Diffusion Weighted Imaging Based On IVIM Mode In Evaluating Liver Function:a Pilot Study

Objective:To explore the clinical application value of diffusion weighted imaging based on IVIM mode with total estimated liver in evaluating liver function. To explore the repeatability of D, D*, f and ADC values, and evaluate the feasibility of its use in clinical practice.Material and methods:Control group (30 cases, male 16 cases; female 14 cases; years of age 22-51 years; average age 45.2 years), liver cirrhosis group (70 cases, male 49 cases; female 21 cases; years of age 20-75 years; average age 46.4 years). According to the model for Child-Pugh liver function classification scores, liver cirrhosis patients were divided into three groups:Child-Pugh class A:29 cases; Child-Pugh class B: 19 cases; Child-Pugh class C:22 cases.The patients and volunteers underwent liver IVIM-DWI examination twice at using the same DWI sequence on a 3.0 Tesla Philips MR scanner (Philips Healthcare, Best, Netherland). Conventional MRI sequences included mDixon, axial T2WI fat suppression. IVIM-DWI examination adopted respiratory triggered single-shot echo planar imaging (SS-EPI) sequence, parallel sampling space sensitive coding technology (ASSET). Axial DWI images (TR 1642 ms, TE 62 ms, field of view (FOV) 375mm×302mm, matrix:256, excitation frequency (NEX) 2, slice thickness 5mm, slices 32, EPI factor 53, ASSET 2) were collected with spectral presaturation inversion recovery and diffusion sensitive gradient pulse in X, Y, Z axis direction applied with weighting factors of b 0,10,30,60,100,150,200, 400,600,1000 s/mm2. The intervals between two scanning were 15 minutes.The bi-exponential model from an IVIM sequence was expressed by equation Sb/S0=(1-f) exp (-bD)+f exp [-b (D*+D)]. The mono-exponential model was expressed by the equation ADC= [In (S0/S1)]/(b1-b0). IVIM-DWI parameters and parametric maps were obtained with software developed by the institute of biological engineering, southern medical university. The data of patients and volunteers were measured by two radiologists respectively. The methods are as follows:(1) Outlining the liver area according to the liver contour, and repeating 7 times for a 3D area by treating the right portal vein level as the middle level of the liver (Abandon part slices near the diaphragmatic surface and the visceral surface in order to remove the lung and intestinal gas and respiratory motion artifact); (2) Then processed by threshold method to vascular removed for the obvious difference between blood vessels and other areas of liver when b= 0, to avoid whole liver IVIM-DWI parameters affected by blood vessels; (3) After removal of blood vessels, fitting the IVIM-DWI parameters of the whole liver areas and calculating the determination coefficient (Rsquare, between 0 and 1) in the process of fitting to characterize the fitting effect. The fitting effect is regarded as good when Rsquare value is equal or greater than 0.9, while the point was removed with the Rsquare value less than 0.9. At the same time, get rid of the points with the D value greater than 0.003, D* value greater than 0.3, or f value greater than 0.5 to avoid the whole liver IVIM-DWI fitting parameters were affected by the points of bad fitting effect or abnomal parameter values. Then calculate the pure diffusion coefficient (D), pseudo-diffusion coefficient (D), and perfusion fraction (f).All statistical analysis were conducted by SPSS for Windows version 13.0 and Medcalc. D, D*, f and ADC values were represented by means±standard deviation. Kolmogorow-Smirnov test was used for Normality tests, and Levene test for homogeneity of variance test. The intraclass correction coefficient (ICC) was calculated to derive the data varability for two different radiologists, a good reproducibility was demonstrated when IC>0.75. The reproducibility of D, D*, f and ADC were evaluated with Bland-Altman method. D, D*, f and ADC values were compared between control group and liver cirrhosis group with student t test. D, D*, f and ADC values were compared between Child-Pugh class A group, B group and C group with one-way ANOVA and Post Hoc Multiple Comparisons (Least-significant Difference, LSD). Spearman test was used for the correlation between parameters and Child-Pugh liver function classification, observe the value and direction of correlation coefficient. Receiver operating characteristic (ROC) curve was used to evaluate the prediction ability of DWI parameters for liver function classification, and estimate the optimal cut-off according to the Youden Index (sensitivity+specificity-1) and the area under the ROC curve (AUC). The confidence interval of average values was set to 95%, P value less than 0.05 was considered to indicate a significant difference in all statistical tests.Results:3.1 IVIM-DWI features3.1.1 Fitting curve of whole liver IVIM bi-exponential model At the low values of b (b< 100s/mm2), the signal attenuation is quick and the slope of curve is steep; While at the high values of b, the signal attenuation is slow and the slope of curve goes fainter. The curve can acquired good fitting effect for all b values to get a hockey-like shape.3.1.2 Normal liver IVIM-DWI parameter valuesThe distribution of D value for normal liver was concentrated, and broadly in line with normal distribution. The average whole liver D value of normal liver group was (1.02±0.11)×10-3 mm2/sec. The distribution of D* value for normal liver was discrete, and broadly in line with positive skewness distribution. The average whole liver D* value of normal liver group was (88.2±13.6)×10-3 mm2/sec. The distribution of f value for normal liver was discrete, and broadly in line with positive skewness distribution. The average whole liver f value of normal liver group was (29.7±5.0)%. The distribution of ADC value for normal liver was concentrated, and broadly in line with normal distribution. The average whole liver ADC value of normal liver group was (1.35±0.11)×10-3mm2/sec.3.1.3 Cirrhotic liver IVIM-DWI parameter valuesIn this study, the average whole liver D value of liver cirrhosis group was (1.00±0.16)×10-3 mm2/sec. The average whole liver D* value of liver cirrhosis group was (76.57±13.88)×10-3 mm2/sec. The average whole liver f value of liver cirrhosis group was (27.85±5.92)%. The average whole liver ADC value of liver cirrhosis group was (1.30±0.16)×10-3mm2/sec.3.2 The repeatability of parameters between different radiologistsA total of 70 cases of liver cirrhosis and 30 cases of volunteers were evaluated by two different radiologists respectively. ICC between two radiologists of the D value was 0.997 (95% CI:0.996-0.998). ICC between two radiologists of the D* value was 0.986(95% CI:0.979-0.990). ICC between two radiologists of the f value was 0.985(95% CI:0.978-0.989). ICC between two radiologists of the f value was 0.995(95% CI:0.993-0.997). The repeatability of D, D*, f and ADC between two radiologists was good, the best of which is D, followed by ADC, D* and f.3.3 The reproducibility of parameters between two scansGraphical representation of consistency limit-Bland-Altman diagram graphically reflect consistency limit. In two-dimensional rectangular coordinates, the horizontal axis x said the average of the two methods for measuring each object, the vertical y said the difference between two methods for each object. The media solid line represented the mean value of the difference between the two different scan and the dotted lines in the top and bottom represented 95% consistency boundaries. The higher the degree of the consistency between two methods, the closer the solid line to the mean difference 0. According to the data points outside the boundaries of 95% consistency and the biggest difference within the boundaries of consistency, along with clinically acceptable level, to evaluate the consistency of the two methods. The reproducibility value of whole liver D, D*, f and ADC were 0.05%,2.9%,7.8%,0.05%, respectively. The Bland-Altman plots demonstrated the good consistency of total liver D, D*, f and ADC values between two scan and the better consistency of D and ADC.3.4 The comparison of D, D*, f and ADC between volunteers and liver cirrhosisThis study was conducted in 70 cases of liver cirrhosis group and 40 cases of normal control group. The D*, f and ADC values of liver cirrhosis group were significantly lower than control group (P<0.001, P=0.016, P=0.042, respectively). The difference of D value between liver cirrhosis group and control group was of no statistical significance (P=0.634).3.5 The comparison of parameters between different liver function groupsAccording to Child-Pugh liver function classification,70 patients of liver cirrhosis were graded into three groups:Child-Pugh class A:29 cases, B class:19 cases, C class:22 cases. The one way ANOVA analysis between liver function groups showed that D* and f values were reduced with the increasing degree of liver function, and the difference were statistically significant (P=0.001, P< 0.001, respectively). D value was reduced with the increasing degree of liver function, but the difference between different degree of liver function were no statistically significant (P= 0.136). There was no obvious difference of ADC value between Child-Pugh class A and class B, While ADC value of class C was lower than class A and B, the difference between liver function groups were significant (P=0.022).The multiple comparison of D, D*, f and ADC values between different degrees of Child-Pugh liver function showed that the difference of D* value between class A and class B, class A and class C were statistically significant (P=0.011, P=0.001, respectively), but between class B and class C was no statistically significant (P= 0.298). The difference of f value between class A and class B, class A and class C, class B and class C were statistically significant(P< 0.011, P< 0.001, P=0.041, respectively). The difference of ADC value between class A and class C(P=0.008), class B and class C(P=0.039) were statistically significant.Spearman correlation analysis results of D, D*, f, ADC and Child-Pugh liver function classification revealed that D*, f and ADC values were negatively correlated with Child-Pugh liver function classification, and the correlation respectively, f value showed the strongest correlation with the liver function classification, and the correlation coefficient was -0.620. D value had no significant correlation with liver function classification (r=-0.183, P= 0.129).ROC curve analysis obtained the cut-off values of D* and f between Child-Pugh class A and class B, C, class B and class C, as well as the area under the ROC curve (AUC) and corresponding sensitivity and specificity. The cut-off value of D* between Child-Pugh class A and class B, C was 79.50,65.31 respectively. The cut-off value of f between Child-Pugh class A and class B, class A and class C, class B and class C was 31.85,28.60,25.23 respectively.Conclusions:1. The repeatability of total liver D, D*, f, ADC values between two radiologists and the reproducibility of D, D*, f and ADC between two scans were good.2. D*, f and ADC values of liver cirrhosis group were reduced, which were related to the decreased perfusion in cirrhotic liver. There was no significant difference of D between liver cirrhosis group and control group.3.D* and f values had significantly negative correlation with Child-Pugh liver function classification, and f value had the strongest correlation with liver function classification, which would make f value be a effective imaging predictor for liver function classification. D* may potentially be a sensitive predictor for early liver function damage.