example of this is the ERETIC method (‘Electronic Reference To access In vivo Concentrations’) providing an electronic reference
signal
95
. Another improvement would be to apply proton decoupling of phospho-
rus metabolites. This would be highly advantageous in the PME and PDE region
where a considerable overlap of signals occurs. Furthermore, using a 3 T system instead of a 1.5 T means better SNR and improves the resolution of the spectrum. The
absolute quantification would be improved if the volume selection could be performed in a more selective manner, e.g. using a chemical shift imaging method. A
disadvantage of the CSI techniques, particularly in a clinical setting and with the
added complication of absolute quantification of metabolite concentrations is the effect of the point spread functions, which ‘smears’ the volume selection across several
spectra. This is not a severe limitation provided that enough phase-encoded spectral
voxels are acquired, however, inevitably at the cost of extended acquisition time.
51
Dynamic Contrast Enhanced MRI
Gd-EOB-DTPA behaves like a standard extracellular contrast agent in the early time
series with an increasing intracellular uptake in the hepatocytes starting within one
minute after injection. In order to quantify the intracellular uptake a pharmacokinetic
two-compartment model is desirable. A technique recently introduced (Dahlqvist
Leinhard et al 74) was used to calculate the contrast agent uptake rate in the hepatocytes, and normalized liver-to-spleen contrast rations (LSC_N) from late time series
(10 and 20 min) were studied.
Gd-EOB-DTPA is not metabolized in hepatocytes and therefore excreted in unchanged form into the bile. The visual image assessment of bile duct excretion in five
anatomical regions in late time series (10-20-30 min) was correlated to the contrast
uptake parameters.
The main finding in the prospective ‘uptake study’ (Paper III) was that liver fibrosis
stage strongly influences the hepatocyte-specific uptake of Gd-EOB-DTPA, but visually assessed biliary excretion of Gd-EOB-DTPA (Paper IV) could not establish a relationship between the fibrosis grade or contrast uptake parameters, expressed as
KHep or LSC_N. There are several reports, both animal and human studies, stating a
clear relationship between fibrosis stage and reduced contrast uptake 11 96 97 98 99 100.
There are likely several mechanisms responsible for the reduced uptake of contrast
agent in the hepatocytes. First, hepatocytes replaced by fibrous tissue yields a reduced number of hepatocytes that mediate transfer of the contrast agent from the
blood pool into the bile ductules. Second, the contrast agent uptake into the hepatocytes, mediated via a member of the organic anion transport proteins, might be competitively reduced owing to high concentrations of other metabolites, e.g. bilirubin
102.
101
Third, the excretion of Gd-EOB-DTPA is a rate-limiting step mediated by an ATP-
driven multi-resistance protein 103 104. Changes in the dual excretion pathways of GdEOB-DTPA may also contribute. Tamada et al
52
105
found a prolonged relative en-
hancement (RE) of the portal vein and a higher RE of the renal medulla in patients
with liver cirrhosis suggesting a compensatory increase in the renal excretion of GdEOB-DTPA.
The properties of Gd-EOB-DTPA lead us to a model-based approach, defining an
intracellular compartment. In order to calculate concentrations of a contrast agent in
different compartments the signal-time curves have be converted to R1 relaxation
time curves that can be interpreted as contrast agent concentrations via the relaxivity
of the contrast agent in the imaged tissue, before a pharmacokinetic model can be
applied. The present results also show, in agreement with a previous study
74
, that
the normalization procedure is an important step for removing patient and system
bias. The normalized LSC ratios used herein show a significant group difference,
which the simplistic SI-based LSC ratios fail to distinguish. The conversion of the
normalized SI to R1, used for contrast uptake rate – KHep – determination, is likely
to further reduce patient and system bias, yielding a robust approach to liver function determination. This finding needs to be confirmed, but in a recent report by Saito et al
11
the intracellular uptake rate (UR) was also studied. They used a dual-inlet
two compartment uptake model in cirrhotic patients and found a significant difference between the control groups and the Child-Pugh class A and B cirrhosis group
and between class A and B groups. The estimation of the intracellular contrast uptake offers an interesting tool for quantification of liver function in a clinical MR protocol. Evidence of the Gd-EOB-DTPA potential to measure liver function has also
been reported by Nilsson et al 12. They found that model-free parameters can quantitatively assess hepatic function and correlates with disease severity in patients with
PBC.
The excretion of non-metabolized Gd-EOB-DTPA, as well as other compounds such
as bilirubin, is mediated via MRP2 – a unidirectional ATP-dependent transporter.
MRP2 excretion has a limited rate yields retention of Gd-EOB-DTPA in the hepato53
cytes 103 104. There is no obvious explanation why we could not establish a relationship
between visually assessed biliary contrast excretion, fibrosis grade or contrast uptake
parameters expressed as KHep or LSC_N (IV). Tschirch et al
106
have reported a
threshold value for total bilirubin of ≥30 µmol/L for insufficient visualization of the
biliary tree for anatomical diagnosis 20 min after contrast administration in cirrhotic
patients. However, in the present study only one patient had a total bilirubin of ≥30
µmol/L.
In the time course of fibrosis development, there is a progressive loss of sinusoid endothelial fenestration preventing a free passage of molecules 15. In a study by Geetha
et al
107
increased oxidative stress has been found in red blood cells in cirrhotic pa-
tients. Oxidative stress induces MRP2 retrieval from the canalicular membranes and
causes cholestasis 51 .
Possibly the patients in the present study were ‘too healthy’ – all of them with grade
A according to the Child-Pugh scoring system. Only two patients had fibrosis stage
F4 and Fig 11 illustrates the variation in contrast excretion within this group. One
may speculate that the dominating feature in healthier patients is a disturbance in the
uptake mechanism while in clinically severe cases a shift towards a gradually more
impaired excretion due to severe oxidative stress predominates? Primarily cholestatic
diseases such as PSC and PBC will have to be excluded from this assumption since
these disorders primarily affect the excretion mechanism and decreased expression
of MRP2 is seen in PBC and PSC patients 108 109.
Interestingly enough, we found a negative correlation between the visual assessment
and ALP, elevated levels indicating intra-hepatic cholestasis. Additional statistical
analysis, comparing the sub-group of cholestatic patients (n = 8), assuming a connection between PSC and PBC and elevated ALP, with fibrosis score and Gd-EOB-DTPA
dynamic parameters did not add additional significant correlation.
54
The uptake study (Paper III) has some limitations. Although all patients were well
defined and liver biopsy was performed on the same day as the MR examination,
they had liver injury for different reasons and limited number of subjects included
precludes us from drawing any strong conclusion as to whether our results are applicable to different types of chronic liver disease. On the other hand, we believe that
the prospective approach applied may be an advantage as this reflects the clinical
routine. Second, sensitivity to spleen disease is incorporated into the method as it is
assumed that the blood and extracellular extravascular space (EES) compartment of
the spleen is similar to that of the liver. Third, in this study we do not present any
results from the early time series but focus on the late ones.
From a methodological point an increased flip angle (FA), e.g. from 10 to 30 degrees,
meaning a stronger T1 weighting and an improved SNR, and quantification of the
native time constant T1 for every subject instead of assuming fixed pre-contrast T1
values thereby improving the contrast concentration calculation would be beneficial.
It would also be desirable to ensure that the same voxels are analyzed throughout
the time series by applying a non-rigid registration method
110.
The design of the ‘excretion study’ (Paper IV) may benefit from using coronal reformats when analyzing the time series. Another improvement could be to use a fourpoint scale (‘no – mild – moderate – good’) when assessing the grade of contrast enhancement in bile ducts. In order to further investigate the hepatocyte excretion
pharmaco-kinetics of Gd-EOB-DTPA, it might also be useful to introduce a contrast
enhancement index based on increased temporal resolution in the late time series,
thereby achieving more detailed information concerning contrast timing and characteristics.
55
Clinical Significance
The long-term prognosis of chronic liver diseases largely depends on the extent and
the progression of liver fibrosis. For instance, in patients with chronic hepatitis C,
precise staging of hepatic fibrosis is paramount, as fibrosis is the most important
predictor of disease outcome and influences the indication for antiviral therapy.
Histopathological examination of liver biopsy, traditionally considered the gold
standard for evaluating hepatic fibrosis, has several drawbacks and uneven distribution of fibrosis in the liver parenchyma
2 3 4 5 6 7
raises questions to what extent a
tissue sample corresponding to about 1/50000 of the liver volume represents overall
findings in the liver. A broad spectrum of different non-invasive techniques has been
developed in order to compete with histopathological fibrosis staging. In Table 13,
examples of AUROC values from the literature are presented together with data
from the present study for comparison.
56
Table 13. AUROC from the literature and present study with regard to fibrosis discrimination
AUROC F≥F2
AUROC F≥F3
Fibrotest
Imbert-Bismuth 111 (2001)
0.74-0.89
Castera 112(2005)
0.85
0.90
Friedrich-Rust 113 (2008)
0.84
0.89
Ziol 114 (2005)
0.79
0.91
Castera112 (2005)
0.83
0.90
Foucher 115(2006)
0.75-0.84
0.86-0.93
Fibroscan
Contrast enhanced Ultrasound
Staub36 (2009)
0.85
Fibroscan +Fibrotest
Castera112 (2005)
0.88
0.95
0.78
0.78
0.53
0.50
0.63-0.66
0.71-0.80
P MRS AC
31
Norén (2008 II)
P MRS PME/PDE ratio
31
Godfrey93 (2012)
DCE-MRI
Norén (2012 Primovist) (III)
Chen99 (2012 Primovist)
Hagiwara
116
0.66
(2008 Magnevist)
0.79-0.82
MRE
Wang
117
(2012 Metanalysis)
Yin
118
(2007)
0.95
0.92
Huwart
119
(2008)
0.96
Godfrey
93
(2012)
0.78
0.90
DWI
Wang
117
(2012 Metanalysis)
Sandrasegeran
Lewin
121
(2009)
120
(2007)
0.86
0.66
0.79
0.92
57
MRE appears to be the superior technique in identifying significant fibrosis(stage ≥
F2) and is not influenced by the degree of steatosis
118
. While an MRE examination
generally includes sections in both the right and left liver lobes it presently does not
allow for a fibrosis scoring of each liver segment (Fig. 15).
The advantage of the DCE-MRI method described in this study is that it utilizes a
clinical MR protocol used in routine practice with the potential to evaluate the liver
function on a segmental as well as a global level. A relationship between liver SI after
Gd-EOB-DTPA injection and liver function measured with ICG R15 clearance has
been demonstrated75
122.
Fig 15. MRE University Hospital Linköping.NAFLD and cirrhosis. A kPa scal is provided.
The advantage of the DCE-MRI method described in this study is that it utilizes a
clinical MR protocol used in daily routine with the potential to evaluate the liver
function on a segmental as well as a global level. A relationship between liver SI after
Gd-EOB-DTPA injection and liver function measured with ICG R15 clearance has
been demonstrated 122.
A possible future scenario is to combine the result of the mechanical properties acquired using MRE with the functional aspects derived from quantitative DCE-MRI
analysis and adding biochemical information from 31P-MRS and DWI strengthening
the staging of fibrosis. Such a multimodal MR approach has the advantage of providing both global and regional information regarding fibrosis distribution, severity and
liver function. Compared to a conventional biopsy the added information is likely to
make it a powerful prognostic tool.
Evaluation of excessive accumulation of triglycerides and fatty acids in the hepatocytes and hepatic iron overload is also important information obtained by a conven58
tional biopsy. MR has proven to add relevant information also concerning these issues. T2* relaxation time have excellent sensitivity and specificity in the diagnosis
and presence of hepatic iron overload 123. 1H MRS is a sensitive method to detect hepatic triglyceride content and has shown a strong correlation with histopathological
steatosis 124.
The non-invasive assessment of the inflammatory component with MR is troublesome. First, there is no straightforward MR application that specifically deals with
the grade of inflammation although there are reports describing a weak to moderate
correlation between DWI and inflammation
125 126.
Second, a higher inflammation
grade tends to yield a higher liver stiffness compared to low-grade inflammation in
patients with fibrosis stages F1-3. In stage F4 the inflammation grade does not significantly affect the liver stiffness 127.
The MR technique also enables detailed morphological assessment, not only limited
to the liver parenchyma, but also covering the upper part of the abdomen. This is
important for the detection of early HCC in patients included in surveillance programs, as well as for the detection of complications such as portal hypertension and
gastroesophageal varices.
Nguyen et al
128
have proposed a general ‘non-invasive algorithm’ for patients with
hepatitis C, starting with Fibrotest, Fibroscan, or MRE. If indeterminate the next step
would be a liver biopsy.
In patients with suspected or manifest diffuse liver disease, a more attractive alternative would be to start with a multimodal MR-examination. Based on a protocol used
in clinical routine, the technique can offer a non-invasive way of minimizing the need
for conventional liver-biopsy, or potentially replacing it, by accurately staging liver
fibrosis and the intracellular lipid content. It is also a reliable technique for diagnosing hepatic iron overload. Contributing with valuable functional and high resolution
59
morphological information globally as well as on a segmental level, it adds properties that no other non-invasive technique presently matches.
This scenario is encouraging to support the clinical decision making and management of patients, which also takes into account a wide range of patient and disease
characteristics – all of them taken together reflecting the versatility of the liver.
A common feature of the non-invasive techniques, possibly with the exception of
quantitative texture analysis in double contrast enhanced MRI, is that the staging of
fibrosis is based on indirect measurements or surrogate serum markers. According to
Schuppan et al15 the ideal non-invasive marker should be liver specific and not affected by renal or reticuloendothelial function. It should also provide exact measurements of one or more of; stage of fibrosis, fibrogenesis and/or fibrolysis. Finally it
should be easy, reproducible and able to predict risk of disease progression or regression.
Molecular imaging of fibrosis using collagen-targeted MRI contrast may be a step in
that direction. Caravan et al
129
have evaluated a type I collagen-specific peptide con-
jugated to a gadolinium complex in a mouse model of chronic myocardial infarction.
High contrast for fibrotic scar versus viable myocardium was noted as well as a remaining enhancement of a collagen-rich liver after 40 min. Chow et al 130 investigated
a fibrin-fibronectin-targeted Gd contrast agent for early detection of liver fibrosis.
They also used a mouse model and found different contrast enhancement between
normal and fibrotic livers indicating that this contrast agent could be used to detect
and characterize liver fibrosis at an early stage.
Another interesting approach is the potential use of a model - or a combination of
models - for advanced decision support. Data retrieved from DCE-MRI time series
may be used in a mathematical mechanistic model describing hepatobiliary transfer
rates for Gd-EOB-DTPA. Taking the above mentioned parameters into account as
60
well as a Bayesian decision support model incorporating mechanistic data is a challenging future scenario.
61
CONCLUSIONS
C1
As making 31P-MRS a useful clinical tool requires a simple and straightforward procedure avoiding complicated methodological issues, Paper I has shown that the use
of a simple pulse sequence, fixed coil geometry and time domain spectral evaluation
can fulfill these demands and provide absolute metabolite concentrations.
C2
Application of a 31P-MRS method for absolute quantification of individual metabolite
concentrations may reliably reflect biochemical processes in the liver, which a metabolite ratio may fail to detect. Paper II indicates that PDE is a marker of degree of
fibrosis and that AC is a potentially clinically useful parameter in differing mild fibrosis from severe.
C3
A new quantification procedure for calculation of the hepatocyte-specific contrast
agent, Gd-EOB-DTPA, uptake applied on a two-compartment pharmacokinetic model (Paper III) shows promising results for achieving a non-invasive approach in separating no/mild from advanced fibrosis with the potential to assess liver function on a
segmental level.
C4
In the final study (Paper IV), visually assessed biliary excretion of Gd-EOB-DTPA
could not establish a relationship between the fibrosis grade or contrast uptake parameters expressed as KHep or LSC_N.
Future Aspect
Based on a clinically applicable protocol, a multimodal MR-approach has the potential to accurately stage liver fibrosis, the intracellular lipid content, hepatic iron overload and to add functional and morphological information, globally as well as on a
segmental level.
62
ACKNOWLEDGEMENTS
I wish to express my deep gratitude to all who have helped and supported me in this
project. In particular, I want to thank:
My supervisor Staffan Wirell, and co- supervisors Peter Lundberg and Sven Almer
for support, advice, enthusiastic encouragement, useful critiques and patient guidance throughout this interdisciplinary work.
Örjan Smedby for invaluable scientific guidance and support.
My close associates and co- authors in the research group; Olof Dahlqvist Leinhard,
Nils Dahlström, Mikael Forsgren, Johan Kihlberg and Stergios Kechagias for valuable and constructive discussions in planning, developing and performing this research work.
My additional co authors Markus Ressner, Mattias Ekstedt, Tobias Romu and Lennart Franzén for appreciated support and team work.
Anders Persson for providing an inspiring and unique research environment at
CMIV.
Sven-Göran Fransson for rapid and constructive feed back.
Pia Säfström for positive support and providing time for research.
My colleagues in the Body Section at the Dept. of Radiology Linköping University
Hospital Karin Heijl, Nina Kämmerling, Jenny Öman and Wolf Bartholomae for
taking on my clinical share while I was occupied with research tasks.
All MR technicians at CMIV and Dept of Radiology Linköping University Hospital
for their help and creating a pleasant working environment.
Virginia Westerberg and Ingela Allert for excellent administrative support and language control.
All volunteers and patients who generously contributed to the different studies.
Finally – my family for all your support!
63
REFERENCES
1.
Högnäs, S. Idéernas Historia - en översikt. Historisk Media (2003).
2.
Goldin, R. D. et al. Intra-observer and inter-observer variation in the
histopathological assessment of chronic viral hepatitis. J Hepatol 25, 649–54
(1996).
3.
Kleiner, D. E. et al. Design and validation of a histological scoring system for
nonalcoholic fatty liver disease. Hepatology 41, 1313–21 (2005).
4.
Bedossa, P., Dargère, D. & Paradis, V. Sampling variability of liver fibrosis in
chronic hepatitis C. Hepatology 38, 1449–57 (2003).
5.
Rousselet, M. C. et al. Sources of variability in histological scoring of chronic viral hepatitis. Hepatology 41, 257–64 (2005).
6.
Piccinino, F., Sagnelli, E., Pasquale, G. & Giusti, G. Complications following percutaneous liver biopsy. A multicentre retrospective study on 68,276 biopsies. J
Hepatol 2, 165–73 (1986).
7.
Regev, A. et al. Sampling error and intraobserver variation in liver biopsy in patients with chronic HCV infection. Am J Gastroenterol 97, 2614–8 (2002).
8.
Martínez, S. M., Crespo, G., Navasa, M. & Forns, X. Noninvasive assessment of
liver fibrosis. Hepatology 53, 325–35 (2011).
9.
Rockey, D. C. Current and future anti-fibrotic therapies for chronic liver disease.
Clin Liver Dis 12, 939–62, xi (2008).
10. Cohen-Naftaly, M. & Friedman, S. L. Current status of novel antifibrotic therapies in patients with chronic liver disease. Therap Adv Gastroenterol 4, 391–417
(2011).
11. Saito, K. et al. Assessing liver function using dynamic Gd-EOB-DTPA-enhanced
MRI with a standard 5-phase imaging protocol. J Magn Reson Imaging
(2012).doi:10.1002/jmri.23907
12. Nilsson, H., Blomqvist, L., Douglas, L., Nordell, A. & Jonas, E. Assessment of
liver function in primary biliary cirrhosis using Gd-EOB-DTPA-enhanced liver
MRI. HPB (Oxford) 12, 567–76 (2010).
13. Poynard, T. Assessment of liver fibrosis: Noninvasive means. The Saudi Journal of
Gastroenterology 14, 163–73 (2008).
14. Bataller, R. & Brenner, D. A. Liver fibrosis. J Clin Invest 115, 209–18 (2005).
15. Schuppan, D. & Afdhal, N. H. Liver cirrhosis. Lancet 371, 838–51 (2008).
16. Hemmann, S., Graf, J., Roderfeld, M. & Roeb, E. Expression of MMPs and TIMPs
in liver fibrosis - a systematic review with special emphasis on anti-fibrotic
strategies. J Hepatol 46, 955–75 (2007).
64
17. Younossi, Z. M. Practical management of liver diseases. Cambridge University
Press (2008).
18. Vårdprogram Levercells cancer. Nationellt vårdprogram (2011).
19. Morris-Stiff, G., Gomez, D. & Prasad, R. Quantitative assessment of hepatic
function and its relevance to the liver surgeon. J Gastrointest Surg 13, 374–85
(2009).
20. Lee, C-F et al. Using Indocyanine Green Test to Avoid Post-hepatectomy Liver
Dysfunction. Chang Gung Med J 30, 333–338 (2007).
21. Limdi, J. K. & Hyde, G. M. Evaluation of abnormal liver function tests. Postgrad
Med J 79, 307–12 (2003).
22. Kwon, A., Ha-Kawa, S., Uetsuji, S., Kamiyama, Y. & Tanaka, Y. Use of technetium 99m diethylenetriamine-pantaacetic acid-galactosylhuman serum albumin
liver scintigraphy in the evaluation of preoperative and postoperative hepatic
functional reserve for hepatectomy. Surgery 117, 429–434. (1995).
23. Pugh, R. N., Murray-Lyon, I. M., Dawson, J. L., Pietroni, M. C. & Williams, R.
Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 60,
646–9 (1973).
24. Wiesner, R. et al. Model for end-stage liver disease (MELD) and allocation of
donor livers. Gastroenterology 124, 91–6 (2003).
25. Durand, F. & Valla, D. Assessment of the prognosis of cirrhosis: Child-Pugh versus MELD. J Hepatol 42 Suppl, S100–7 (2005).
26. Kamath, P. S., Kim, W. R. & Group, A. L. D. S. The model for end-stage liver
disease (MELD). Hepatology 45, 797–805 (2007).
27. Batts, K. P. & Ludwig, J. Chronic hepatitis. An update on terminology and reporting. Am J Surg Pathol 19, 1409–17 (1995).
28. Bedossa, P. & Poynard, T. An algorithm for the grading of activity in chronic
hepatitis C. The METAVIR Cooperative Study Group. Hepatology 24, 289–93
(1996).
29. Ishak, K. et al. Histological grading and staging of chronic hepatitis. J Hepatol 22,
696–9 (1995).
30. Sebastiani, G. & Alberti, A. How far is noninvasive assessment of liver fibrosis
from replacing liver biopsy in hepatitis C? J Viral Hepat 19 Suppl 1, 18–32 (2012).
31. Rockey, D. C. Noninvasive assessment of liver fibrosis and portal hypertension
with transient elastography. Gastroenterology 134, 8–14 (2008).
32. Castéra, L. et al. Pitfalls of liver stiffness measurement: a 5-year prospective
study of 13,369 examinations. Hepatology 51, 828–35 (2010).
33. Blomley, M. J. et al. Liver microbubble transit time compared with histology and
Child-Pugh score in diffuse liver disease: a cross sectional study. Gut 52, 1188–93
(2003).
65
34. Lim, A. K. et al. Hepatic vein transit times using a microbubble agent can predict
disease severity non-invasively in patients with hepatitis C. Gut 54, 128–33
(2005).
35. Searle, J. et al. Non-invasive prediction of the degree of liver fibrosis in patients
with hepatitis C using an ultrasound contrast agent. A pilot study. J Med Imaging
Radiat Oncol 52, 130–3 (2008).
36. Staub, F. et al. Liver fibrosis staging with contrast-enhanced ultrasonography:
prospective multicenter study compared with METAVIR scoring. Eur Radiol 19,
1991–7 (2009).
37. Castéra, L. et al. Prospective comparison of transient elastography, Fibrotest,
APRI, and liver biopsy for the assessment of fibrosis in chronic hepatitis C. Gastroenterology 128, 343–50 (2005).
38. Van Beers, B. E. et al. Hepatic perfusion parameters in chronic liver disease: dynamic CT measurements correlated with disease severity. AJR Am J Roentgenol
176, 667–73 (2001).
39. Miles, K. A., Hayball, M. P. & Dixon, A. K. Functional images of hepatic perfusion obtained with dynamic CT. Radiology 188, 405–11 (1993).
40. Ronot, M. et al. Liver fibrosis in chronic hepatitis C virus infection: differentiating minimal from intermediate fibrosis with perfusion CT. Radiology 256, 135–42
(2010).
41. Gautam Bahl, I. C., Tanya Wolfson,, A. C. G., Julie M. Collins,, A. D. C., Fatma
Barakat,, T. H. & Claude B. Sirlin, Noninvasive Classification of Hepatic Fibrosis
Based on Texture Parameters From Double Contrast- Enhanced Magnetic Resonance Images. J Magn Reson Imaging 36, 1154–1161 (2012).
42. Aguirre, D. A., Behling, C. A., Alpert, E., Hassanein, T. I. & Sirlin, C. B. Liver
fibrosis: noninvasive diagnosis with double contrast material-enhanced MR imaging. Radiology 239, 425–37 (2006).
43. Taouli, B., Ehman, R. L. & Reeder, S. B. Advanced MRI methods for assessment
of chronic liver disease. AJR Am J Roentgenol 193, 14–27 (2009).
44. Bloch, F. Nuclear induction. Phys Rev 70, 460–474 (1946).
45. Purcell, E. Resonance absorption by nuclear magnetic moments in a solid. Phys
Rev 69, 37–38 (1946).
46. Corbin, I. R., Ryner, L. N., Singh, H. & Minuk, G. Y. Quantitative hepatic phosphorus-31 magnetic resonance spectroscopy in compensated and decompensated cirrhosis. Am J Physiol Gastrointest Liver Physiol 287, G379–84 (2004).
47. Lundberg, P., Harmsen, E., Ho, C. & Vogel, H. J. Nuclear magnetic resonance
studies of cellular metabolism. Anal Biochem 191, 193–222 (1990).
48. Taylor-Robinson, S. D. et al. In vivo and in vitro hepatic 31P magnetic resonance
spectroscopy and electron microscopy of the cirrhotic liver. Liver 17, 198–209
(1997).
66
49. Menon, D. K. et al. Effect of functional grade and etiology on in vivo hepatic
phosphorus-31 magnetic resonance spectroscopy in cirrhosis: biochemical basis
of spectral appearances. Hepatology 21, 417–27 (1995).
50. Hamm, B et al. Phase I clinical evaluation of Gd-EOB-DTPA as a hepatobiliary
MR contrast agent: safety, pharmacokinetics, and MR imaging. Radiology 195,
785–792 (1995).
51. Van Beers, B. E., Pastor, C. M. & Hussain, H. K. Primovist, Eovist: what to expect? J Hepatol 57, 421–9 (2012).
52. Trauner, M. & Boyer, J. L. Bile salt transporters: molecular characterization,
function, and regulation. Physiol Rev 83, 633–71 (2003).
53. Atkinson, D. E. The energy charge of the adenylate pool as a regulatory parameter. Interaction with feedback modifiers. Biochemistry 7, 4030–4 (1968).
54. Sommer, W. H. et al. Contrast agents as a biological marker in magnetic resonance imaging of the liver: conventional and new approaches. Abdom Imaging
37, 164–79 (2012).
55. Peters, A. M. Fundamentals of tracer kinetics for radiologists. Br J Radiol 71,
1116–29 (1998).
56. Taouli, B., Ehman, R. L. & Reeder, S. B. Advanced MRI methods for assessment
of chronic liver disease. AJR Am J Roentgenol 193, 14–27 (2009).
57. Ressner M. In vivo Quantification of absolute Liver Metabolite Concentrations
by 31P NMR Spectroscopy. M sc . Linköping University (1998).
58. Naressi, A. et al. Java-based graphical user interface for the MRUI quantitation
package. MAGMA 12, 141–52 (2001).
59. Vanhamme, L., Van den Boogaart A & Van Huffel S Improved method for accurate and efficient quantification of MRS data with use of prior knowledge. J
Magn Reson 129, 35–43 (1997).
60. Golub G. H., Pereyra V. The differentiation of pseudoinverses and nonlinear
least squares problems whose variables separate. SIAM J. Numer. Anal. 10, 413–
432 (1973).
61. Robotaille, P.-M., Robitaille, P. & Browns, G. An analysis of the pH-dependent
chemical shift behavios of phosphorus containing metabolites. J Magn Reson 92,
73–84 (1991).
62. Graham, R., Meyer, R., Szwergold, B. & Brown, T. Observation of myo-inositol
1,2- (cyclic) phosphate in a Morris hepatoma by 31P NMR. J Biol Chem 262, 35–37
(1987).
63. Bates, T. E., Williams, S. R. & Gadian, D. G. Phosphodiesters in the liver: the effect of field strength on the 31P signal. Magn Reson Med 12, 145–50 (1989).
67
64. De Kruijff, B., Van den Besselaar, A. M., Cullis, P. R., Van den Bosch, H. & Van
Deenen, L. L. Evidence for isotropic motion of phospholipids in liver microsomal membranes. A 31P NMR study. Biochim Biophys Acta 514, 1–8 (1978).
65. Murphy, E. J., Rajagopalan, B., Brindle, K. M. & Radda, G. K. Phospholipid bilayer contribution to 31P NMR spectra in vivo. Magn Reson Med 12, 282–9 (1989).
66. Dixon, W. T. Simple proton spectroscopic imaging. Radiology 153, 189–94 (1984).
67. Rydell, J. et al. Phase sensitive reconstruction for water/fat separation in MR
imaging using inverse gradient. Med Image Comput Comput Assist Interv 10, 210–8
(2007).
68. Rydell, J. et al. Three Dimensional Phase Sensitive Reconstruction for Water/Fat
Separation in MR Imaging using Inverse Gradient. In Proc of the Int Soc Mag
Reson in Med (ISMRM’08) 16, 1519 (2008).
69. Romu, T. et al. Fat Water Classification of Symmetrically Sampled Two-Point
Dixon Images Using Biased Partial Volume Effects. In Proc of the Int Soc Mag
Reson in Med (ISMRM’11) 19, 2711 (2011).
70. Dahlqvist Leinhard, O. et al. Quantitative Abdominal Fat Estimation using MRI.
In Proceedings of the 19th international conference of pattern recognition (ICPR ’ 08)
(2008).
71. Romu, T., Borga, M. & Dahlqvist, O. MANA - Multi scale adaptive normalized
averaging. In Proceedings in the IEEE international symposium on biomedical imaging; from nano to macro(ISBI’11) 361–364 (2011).
72. Andersson, T. et al. Selfcalibrated DCE-MRI using Scale Adaptive Normaluzed
Averaging (MANA. In Proc of the Int Soc Mag Reson in Med (ISMRM’12) 20, 2445
(2012).
73. Couinaud, C. Le foie; etudes anatomiques et chirurgicales. Paris, Masson (1957).
74. Dahlqvist Leinhard, O. et al. Quantifying differences in hepatic uptake of the
liver specific contrast agents Gd-EOB-DTPA and Gd-BOPTA: a pilot study. Eur
Radiol 22, 642–53 (2012).
75. Motosugi, U. et al. Liver parenchymal enhancement of hepatocyte-phase images
in Gd-EOB-DTPA-enhanced MR imaging: which biological markers of the liver
function affect the enhancement? J Magn Reson Imaging 30, 1042–6 (2009).
76. Mathiesen, U. L. Increased liver echogenicity at ultrasound examination reflects
degree of steatosis but not of fibrosis in asymptomatic patients with
mild/moderate abnormalities of liver transaminases. Digest Liver dis 34, 516–522
(2002).
77. Tosner, Z., Dezortová, M., Tintĕra, J. & Hájek, M. Application of twodimensional CSI for absolute quantification of phosphorus metabolites in the
human liver. MAGMA 13, 40–6 (2001).
68
78. Sijens, P. E. et al. Understanding the discrepancies between 31P MR spectroscopy assessed liver metabolite concentrations from different institutions. Magn
Reson Imaging 16, 205–11 (1998).
79. Li, C. W., Negendank, W. G., Murphy-Boesch, J., Padavic-Shaller, K. & Brown,
T. R. Molar quantitation of hepatic metabolites in vivo in proton-decoupled, nuclear Overhauser effect enhanced 31P NMR spectra localized by threedimensional chemical shift imaging. NMR Biomed 9, 141–55 (1996).
80. Buchli, R., Meier, D., Martin, E. & Boesiger, P. Assessment of absolute metabolite
concentrations in human tissue by 31P MRS in vivo. Part II: Muscle, liver, kidney. Magn Reson Med 32, 453–8 (1994).
81. Rajanayagam, V., Lee, R. R., Ackerman, Z., Bradley, W. G. & Ross, B. D. Quantitative P-31 MR spectroscopy of the liver in alcoholic cirrhosis. J Magn Reson Imaging 2, 183–90 (1992).
82. Oberhaensli, R., Rajagopalan, B., Galloway, G. J., Taylor, D. J. & Radda, G. K.
Study of human liver disease with P-31 magnetic resonance spectroscopy. Gut
31, 463–7 (1990).
83. Meyerhoff, D. J., Boska, M. D., Thomas, A. M. & Weiner, M. W. Alcoholic liver
disease: quantitative image-guided P-31 MR spectroscopy. Radiology 173, 393–
400 (1989).
84. Dezortova, M., Taimr, P., Skoch, A., Spicak, J. & Hajek, M. Etiology and functional status of liver cirrhosis by 31P MR spectroscopy. World J Gastroenterol 11,
6926–31 (2005).
85. Munakata, T. An in vivo 31P MRS study of patients with liver cirrhosis: progress
towards a non-invasive assessment of disease severity. NMR Biomed 6, 168–172
(1993).
86. Menon, D. K. et al. Effect of functional grade and etiology on in vivo hepatic
phosphorus-31 magnetic resonance spectroscopy in cirrhosis: biochemical basis
of spectral appearances. Hepatology 21, 417–27 (1995).
87. Taylor-Robinson, S. D. et al. In vivo and in vitro hepatic 31P magnetic resonance
spectroscopy and electron microscopy of the cirrhotic liver. Liver 17, 198–209
(1997).
88. Jalan, R. et al. Hepatic phosphorus-31 magnetic resonance spectroscopy in primary biliary cirrhosis and its relation to prognostic models. Gut 39, 141–6 (1996).
89. Kiyono, K. et al. Relationship of 31P MR spectroscopy to the histopathological
grading of chronic hepatitis and response to therapy. Acta Radiol 39, 309–14
(1998).
90. Van Wassenaer-van Hall, H. N. et al. 31P magnetic resonance spectroscopy of
the liver: correlation with standardized serum, clinical, and histological changes
in diffuse liver disease. Hepatology 21, 443–9 (1995).
69
91. Angus, P. W. et al. A study of patients with alcoholic liver disease by 31P nuclear
magnetic resonance spectroscopy. Clin Sci (Lond) 78, 33–8 (1990).
92. Lim, A. K. et al. The relationship of in vivo 31P MR spectroscopy to histology in
chronic hepatitis C. Hepatology 37, 788–94 (2003).
93. Godfrey, E. M. et al. Comparison of MR elastography and 31P MR spectroscopy
with histological staging of liver fibrosis. Eur Radiol DOI 10.1007/s00330-0122527-x, (2012).
94. Poynard, T. et al. Impact of pegylated interferon alfa-2b and ribavirin on liver
fibrosis in patients with chronic hepatitis C. Gastroenterology 122, 1303–13 (2002).
95. Akoka, S., Barantin, L. & Trierweiler, M. Concentration Measurement by Proton
NMR Using the ERETIC Method. Anal Chem 71, 2554–7 (1999).
96. Tsuda, N., Okada, M. & Murakami, T. New proposal for the staging of nonalcoholic steatohepatitis: evaluation of liver fibrosis on Gd-EOB-DTPA-enhanced
MRI. Eur J Radiol 73, 137–42 (2010).
97. Lee, W. et al. Quantitative evaluation of the hepatic parenchymal change in patients with chronic liver disease using Gd-EOB-DTPA-enhanced MRI: Comparison with normal liver. J Korean Soc Radiol 64, 49–55 (2011).
98. Watanabe, H. et al. Staging hepatic fibrosis: comparison of gadoxetate disodiumenhanced and diffusion-weighted MR imaging--preliminary observations. Radiology 259, 142–50 (2011).
99. Chen, B. B. et al. Dynamic contrast-enhanced magnetic resonance imaging with
Gd-EOB-DTPA for the evaluation of liver fibrosis in chronic hepatitis patients.
Eur Radiol 22, 171–80 (2012).
100. Goshima, S. et al. Gd-EOB-DTP-Enhanced MR imaging: Prediction of hepatic
fibrosis stages using liver contrast enhancement index and liver-to-spleen volumetric ratio. JMRI (2012).doi:10.1002/jmri.23758
101. Leonhardt, M. et al. Hepatic uptake of the magnetic resonance imaging contrast
agent Gd-EOB-DTPA: role of human organic anion transporters. Drug Metab
Dispos 38, 1024–8 (2010).
102. Narita, M. et al. Expression of OATP1B3 determines uptake of Gd-EOB-DTPA in
hepatocellular carcinoma. J Gastroenterol 44, 793–8 (2009).
103. Muhler, A. Complete elimination of the hepatobiliary MR contrast agent GdEOB-DTPA in hepatic dysfunction: An experimental study using transportdeficient, mutant rats. MAGMA 1, 134–139 (1993).
104. König, J., Nies, A. T., Cui, Y., Leier, I. & Keppler, D. Conjugate export pumps of
the multidrug resistance protein (MRP) family: localization, substrate specificity,
and MRP2-mediated drug resistance. Biochim Biophys Acta 1461, 377–94 (1999).
105. Tamada, T. et al. Gd-EOB-DTPA enhanced MR imaging: evaluation of biliary
and renal excretion in normal and cirrhotic livers. Eur J Radiol 80, e207–11 (2011).
70
106. Tschirch, F. T. et al. Contrast-enhanced MR cholangiography with Gd-EOBDTPA in patients with liver cirrhosis: visualization of the biliary ducts in comparison with patients with normal liver parenchyma. Eur Radiol 18, 1577–86
(2008).
107. Geetha, A., Lakshmi Priya, M. D., Jeyachristy, S. A. & Surendran, R. Level of
oxidative stress in the red blood cells of patients with liver cirrhosis. Indian J Med
Res 126, 204–10 (2007).
108. Zollner, G. & Trauner, M. Mechanisms of cholestasis. Clin Liver Dis 12, 1–26, vii
(2008).
109. Kullak-Ublick, G. A. et al. Expression of the hepatocyte canalicular multidrug
resistance protein (MRP2) in primary biliary cirrhosis. Hepatol Res 23, 78–82
(2002).
110. Wrangsjö, A. et al. Non-rigid registration using morphons. In Proc. SCIA, 501–
510 (2005).
111. Imbert-Bismut, F. et al. Biochemical markers of liver fibrosis in patients with
hepatitis C virus infection: a prospective study. Lancet 357, 1069–75 (2001).
112. Castéra, L. et al. Prospective comparison of transient elastography, Fibrotest,
APRI, and liver biopsy for the assessment of fibrosis in chronic hepatitis C. Gastroenterology 128, 343–50 (2005).
113. Friedrich-Rust, M. et al. Performance of transient elastography for the staging of
liver fibrosis: a meta-analysis. Gastroenterology 134, 960–74 (2008).
114. Ziol, M. et al. Noninvasive assessment of liver fibrosis by measurement of stiffness in patients with chronic hepatitis C. Hepatology 41, 48–54 (2005).
115. Foucher, J. et al. Diagnosis of cirrhosis by transient elastography (FibroScan): a
prospective study. Gut 55, 403–8 (2006).
116. Hagiwara, M. et al. Advanced liver fibrosis: diagnosis with 3D whole-liver perfusion MR imaging--initial experience. Radiology 246, 926–34 (2008).
117. Wang, Q. B., Zhu, H., Liu, H. L. & Zhang, B. Performance of magnetic resonance
elastography and diffusion-weighted imaging for the staging of hepatic fibrosis:
A meta-analysis. Hepatology 56, 239–47 (2012).
118. Yin, M. A preliminary assessment of hepatic fibrosis with magnetic reasonance
elastography. Clin Gastroenterol Hepatol 5, 1207–1213 (2007).
119. Huwart, L. Magnetic resonance elastography for the non invasive staging of
liver fibrosis. Gastroenterology 135, 32–40 (2008).
120. Sandrasegaran1, K. et al. Value of Diffusion-Weighted MRI for Assessing Liver
Fibrosis and Cirrhosis. AJR Am J Roentgenol 193, 1556–1560 (2009).
121. Lewin, M. et al. Diffusion-weighted magnetic resonance imaging for the assessment of fibrosis in chronic hepatitis C. Hepatology 46, 658–65 (2007).
71
122. Yamada, A. et al. Quantitative evaluation of liver function with use of
gadoxetate disodium-enhanced MR imaging. Radiology 260, 727–33 (2011).
123. Henninger, B. et al. Evaluation of MR imaging with T1 and T2* mapping for the
determination of hepatic iron overload. Eur Radiol 22, 2478–86 (2012).
124. Van Werven, J. R. et al. Assessment of hepatic steatosis in patients undergoing
liver resection: comparison of US, CT, T1-weighted dual-echo MR imaging, and
point-resolved 1H MR spectroscopy. Radiology 256, 159–68 (2010).
125. Taouli, B. et al. Diffusion-weighted MRI for quantification of liver fibrosis: preliminary experience. AJR Am J Roentgenol 189, 799–806 (2007).
126. Zhou, M. L. et al. Comparative study on clinical and pathological changes of
liver fibrosis with diffusion-weighted imaging. Zhonghua Yi Xue Za Zhi 89,
1757–61 (2009).
127. Yin, M. MR elastography of the liver: Observations from a review of 1377 exams.
Proc. Intl. Soc. Mag. Reson. Med. 19, 390 (2011).
128. Nguyen, D. & Talwalkar, J. A. Noninvasive assessment of liver fibrosis. Hepatology 53, 2107–10 (2011).
129. Caravan, P. et al. Collagen-targeted MRI contrast agent for molecular imaging of
fibrosis. Angew Chem Int Ed Engl 46, 8171–3 (2007).
130. Chow A.M. et al Molecular MRI of liver fibrosis by Fibrin-Fibronectin targeted
contrast agent in an experimental mouse model. Proc. Intl. Soc. Mag. Reson. Med
19 (2011).
72
Fly UP