A study of T 1 relaxation time as a measure of liver fibrosis and the influence of confounding histological factors

A study of T 1 relaxation time as a measure of liver fibrosis and the influence of confounding histological factors
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A study of   T  1  relaxation time as a measure of liver  fi brosis and the influence of confoundinghistological factors Caroline L. Hoad a,b , Naaventhan Palaniyappan b , Philip Kaye b,c ,Yulia Chernova b , Martin W. James b , Carolyn Costigan a , Andrew Austin d ,Luca Marciani b , Penny A. Gowland a , Indra N. Guha b , Susan T. Francis a and Guruprasad P. Aithal b, * Liver biopsy is the standard test for the assessment of   fi brosis in liver tissue of patients with chronic liver disease.Recent studies have used a non-invasive measure of   T  1  relaxation time to estimate the degree of   fi brosis in a singleslice of the liver. Here, we extend this work to measure  T  1  of the whole liver and investigate the effects of additionalhistological factors such as steatosis, in fl ammation and iron accumulation on the relationship between liver  T  1  and fi brosis. We prospectively enrolled patients who had previously undergone liver biopsy to have MR scans. A non-breath-holding, fast scanning protocol was used to acquire MR relaxation time data ( T  1  and  T  2 *), and blood serumwas used to determine the enhanced liver  fi brosis (ELF) score. Areas under the receiver operator curves (AUROCs)for  T  1  to detect advanced  fi brosis and cirrhosis were derived in a training cohort and then validated in a second co-hort. Combining the cohorts, the in fl uence of various histology factors on liver  T  1  relaxation time was investigated.The AUROCs (95% con fi dence interval (CI)) for detecting advanced  fi brosis ( F  ≥ 3) and cirrhosis ( F  =4) for the trainingcohort were 0.81 (0.65 – 0.96) and 0.92 (0.81 – 1.0) respectively (  p < 0.01). In fl ammation and iron accumulation wereshown to signi fi cantly alter  T  1  in opposing directions in the absence of advanced  fi brosis; in fl ammation increasing T  1  and iron decreasing  T  1 . A decision tree model was developed to allow the assessment of early liver disease basedon relaxation times and ELF, and to screen for the need for biopsy.  T  1  relaxation time increases with advanced  fi bro-sis in liver patients, but is also in fl uenced by iron accumulation and in fl ammation. Together with ELF, relaxation timemeasures provide a marker to stratify patients with suspected liver disease for biopsy. Copyright © 2015 John Wiley& Sons, Ltd.  Additional supporting information may be found in the online version of this article at the publisher  ’  s web site. Keywords:  relaxometry; hepatobiliary; other applications;  fi brosis; in fl ammation; iron INTRODUCTION Clinical decision making and prognosis in chronic liver diseasedepends upon the degree of necroin fl ammation and  fi brosis.Liver biopsy is the primary tool used in this evaluation (1), butsince a liver biopsy samples only one-50 000th of the liver volumeit is prone to sampling errors and intra- and inter-observervariability, particularly in biopsies less than 25mm in length(2,3). Furthermore, the procedure can cause bleeding and pain(4). Therefore, the development of new non-invasive tests toevaluate both necroin fl ammation and  fi brosis is highly desirable.Serum biomarker panels (5,6) currently being evaluated arenot organ speci fi c; for instance, they are elevated in in fl amma-tory joint conditions as well as liver disease (7). Transientelastography is used increasingly to detect advanced  fi brosis,but this has a number of pitfalls (8); importantly, the resultscan be confounded by acute in fl ammation (9 – 11) and cholesta-sis (12). Recently, phase contrast magnetic resonanceelastography (MRE) has attracted considerable interest for thedetection of   fi brosis within the liver (13,14), with tissue stiffnessincreasing with  fi brosis. However, MRE stiffness measures donot directly relate to  fi brosis and are also in fl uenced by *  Correspondence to: Guruprasad P. Aithal, NIHR Nottingham Digestive DiseasesBiomedical Research Unit at the Nottingham University Hospitals NHS Trust and University of Nottingham, UK.E-mail: Guru.Aithal@nottingham.ac.uk  aa  C. L. Hoad, C. Costigan, P. A. Gowland, S. T. FrancisSPMIC, School of Physics and Astronomy, University of Nottingham, UK  bb  C. L. Hoad, N. Palaniyappan, P. Kaye, Y. Chernova, M. W. James, L. Marciani, I.N. Guha, G. P. Aithal NIHR Nottingham Digestive Diseases Biomedical Research Unit at the Notting-ham University Hospitals NHS Trust and University of Nottingham, UK  cc  P. KayeDepartment of Cellular Pathology, Nottingham University Hospitals NHS Trust, UK  dd  A. AustinDerby Royal Hospital, Derby, UK  Abbreviations used :  AUROC, area under the receiver operator curve; B-A,Bland  –  Altman; CI, con fi dence interval; CV, coef  fi cient of variance; ELF, en-hanced liver   fi brosis; HA, hyaluronic acid; ICC, intra-class correlation coef  fi -cient; MRE, magnetic resonance elastography; NPV, negative predictivevalue; PIIINP, procollagen III amino-terminal peptide; PPV, positive predictivevalue; ROC, receiver operator curve; SD, standard deviation; TIMP-1, tissue in-hibitor of metalloproteinase-1. Research article Received: 22 January 2015, Revised: 04 March 2015, Accepted: 11 March 2015, Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/nbm.3299 NMR Biomed.  2015 Copyright © 2015 John Wiley & Sons, Ltd.  in fl ammation and hepatic congestion (15). Other MR measures,such as relaxation time and fat fraction, provide a method of monitoring the effects of various different pathological processesoccurring in the liver, across the whole liver, non-invasively, andin a short scan session (16,17). However, the measurement of any MR parameter in isolation ignores its dependence on multi-ple histological factors (18). The aim of this study was to investigate the relationshipbetween the longitudinal relaxation time ( T  1 ), measured with MRIin a short scan time across the whole liver, and the stage of liver fi brosis, measured by liver biopsy (19,20). First a training cohortwas used to determine the receiver operator curves (ROCs) forthe diagnosis of advanced  fi brosis and cirrhosis, and a subsequentcohortwasusedtovalidatetheresults.Subsequently,thedatafromthe combined cohort was used to determine how the relationshipbetween  fi brosis and  T  1  was confounded by other histologicalfactors such as steatosis, in fl ammation and iron accumulation. METHODS Study population  This study was conducted with approval from the NottinghamNHS Ethics Committee at Nottingham University Hospitals NHS Trust and Royal Derby Hospital between May 2009 and Septem-ber 2012; all patients gave written, informed consent (NationalResearch Register Document NCT01572064). We included con-secutive patients investigated for chronic liver disease with aliverbiopsy core length of at least 25mm. MRI was performed within3months of biopsy and patients received no therapeutic inter-ventions between biopsy and imaging. Five patients ’  data wereused as pilot data to optimize imaging parameters. 134 patientswere enrolled to the study, 81 to the training (including  fi ve pa-tients for pilot data) and 53 to the validation cohort. Technical is-sues associated with the MR scanner resulted in incompleterelaxometry data for six subjects, and two further subjects wereexcluded from the analysis due to inadequate biopsy. Biopsy procedure  The liver biopsies were obtained via either the percutaneous orthe transjugular route. Patients were fasted overnight beforethe procedure and biopsies were carried out by experienced op-erators. The majority of the biopsies acquired were percutaneousand these were all carried out with ultrasound guidance, withsamples taken from the right lobe of the liver. No major compli-cations from the liver biopsy were reported in this cohort. MR data acquisition All patients were scanned once on a 1.5T scanner (Achieva,Philips Medical Systems, Best, Netherlands) with body transmitand a  fi ve-element SENSE cardiac receive coil placed over theliver. Longitudinal ( T  1 ) and transverse ( T  2 *) relaxation time mapssampling the whole liver were measured from MR images gener-ated using single shot EPI data with nine slices per volume, a3×3×8mm 3 voxel size, 4mm slice gap (33%), 96×96 imagematrix and SENSE factor 2. All other sequence parameters spe-ci fi c to each measurement are given in Table 1. The position of slices was planned from coronal and transverse localizers, so asto maximize coverage of the left lobe of the liver and ensure thatthe right lobe was present in all nine slices. Data were acquiredusing a respiratory triggered acquisition near the end of the ex-piration phase of the breathing cycle, avoiding the need forbreath-holding. To avoid the in fl uence of liver fat on the relaxa-tion time measurements, fat saturation pulses were used for allsequences (details provided in Table 1). For the  T  1  data (21) a sin-gle inversion pulse was applied per volume, with nine slices ac-quired in a period of 432ms following an inversion delay. Thiswas repeated for 10 inversion times ( T  I ). The acquisition wasperformed twice with the order of slice acquisition reversed,resulting in slices being acquired at 20  T  I  values ranging from100 to 1484ms. Scan times for the acquisition of therelaxometry data ranged from 2 to 3min for the  T  1  data and1 to 2min for the  T  2 * data, dependent on the duration of the subject ’ s respiratory cycle. MR data analysis MRI data were analysed by observers blind to the histopathologyresults.A mask was drawn around the liver from a single  T  E  /  T  I  volume(nine slices; shortest  T  E  for  T  2 * data, and longest  T  I  for  T  1  data)and each voxel was  fi tted for the relaxation parameter under in-vestigation. To avoid noise bias in the  T  2 *  fi tting, a weighted lin-ear least squares  fi t was used, with 1/  T  E  (in ms) as the weightingfactor. Histogram analysis was used to assess the distribution of each relaxation time parameter ( T  1  and  T  2 *) across the wholeliver. For each subject, each histogram of voxel values was  fi ttedto a Gaussian function and the position of the peak (the mode of the distribution) was used to represent the liver tissue relaxationtime (termed tissue  T  1  /  T  2 *). This provided an automated methodto eliminate those voxels containing blood in vessels (which hasa longer relaxation time than liver tissue ( T  1  of 1.2s (22) versus Table 1.  MR sequence parameters for  T  1  and  T  2 * relaxation time measurementsSequence parameter  T  1  (varying  T  I )  T  2* (varying  T  E )Sequence type IR SE-EPI GE-EPIMinimum  T  R  (s) 4 3No of different  T  I  /  T  E  10 5No of repeats 2 a 3 Total no of images for measurement 20 15 T  I  /  T  E  (s) 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9,1.0 0.012, 0.015, 0.02, 0.03, 0.04Fat saturation pulse water only excitation SPIR a Slices acquired in ascending and descending order to give a larger range of inversion times for a given slice.  T  I  provides theinversion time for the  fi rst slice in the multi-slice dataset, for which there was a temporal slice spacing of 48ms.C. L. HOAD  ET AL. wileyonlinelibrary.com/journal/nbm  Copyright © 2015 John Wiley & Sons, Ltd.  NMR Biomed.  2015  T  1 ~0.6s at 1.5T (23)) without having to draw detailed masks toexclude these voxels. Intra-observer and inter-observer variability  For the intra-observer measurements, one observer carried outthe data analysis on two separate occasions at least 3monthsapart with no speci fi c interventions to data analysis between;for the inter-observer measurements, two observers performedthe data analysis. This was assessed on data from 20 chronic liverdisease patients to provide realistic biological variation. Intra-observer and inter-observer variability was determined for allMR parameters using the limits of agreement on a Bland – Altman(B-A) plot (24). The coef  fi cient of variance (CV) and the intra-classcorrelation coef  fi cient (ICC) for both intra- and inter- observermeasures were also determined. Repeatability study  Healthy volunteers were used to assess repeatability of mea-sures, to avoid variations between scan sessions due to patho-logical changes in patients. Fourteen healthy subjects (sevenmale, mean age 30years, range 24 – 55 years) were scanned ontwo separate occasions (mean time between scan sessions16days) after an overnight fast. The scanning protocol and dataanalysis was identical to that performed on patients. The CV intissue  T  1  and  T  2 * between the two visits was calculated; B-A plotswere used to determine the limits of agreement between visits; aPearson correlation coef  fi cient between the parameters mea-sured at the two visits was also determined. Histopathology and blood samples Biopsies were stained with hematoxylin and eosin, picrosirius redand Perls ’  Prussian blue stains prior to assessment by one pathol-ogist, blinded to the MR data. Due to the varied aetiologies, acomposite grading system (see Supplementary Information)was used to categorize biopsies based on necroin fl ammation(none/mild, moderate, severe) and  fi brosis was de fi ned accord-ing to the NASH-CRN scoring (F0 – 4). Fat (S0 – 3) and ballooning(0 – 2) were graded as described according to the NASH-CRNgrading system (25). Amount of iron was graded as I0 – 4. In addi-tion, patients had blood samples taken to assess their enhancedliver  fi brosis (ELF) score (6). These serum samples were analyzedfor levels of tissue inhibitor of metalloproteinase-1 (TIMP-1),hyaluronic acid (HA) and procollagen III amino-terminalpeptide (PIIINP) at an independent reference laboratory(iQur, UK) using an ADVIA Centaur and the manufacturer ’ sreagents in accordance with the manufacturer ’ s instructions(Siemens Healthcare Systems, Tarrytown, NY, USA). ELFscores were calculated using the manufacturer ’ s publishedupdated algorithm (26), with  C   the concentration of the as-say in ng/mL:ELF score  ¼  2 : 494  þ 0 : 846 ln  C  HA ð Þ þ 0 : 735 ln  C  PIIINP ð Þþ 0 : 391 ln  C   TIMP  1 ð Þ : Statistical analysis Statistical analysis of the data was carried out using IBMSPSS® 20.0.Subjects were grouped according to their NASH-CRN histolog-ical  fi brosis score (F0 – 4), and liver  T  1  (median across group) wasplotted against  fi brosis score for both the training and validationcohorts. This was repeated for ELF score against NASH-CRN his-tological  fi brosis score. Areas under the receiver operator curve(AUROCs) were calculated for  T  1  to determine advanced  fi brosis( F  ≥ 3) and cirrhosis ( F  =4), for the training and validation cohorts. Training and validation data were then combined to assessAUROCs and cut-off values for  T  2 *, to determine signi fi cant iron( I  ≥ 2), in the combined cohort. Patient data were then dividedinto two groups based on using their histological  fi brosis gradingto determine the in fl uence of other histological factors, in addi-tion to  fi brosis, on the measured value of   T  1 : Group 1 comprisedall patients without advanced  fi brosis (F0 – 2) and Group 2 com-prised those patients with advanced  fi brosis and cirrhosis (F3 – 4).Group 1 and 2 were then sub-divided into patients with ( ≥ S2)and without ( < S2) signi fi cant steatosis determined fromhistology, and the mean liver tissue  T  1  values of the patients inthese two sub-divisions were compared using an unpaired  t  -test. This was then repeated for Groups 1 and 2 sub-divided intopatients with ( ≥ I2) and without ( < I2) signi fi cant iron accumulationdetermined from histology. For those patients with low iron ( < I2),Groups 1 and 2 were sub-divided into patients with none – mildin fl ammation and moderate – severe from histology, and livertissue  T  1  values compared between these sub-divisions. ELFscores were also assessed as a method to separate  fi brosis andin fl ammation.A simple decision tree model combining relaxation time ( T  1 and  T  2 *) data and blood serum ELF scores was formed to predictthe presence of signi fi cant iron,  fi brosis and in fl ammation in theliver tissue and hence the need for biopsy con fi rmation. Thismodel was  fi rst split on signi fi cant iron accumulation as pre-dicted by  T  2 * cut-off value, as iron has already been shown to in- fl uence  T  1  relaxation time measurements (18). RESULTS Seven patients withdrew due to claustrophobia and four pa-tients ’  datasets were excluded due to excessive iron accumula-tion. For the remaining 110 subjects (64 training, 46 validation)all biopsy and MR measurements were completed; patient de-mographics are shown in Table 2. For three subjects no bloodsamples were taken and so no ELF scores were available.Representative  T  1  and  T  2 * maps (taken from the multi-slicedata set) are shown in Fig. 1 for different histological stages. The effects of different stages of   fi brosis and iron accumulation,and of tissue heterogeneity and vessels, can clearly be seen. AllMR measurements were normally distributed. Training and validation cohort Figure 2 shows the variation in liver tissue  T  1  with NASH-CRN fi brosis scores, and the variation of ELF scores with NASH-CRN fi brosis scores, for the training and validation cohorts. In thetraining cohort, the AUROCs (95% con fi dence interval, CI) for de-tecting advanced  fi brosis ( F  ≥ 3) and cirrhosis ( F  =4) were 0.81(0.65 – 0.96) and 0.92 (0.81 – 1.0) respectively, and were statisticallysigni fi cant at  p < 0.01. For the validation cohort, the AUROCs(95% CI) for detecting advanced  fi brosis and cirrhosis were 0.77(0.62 – 0.92) and 0.83 (0.61 – 1.0), signi fi cant at  p < 0.01 for ad-vanced  fi brosis and  p < 0.05 for cirrhosis. In the training cohort, T  1  cut-off values of 727ms and 682ms for cirrhosis (F4) and T1 RELAXATION TIME AS A MEASURE OF LIVER FIBROSIS NMR Biomed.  2015 Copyright © 2015 John Wiley & Sons, Ltd.  wileyonlinelibrary.com/journal/nbm  advanced  fi brosis ( F  ≥ 3) respectively had sensitivities of 0.8 and0.71 and speci fi cities of 0.93 and 0.8. In the validation cohort,for these  T  1  cut-off values the sensitivities dropped to 0.61 and0.67 for advanced  fi brosis and cirrhosis respectively, with speci- fi cities of 0.89 and 0.8, indicating that  T  1  values are in fl uencedby factors other than  fi brosis. Combined cohort: in fl uence of histological factors In the combined cohort the AUROCs of   T  1  for detecting ad-vanced  fi brosis ( F  ≥ 3) and cirrhosis ( F  =4) were 0.78 (0.68 – 0.89)and 0.87 (0.76 – 0.98) respectively,  p < 0.001 for both. In addition,the AUROC of   T  2 * for detecting signi fi cant iron ( ≥ I2) was 0.91(0.84 – 0.96),  p < 0.001. A  T  2 * cut-off value of 22.6ms was foundto de fi ne subjects without iron accumulation in their liver (sensi-tivity 0.89, speci fi city 0.71, positive predictive value (PPV) 40%and negative predictive value (NPV) 97%), resulting in 67subjects classi fi ed as having no signi fi cant iron in the combinedcohort. An ELF score cut-off value for advanced  fi brosis (F3 – 4) of 9.29 was found, with a sensitivity of 0.67 and speci fi city of 0.75. The in fl uences of the different histological factors (iron,steatosis and in fl ammation) on liver  T  1  are summarized in Table 3and Fig. 3, for all subjects without advanced  fi brosis  –  Group 1(F0 – 2)  –  and those with advanced  fi brosis and cirrhosis  –  Group2 (F3 – 4). There was no in fl uence of signi fi cant steatosis ( ≥ S2) on T  1  (Fig. 3A). However, iron accumulation ( ≥ I2) signi fi cantly re-duced the measured liver tissue  T  1  for Group 1 (F0 – 2) in the ab-sence of advanced  fi brosis, with a similar trend seen in Group 2with advanced  fi brosis (F3 – 4) (Fig. 3(B)). Fig. 3(C) considers onlythose 67 patients without signi fi cant iron ( T  2 * ≥ 22.6ms), andhighlights that moderate to severe in fl ammation signi fi cantly in-creased  T  1  independently from  fi brosis in Group 1 (F0 – 2), but notin Group 2 (F3 – 4). A  T  1  cut-off for in fl ammation derived from thesubset of subjects in Group 1 ( N  =47 patients) was calculated tobe 678ms, with a sensitivity of 0.67 and speci fi city of 0.83. Fig. 3(D) shows the ELF scores for the combined cohort ( N  =107 withblood samples taken) for patient groups with and without signif-icant in fl ammation, for Group 1 (F0 – 2) and Group 2 (F3 – 4). Atlow  fi brosis levels (F0 – 2), ELF scores were not signi fi cantlyaltered by in fl ammation. Decision tree model Combining the  T  1  cut-off value for in fl ammation and the ELFscore cut-off value for advanced  fi brosis (F3 – 4) produced asimple decision tree model to stratify patients with  fi brosis andin fl ammation, as illustrated in Fig. 4. Table 4 summarizes theresults of using the MRI  T  1  values and ELF scores in this modelto stratify  fi brosis and in fl ammation in the 65 patients withoutsigni fi cant iron as determined by MRI  T  2 * (2 of 67 patients didnot have blood serum taken). If the results of this decision treemodel are looked at simplistically such that the aim is simplyto indicate correctly the need for a biopsy if the subject haseither signi fi cant in fl ammation or  fi brosis (positive result), thismodel produces a 72% PPV, 89% NPV, with sensitivity of 0.90and speci fi city of 0.71. Repeatability and intra-/inter-observer variability  The healthy subject repeatability data for MR relaxation times ( T  1 and  T  2 *) and inter- and intra-observer variability are summarizedin Table 5. The variance in liver tissue  T  1  values between healthysubjects (645±44ms (mean±standard deviation, SD)) wasmuch larger than the difference between visits (  3±24ms, CV 1.8±1.8%). No data showed any large bias between visits, ob-server or analysis session, as assessed by B-A analysis. Inter-and intra-observer variability was low, with ICCs > 0.99 for allMR relaxation times. DISCUSSION  This study has con fi rmed that liver longitudinal relaxation time( T  1 ) varies with degree of   fi brosis, as reported in previous studies(16,19). However, importantly it has also shown that the mea-sured  T  1  value is affected by levels of iron and in fl ammation inthe liver. This suggests that it would be simplistic to attempt touse  T  1  alone to assess liver  fi brosis or indeed in fl ammation. Thismay explain why, in this study, when the  T  1  cut-off values corre-sponding to  fi brosis derived from the training cohort were Table 2.  Patient demographics (mean (SD) presented if notstated otherwise) Training( N  =64)Validation( N  =46)Combined( N  =110) Biopsy measures Fibrosis stage,  N   (%)0 12 (18.8) 12 (26.1) 24 (21.8)1 22 (34.4) 8 (17.4) 30 (27.3)2 16 (25) 8 (17.4) 24 (21.8)3 9 (14.1) 12 (26.1) 21 (19.1)4 5 (7.8) 6 (13.0) 11 (10.0)Steatosis stage,  N   (%)0 19 (29.7) 8 (17.4) 27 (24.6)1 22 (34.4) 22 (47.8) 44 (40)2 13 (20.3) 6 (13.0) 19 (17.3)3 10 (15.6) 10 (21.7) 20 (18.2)Iron content score,  N   (%)0 39 (60.9) 21 (45.7) 60 (54.6)1 16 (25.0) 15 (32.6) 31 (28.2)2 8 (12.5) 8 (17.4) 16 (14.6)3 1 (1.6) 1 (2.2) 2 (1.8)4 0 (0) 1 (2.2) 1 (0.9)In fl ammation score,  N   (%)None/mild 37 (57.8) 29 (63.0) 66 (60.0)Moderate 16 (25.0) 8 (17.4) 24 (21.8)Severe 11 (17.2) 9 (19.6) 20 (18.2) Time between biopsyand MRI (days)58 (18) 67 (19) 62 (19)Biopsy length (mm) 28.7 (7.3) 30.8 (6.4) 29.6 (7.0)Biopsy width (mm) 1.0 (0.2) 1.0 (0.1) 1.0 (0.2) Other  Age (years) 49 (12) 51 (11) 50 (11)Height (m) 1.72 (0.10) 1.72 (0.08) 1.72 (0.09)Waist circumference (cm) 90.5 (16.2) 98.3 (14.0) 93.7 (15.7)Weight (kg) 79.9 (13.6) 83.8 (14.7) 81.5 (14.1)Aetiology,  N   (%)Alcoholic liver disease 7 (11.1) 6 (13.6) 13 (12.2)Haemochromatosis 2 (3.2) 1 (2.3) 3 (2.8)Hepatitis B and C 13 (20.6) 8 (18.2) 21 (19.6)Non-alcoholic fatty liver 36 (57.1) 26 (59.1) 62 (57.9)Normal 3 (4.8) 3 (6.8) 6 (5.6)Other 2 (3.2) 0 (0) 2 (1.9)C. L. HOAD  ET AL. wileyonlinelibrary.com/journal/nbm  Copyright © 2015 John Wiley & Sons, Ltd.  NMR Biomed.  2015  applied to the validation cohort, the resulting sensitivities werelower than for the training cohort. In addition, variability inbiopsy samples may also confound such results. The technique for acquiring  T  1  data in this study differs fromprevious studies in that it uses a respiratory triggered, multi-slice, EPI readout, allowing the whole liver to be sampled in Figure 1 . Example maps showing (A)  T  1  relaxation time with NASH-CRN histological scoring for  fi brosis (F0 – 4); (B)  T  2 * relaxation times with histologicalscoring for iron accummulation (I0 – 4). Figure 2 . Box plots of the distribution of (A) liver tissue  T  1  and (B) blood serum ELF score with NASH-CRN histological scoring for  fi brosis, split by train-ing (T) and validation (V) cohorts. Table 3.  Inter-dependence of histological factors on  T  1 Histological factor Group 1: no advanced  fi brosisF0 – 2Mean  T  1  (SD)/ms N   (number of subjects)Group 2: advanced  fi brosisF3 – 4Mean  T  1  (SD)/ms N   (number of subjects)Steatosis S0 – 1 S2 – 3 S0 – 1 S2 – 3640 (42) 643 (34) 720 (85) 697 (65) N  =110 51 27 20 12Iron I0 – 1 I2 – 4 I0 – 1 I2 – 4646 (38) 610* (33) 717 (79) 694 (77) N  =110 67 11 24 8In fl ammation none/mild moderate/severe none/mild moderate/severe(non-signi fi cant iron only) 643 (33) 683* (46) 750 (169) 732 (68) N  =67 35 12 2 18*Independent samples  t  -test signi fi cant differences between means for histological factor in LHS column,  p < 0.01. T1 RELAXATION TIME AS A MEASURE OF LIVER FIBROSIS NMR Biomed.  2015 Copyright © 2015 John Wiley & Sons, Ltd.  wileyonlinelibrary.com/journal/nbm
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks