| Objective:1. To determine the reproducibility of intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) parameters during repeated examination as well as their intraobserver and interobserver agreement in measurement in hepatic ischemia-reperfusion injury (HIRI).2. To investigate the correlation between IVIM-DWI parameters and serum biochemical parameters which indicated hepatic cellular injury as well as histological alterations in HIRI.Materials and Methods:1. Animal Model and Experimental ProtocolThirty-two healthy New Zealand white rabbits (female,2kg-3kg, provided by Guangdong Animal Medical Laboratory Animal Center) were randomly and equally assigned to four groups (Group A-the sham group, Group B-1h, Group C-4h, Group D-12h) according to the reperfusion time after ischemia, with 8 animals for each group and none of them were excluded during the research.For group B, C and D, the portal triad of the left lobe and right anterior lobe was occluded with a vascular clamp for 60min to make a 70% HIRI model. Subsequently, the clamp was removed to allow reperfusion for 1h,4h and 12h in group B, C and D, respectively. For group A, hepatoduodenal ligament was exposed without clamping of the portal triad.2. MR Imaging Protocol and Image Analysis2.1 MR Imaging ProtocolThe study of IVIM-DWI was performed with Achieva 1.5T (Philips Healthcare, Best, Netherlands).The subjects were anesthetized and placed in an 8-channel phased-array coil specialized for rabbits (Product model:CG-RBC18-H150-AP, Shanghai Chenguang Medical Technologies Co., Ltd., China) in a supine position (head first). Axial images were acquired, covering the whole liver.The IVIM-DW images were acquired with a single-shot echo-planar imaging pulse sequence using a free-breathing method and parallel imaging. For fat suppression, we employed the spectral presaturation with inversion recovery (SPIR) technique,.12 b values (0,10,20,30,40,50,75,100,150,300,500, and 800 sec/mm2) were chosen for bi-exponential fitting of the signal intensities, among which 9 b values were lower than 200 sec/mm2 for a precise evaluation of the perfusion-related parameters. The diffusion gradients were applied along 3 orthogonal axis simultaneously. Other parameters of IVIM-DWI sequence included:echo time= 69.8, repetition time (TR)=around 1310 ms (the "shortest" mode was chosen, in which a minimum TR would be proposed by the system according to the FOV and number of slices.); field of view (FOV)= 150mm×150mm×45mm(AP×PL x FH); section thickness/gap=3/0 mm; number of signal averages= 6; echo train length= 53.2.2 Image Processing and AnalysisTo extract the parametric maps, raw data of IVIM-DWI were transferred and post-processed on a vendor-supplied workstation (Extended Workspace, Philips Healthcare) and a software program (PRIDE DWI Tool, version 1.5, Philips Healthcare).The nonlinear least-squares curve fittings based on the Levenberg-Marquardt algorithm were performed in bi-exponential fitting of 12 b value data to calculate the D, D* and f value with a publically available IDL routine MPFIT, using the equation as follows:SI/SIO= (1-f)·exp(-bD)+f·exp(-bD*),where SI represents the signal intensity observed at a specific b value> 0 sec/ mm2, while SIO represents the signal intensity at b=0 sec/mm2.To calculate the ADC, a monoexponential fitting method was employed with data of 12 b values, using the equation as follows:SI/SIO= exp(-bADC).All the parametric maps were generated on a pixel-by-pixel basis.Two radiologists (Observer A and Observer B, with 3 and10 years of experience in imaging, respectively) blinded to the group of every subject measured the ADC and IVIM parameters independently with operator-defined regions of interest (ROIs). To ensure the consistency of ROIs and sampling of specimens, only the left medial lobe was chosen and three ROIs were randomly placed in the middle sections, avoiding inhomogeneous areas, prominent artifacts and large hepatic vessels. For each parameter, the averages of measurements from three ROIs were calculated as the representative results. To assess the intraobserver agreement, these parameters were measured twice by Observer A (the interval was 2 weeks). To assess the interobserver agreement, the averages of parameters measured twice by Observer A were compared with those measured once by Observer B.3. Serum Biochemical Parameters and Histological AnalysisAfter MRI scanning, blood samples were acquired from each subject via the auricular vein for measurement of serum aspartate aminotransferase (serum AST), serum alanine aminotransferase (serum ALT), the ratio of serum aspartate aminotransferase to serum alanine aminotransferase (serum AST/ALT) and y-gltamyltranspeptidase (γ-GGT).4. Tissue Specimens and Histological Grading SystemEach subject was sacrificed through intravenous injection of overdose of pentobarbital. Tissue specimens of the left medial lobe were obtained and stained with hematoxylin-eosin.Histological analysis included qualitative analysis and hemi-quantitative analysis adapted from a previous scoring system. Based on five random areas per slide (200×), the hemi-quantitative analysis of stenosis in the hepatic sinusoids (HSs) were categorized into five grades:Grade 1, no stenosis or occlusion; Grade 2,≤50% stenosis in HSs; Grade 3,>50% stenosis in HSs or<50% occlusion in a minority of HSs; Grade 4,>50% occlusion in a majority of HSs; Grade 5=total occlusion in HSs. The hemi-quantitative analysis of inflammatory cell infiltration in the HSs was categorized into four grades:Grade 1= no infiltration of inflammatory cells observed; Grade 2= mild infiltration (focal involvement); Grade 3= moderate infiltration (diffuse involvement in a minority of HSs); Grade 4= severe infiltration (diffuse involvement in a majority of HSs). The hemi-quantitative analysis of congestion in the HSs was categorized into four grades:Grade 1= no congestion observed; Grade 2= mild congestion (focal involvement); Grade 3= moderate congestion (diffuse involvement in a minority of HSs); Grade 4= severe congestion (diffuse involvement in a majority of HSs). The sum of three sets of hemi-quantitative analysis was calculated as histological score for each subject, which was believed to reflect the extent of cellular swelling and subsequent microcirculation-related alteration. Two pathologists (Observer C and Observer D, with 2 and 8 years of experience, respectively) independently assessed the histological score and the averages of their results were regarded as the final score.5. Statistical AnalysisAll the statistical analyses were performed with SPSS version 19.0 (SPSS, Chicago,Ill) and MedCalc software (MedCalc, Mariakerke, Belgium).To determine the reproducibility of IVIM parameters and ADCs within a scan session, a healthy New Zealand white rabbit (female,2.8kg) underwent IVIM-DWI for 8 times and the imaging parameters were measured. Coefficients of variation (CV) were used.To determine the intraobserver and interobserver agreement of IVIM parameters and ADC, intraclass correlation coefficients (ICCs) and Bland-Altman plots with 95% limits of agreement were used.To evaluate the reproducibilities of parameters within subjects in each group, coefficients of variation (CV) were used.The Shapiro-Wilk test and Levine’s test were applied to test the normality and homogeneity of variance.Data which were in normal distribution and assumed homogeneity of variance among different groups were compared with one-way analysis of variance (ANOVA) and Bonferroni post hoc test. While Data which were not in normal distribution or did not assume homogeneity of variance among different groups, or ranked data were compared with Kruskal-Wallis test and Dunn-Bonferroni post hoc test.For analysis of correlation between f, D, ADC value and histological changes, Spearman rank correlation analysis was employed. And the correlation between f, D, ADC value and serum biochemical parameters was investigated with Pearson correlation analysis. Both analysis with inclusion of all the cases and analysis with the SHAM cases excluded were performed.A P value<0.05 indicated statistical significance.Results:1. Measurement of Imaging Parameters1.1 Reproducibility, the Intraobserver and Interobserver Agreement of Imaging Parameters1.1.1 Reproducibility IVIM Parameters and ADCs in repeated scanning of the same sequenceThe CV of D value in repeated scanning of the same sequence was 3.44%, which indicated good reproducibility. The CV of f value in repeated scanning of the same sequence was 9.86%, which indicated good reproducibility. The CV of D* value in repeated scanning of the same sequence was 44.30%, which indicated poor reproducibility. The CV of ADC value in repeated scanning of the same sequence was 3.62%, which indicated good reproducibility.1.1.2 the Intraobserver and Interobserver Agreement of Imaging ParametersThe intraobserver agreement (expressed as ICC) were shown as follows:D value, 0.973(0.945-0.987), excellent agreement; f value,0.951(0.899-0.976), excellent agreement; D* value,0.597(0.174-0.803), fair agreement; ADC value, 0.974(0.947-0.987), excellent agreement.The intrerbserver agreement (expressed as ICC) were shown as follows:D value, 0.977(0.953-0.989), excellent agreement; f value,0.987(0.974-0.994), excellent agreement; D* value,0.548(0.073-0.779), fair agreement; ADC value, 0.967(0.933-0.984), excellent agreement.1.2 Comparison of IVIM Parameters and ADCs between the Four GroupsThe Shapiro-Wilk test and Levine’s test showed that the IVIM parameters and ADCs in each group were in normal distribution, and assumed homogeneity of variance, with P>0.05 in each group.Therefore, one-way ANOVA was used in comparison of parameters in each group, and Bonferroni post hoc test was employed in pair-wise comparison.The ADC values of each group (mean± standard deviation, expressed in×10-3 mm2/sec) were shown as follows:Group A 1.29±0.09, Group B 1.04±0.10, Group C 0.87±0.07, Group D 0.94±0.11. The ANOVA showed that there was significant difference in ADC (F=31.48) among the four groups (P<0.001). The results of Bonferroni post hoc test for ADC value showed that there were significant differences between Group A (the SHAM group) and Group B/Group C/Group D, with all P< 0.001; there was significant difference between Group B and Group C, with P= 0.006; there was no significant difference between Group B and D (P= 0.314) or Group C and D(P= 0.676).The D values of each group (mean± standard deviation, expressed in ×10-3 mm2/sec) were shown as follows:Group A 1.14±0.07, Group B 0.93±0.08, Group C 0.82±0.05, Group D 0.88±0.10. The ANOVA showed that there was significant difference in D (F=26.97) among the four groups (P<0.001). The results of Bonferroni post hoc test for D value showed that there were significant differences between Group A and Group B/Group C/Group D, with all P< 0.001; there was no significant difference between Group B and C (P= 0.054), Group B and D (P= 1.000) or Group C and D (P= 0.794).The D* values of each group (meant standard deviation, expressed in ×10-3 mm2/sec) were shown as follows:Group A 143.3±39.3, Group B 141.6±35.5, Group C 124.5±47.4, Group D 142.7±22.9. The ANOVA showed that there was no significant difference in D* (F=0.47) among the four groups (P= 0.706).The f values of each group (mean± standard deviation, expressed in%) were shown as follows:Group A 14.6±3.67, Group B 12.2±2.8, Group C 5.0±2.2, Group D 7.5±3.1. The ANOVA showed that there was significant difference in f (F=16.99) among the four groups (P<0.001). The results of Bonferroni post hoc test for f value showed that there was no significant difference between Group A and B (P= 0.708), but there were significant differences between Group A and Group C/Group D, with both P< 0.001; there were significant differences between Group B and C (P< 0.001), Group B and D (P= 0.023); there was no significant difference between Group C and D (P= 0.664).2. Serum Biochemical ParametersThe Shapiro-Wilk test showed that AST, ALT, serum AST/ALT were not all in normal distribution. Therefore, Kruskall-Wallis test with Dunn-Bonferroni post hoc test was used in comparison.The Shapiro-Wilk test showed that y-GGT was in normal distribution. While Levine’s test showed that y-GGT did not assume homogeneity of variance. Therefore, Kruskall-Wallis test with Dunn-Bonferroni post hoc test was also used in comparison.The medians and interquartile ranges (in parentheses) of AST in each group (expressed in IU/L) were shown as follows:Group A 54.5 (55.0), Group B 237 (393.8), Group C 557.5 (615.5), Group D 1758.5 (2417.8). Kruskall-Wallis test showed that there was significant difference in AST among the four groups (P= 0.002). The results of Dunn-Bonferroni post hoc test for AST showed that there were significant differences between Group A and Group B (P= 0.016)/Group C (P= 0.016)/Group D (P< 0.001); there was no significant difference between Group B and C (P= 1.000), Group B and D (P= 0.273) or Group C and D (P= 0.273).The medians and interquartile ranges (in parentheses) of ALT in each group (expressed in IU/L) were shown as follows:Group A 62.0 (97.8), Group B 725.5 (2265.5), Group C 2926.5 (3261.0), Group D 11002.5 (12647.8). Kruskall-Wallis test showed that there was significant difference in ALT among the four groups (P= 0.002). The results of Dunn-Bonferroni post hoc test for ALT showed that there were significant differences between Group A and Group B (P= 0.016)/Group C (P< 0.001)/Group D (P< 0.001); there was no significant difference between Group B and C (P= 1.000), Group B and D (P= 1.000) or Group C and D (P= 1.000).The medians and interquartile ranges (in parentheses) of serum AST/ALT in each group were shown as follows:Group A 0.9 (0.5), Group B 0.3 (0.2), Group C 0.2 (0.1), Group D 0.3 (0.4). Kruskall-Wallis test showed that there was significant difference in AST/ALT among the four groups (P= 0.005). The results of Dunn-Bonferroni post hoc test for AST/ALT showed that there were significant differences between Group A and Group B (P= 0.016)/Group C (P= 0.016)/Group D (P= 0.016); there was no significant difference between Group B and C (P= 1.000), Group B and D (P= 1.000) or Group C and D (P= 1.000).The medians and interquartile ranges (in parentheses) of serum γ-GGT (expressed in IU/L) in each group were shown as follows:Group A 10.5 (6.5), Group B 20.5 (26.8), Group C 28.5 (57.8), Group D 58.5 (84.0). Kruskall-Wallis test showed that there was significant difference in γ-GGT among the four groups (P= 0.012). The results of Dunn-Bonferroni post hoc test for γ-GGT showed that there was significant difference only between Group A and Group C (P= 0.016); while there was no significant difference between Group A and B (P= 1.000), Group A and D (P = 0.706), Group B and C (P= 1.000), Group B and D (P= 1.000) or Group C and D (P= 1.000).3. Histological ScoresThe agreement in histological scores between two pathologists was good (kappa value= 0.706, P< 0.001). The averages of their scores were considered as the final results, which were expressed as median and interquartile ranges (in the parenthesis) and shown as follows:Group A 5 (1), Group B 6.75 (1.25), Group C 8.25 (2.38), Group D 8.5 (3.25). Kruskall-Wallis test showed that there was significant difference in histological scores among the four groups (P= 0.002). The results of Dunn-Bonferroni post hoc test for histological scores showed that there were significant differences between Group A and Group B (P< 0.001)/Group C (P< 0.001)/Group D (P< 0.001); there was significant difference between Group B and C (P= 0.012), but there was no significant difference between Group B and D (P= 0.783) or Group C and D (P= 1.000).4. Correlation between Imaging Parameters and Serum Biochemical ParametersAmong all the statistically significant results of correlation analysis between imaging parameters and serum biochemical parameters, only the correlation coefficients between f value, D value, ADC value and serum AST/ALT, which were calculated with all the cases included, were above 0.5 (indicating moderate or better correlation). There was moderate correlation between serum AST/ALT and f value (r = 0.538, P= 0.002), and good correlation between serum AST/ALT and D value (r= 0.747, P< 0.001) or ADC value (r= 0.748, P< 0.001).5. Correlation between Imaging Parameters and Histological ScoresCalculated with all the cases, there was exellent correlation between histological scores and f value (rs=-0.753, P<0.001), D value (rs=-0.824, P<0.001) or ADC value (rs=-0.861, P<0.001).There was fair to good correlation between histological scores and f value (rs=-0.680, P<0.001), D value (r,=-0.598, P<0.001) or ADC value (rs=-0.707, P<0.001) when the SHAM cases were excluded.Conclusion:In our study, the IVIM-DWI parameters D value, f value and ADC value demonstrated good reproducibility in repeated scanning of the same sequence, and excellent intraobserver and interobserver agreement in measurement, which can be reliably used in evaluation of HIRI of a rabbit model. There was good to excellent correlation between f value and histological scores (based on hepatic sinusoid stenosis, inflammatory cell infiltration and congestion), which can be employed as an non-invasive indicator of severity in microcirculatory disturbance, but this parameter was not sensitive enough in early detection of HIRI. D value was a biomarker sensitive in early detection of HIRI, but the correlation with serum biochemical parameters of hepatic injury was not significant during reperfusion. In HIRI, microcirculatory disturbance might be the major reason that accounted for changes in ADC value. ADC value can reflect the severity of microcirculatory disturbance and it’s sensitive in early detection of HIRI in the meantime. |
PDF Full Text Request