Value of Whole-Liver Apparent Diffusion Coefficient Histogram Analysis for Quantification of Liver Fibrosis Stages
Zhaoxia Yang1, Xinxing Hu1, Heyue Liang1, Robert Grimm2, Caixia Fu3, Xu Yan4, Hui Liu4, Mengsu Zeng1, and Shengxiang Rao1

1Department of Radiology, Zhongshan Hospital, Fudan University, Shanghai, China, People's Republic of, 2MR Application Predevelopment, Siemens Healthcare, Erlangen, Germany, 3Application department, Siemens Shenzhen Magnetic Resonance Ltd., Shenzhen, China, People's Republic of, 4MR Collaboration NE Asia, Siemens Healthcare, Shanghai, China, People's Republic of


Histogram analysis of whole-liver ADC maps was applied for patients with liver fibrosis to assess its diagnostic value and correlation with fibrosis stages achieved from liver biopsy. Our study shows that the mean and the 75th percentile ADC were significantly lower with the increasing grade of liver fibrosis. The median, standard deviation (SD), 5th, 10th, 25th, 90th and 95th percentiles also showed statistical differences among pathological fibrosis grades. In conclusion, histogram analysis of whole-liver ADC maps has a potential value for quantifying liver fibrosis grades.

Introduction and purpose

Accurate assessment of fibrosis stage and early detection of cirrhosis are critical to identify patients in need of close monitoring, management of complications and treatment of underlying liver disease [1]. Liver biopsy is commonly used in clinical practice, but it is invasive and has poor acceptance among patients. A reliable and non-invasive tool is strongly needed. Various MRI methods including DWI [2], MRE [3] etc. have been proposed to access the liver fibrosis stage and have shown promising results. The goal of this study was to explore whether histogram analysis of the whole-liver parenchyma’s apparent diffusion coefficient (ADC) values could quantify liver fibrosis stages.

Materials and Methods

Eighty-one patients underwent MRI exams including DWI on a 3T MR scanner (MAGNETOM Verio, Siemens Healthcare, Erlangen, Germany), followed by a liver biopsy as parts of their clinical care. The parameters of the free-breathing spin-echo EPI DWI sequence are: TR =3400ms, TE = 70ms, b = 0, 500 s/mm2, FOV = 400x312mm2, scan matrix = 128x80, slice thickness = 6mm, averages = 4, transversal orientation with coverage of whole liver, total scan time = 65s. The number of patients in each pathological fibrosis grade was 14, 10, 13, 15, 29 for F0, F1, F2, F3, F4 (METAVIR system), respectively. Histogram analyses of ADC maps were performed on a prototype software (MR OncoTreat, Siemens Healthcare, Erlangen, Germany). Semi-automated segmentation was executed within the software by delineating the liver parenchyma without focal lesions (excluding iron overload patients, fatty liver, prior partial hepatectomy or anti-cancer treatments, metastases or other focal liver lesions). Quantitative histogram parameters of mean, median, skewness, kurtosis, standard deviation (SD), 5th, 10th, 25th, 75th, 90th and 95th percentiles were calculated with MR OncoTreat, and the correlation between different histopathologic grades and histogram analysis parameters was determined through nonparametric Spearman correlation coefficients.


The mean and 75th percentile were significantly lower with the increasing grade of liver fibrosis (P<0.0001, rho= -0.469 for mean and P<0.0001, rho= -0.460 for 75th percentile; respectively), see figure 1. Median, standard deviation (SD), 5th, 10th, 25th, 90th and 95th percentile also showed statistical differences among pathological fibrosis grades with P value from 0.0001 to 0.0226, rho value from -0.430 to 0.253. Skewness and kurtosis demonstrated no statistical differences (P=0.3771, rho= 0.0994 for skewness, P=0.0541, rho=0.215 for kurtosis; respectively), see table 1.


Liver fibrosis is the accumulation of extracellular matrix proteins in chronic liver disease [4]. Several studies have shown that ADC measured with DWI is significantly lower in fibrotic and cirrhotic livers compared with normal livers [2, 5-6]. Contradictive studies found that mean ADC values cannot be used for accurate staging [7]. Histogram analysis reveals the distribution of ADC values within the liver and has the potential to detect changes in the overall structure of the liver parenchyma indicating the presence of fibrosis at a level. Furthermore, histogram analysis was performed for the whole liver volume in our study, which is more representative with respect to tissue heterogeneity and is less susceptible to measuring variations compared with single-slice analyses. Our results showed that especially the volumetric ADC mean and 75th percentile values have a significantly negative correlation with the increasing stages of liver fibrosis.


Histogram analysis of whole-liver ADC maps has a potential value for quantifying liver fibrosis grades.


No acknowledgement found.


[1] Taouli B et al. AJR Am J Roentgenol. 2009 Jul; 193(1): 14–27. [2] Koinuma M et al. J Magn Reson Imaging. 2005; 22: 80–85. [3]Huwart L et al. NMR Biomed. 2006. [4] Friedman SL. J Hepatol. 2003; 38 Suppl 1:S38-53. [5] Bakan AA et al. Eur Radiol. 2011; 22:682–687. [6] Annet L et al. J Magn Reson Imaging. 2007; 25:122–128. [7] Bülow R et al. Eur Radiol 2013;23:1281–1287.


Table 1: Spearman correlation coefficients for whole-liver histogram analysis parameters of ADC maps

Figure 1: Box plot of whole-liver histogram analysis of mean ADC values (A) and 75th percentile (B) for fibrosis stages F0, F1, F2, F3, and F4. Both mean and 75th percentile values were significantly lower with increasing fibrosis grade

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)