Comparison between Intravoxel Incoherent Motion and Splenic Volumetry to Predict Hepatic Fibrosis Staging in Preoperative Patients

Intravoxel incoherent motion (IVIM) and splenic volumetry (SV) for hepatic fibrosis (HF) prediction have been reported to be effective. Our purpose is to compare the HF prediction of IVIM and SV in 67 patients with pathologically staged HF. SV was divided by body surface area (BSA). IVIM indices, such as slow diffusion-coefficient related to molecular diffusion (D), fast diffusion-coefficient related to perfusion in microvessels (D*), apparent diffusion-coefficient (ADC), and perfusion related diffusion-fraction (f), were calculated by two observers (R1/R2). D (p = 0.718 for R1, p = 0.087 for R2) and D* (p = 0.513, p = 0.708, respectively) showed a poor correlation with HF. ADC (p = 0.034, p = 0.528, respectively) and f (p < 0.001, p = 0.007, respectively) decreased as HF progressed, whereas SV/BSA increased (p = 0.015 for R1). The AUCs of SV/BSA (0.649–0.698 for R1) were higher than those of f (0.575–0.683 for R1 + R2) for severe HF (≥F3–4 and ≥F4), although AUCs of f (0.705–0.790 for R1 + R2) were higher than those of SV/BSA (0.628 for R1) for mild or no HF (≤F0–1). No significant differences to identify HF were observed between IVIM and SV/BSA. SV/BSA allows a higher estimation for evaluating severe HF than IVIM. IVIM is more suitable than SV/BSA for the assessment of mild or no HF.


Introduction
Hepatic fibrosis (HF) is an important factor in patients with chronic liver disease and those requiring a surgical operation of the liver [1] because HF may lead to surgical restrictions and affect the patient's prognosis.A liver biopsy for histopathological assessment is commonly performed to stage HF.However, a biopsy can cause complications such as haemorrhage and infection, as well as inherent drawbacks, including sampling error.Therefore, non-invasive imaging-based methods [2][3][4] have been developed to assess HF.Diffusion-weighted imaging (DWI) represents the motion or diffusivity of the molecules.The apparent diffusion-coefficient (ADC) is calculated as the diffusivity and consists of substantial molecular diffusion in solid tissues and molecular movements in the vascular microcirculation (perfusion) [5].Intravoxel incoherent motion (IVIM) is used for microscopic translations which occur in voxels on MRI [6], and the restricted diffusion observed in patients with cirrhosis may be related to D* variations according to Luciani et al. [7] and Ichikawa et al. [4].Several researchers have reported that the ADC value is significantly reduced compared to the non-cirrhotic liver [8][9][10].The measurement of splenic volumetry (SV) is simple, SV/body surface area (BSA) is a better predictor of HF, and SV is related to the severity of HF [11,12].Although analyses by IVIM and SV are promising techniques used for staging HF, to our knowledge, no comparative study of these analyses has been conducted.This study aimed to compare IVIM imaging and SV measurements for staging HF.

Patients
This retrospective study was performed in accordance with the principles of the Declaration of Helsinki, and approved by the relevant Institutional Review Board of Nihon University Itabashi Hospital (RK-210413-9).Written informed consent was waived by the Institutional Review Board (Nihon University Itabashi Hospital, Clinical Research Judging Committee) because of the retrospective nature of this study.However, informed consent was obtained for MR and CT examinations and for publication of the patient's images.This study included 88 consecutive patients who underwent liver MRI including 10-b-values DWI from November 2018 to January 2020.The inclusion criteria for our study were as follows: consecutive patients aged ≥ 20 years who were candidates for initial liver resection for liver tumour and those with available IVIM and dynamic CT imaging data within 3 months prior to liver resection.The exclusion criteria were as follows: patients without liver resection because of clinical or biochemical evidence of decompensated liver function (Child-Pugh classification C, ICG-R15 ≥ 35%, or serum total bilirubin level ≥ 2.0 mg/dL), tumour status, or portal hypertension (including the presence of high-risk oesophageal varices), and those who had already undergone hepatectomy or splenectomy prior to the IVIM-MRI and CT.

IVIM Imaging
Discovery 750 3.0T (GE Medical Systems, Chicago, IL, USA) with GEM Body Array (GEM Anterior Array + GEM Posterior Array), Ingenia 3.0T, and Achieva 1.5T Nova (Philips Healthcare, Best, The Netherlands) with SENSE Torso/Cardiac Coil were used for IVIM imaging, which was acquired in the transverse plane by respiratory-triggered fat-saturated spin echo-echo planar imaging.Motion-proving gradient pulses were applied concurrently in three directions (x, y, and z).The IVIM parameters are listed in Tables 1 and 2. The calculation for IVIM was performed using the Synapse Vincent software version 5.5 (Fujifilm Medical, Tokyo, Japan), which provided the D, D*, ADC, and f parameters mapped on a pixel-by-pixel basis (Figure 1).The calculation for IVIM was performed using the Synapse Vincent software version 5.5 (Fujifilm Medical, Tokyo, Japan), which provided the D, D*, ADC, and f parameters mapped on a pixel-by-pixel basis (Figure 1).The decrease in signal intensity of the hepatic lesions was compared with that of the spleen.We estimated signal attenuation using the following equation [6,13]: where D and D* are the slow diffusion-coefficients related to molecular diffusion, and the fast diffusion-coefficient related to perfusion in micro-vessels, respectively, and f is the perfusion-related diffusion fraction.
To obtain IVIM parameters, three regions of interest (ROIs) were placed in the liver parenchyma (left lobe, anterior right lobe, and posterior right lobe), excluding artefacts, large intrahepatic vessels, and liver tumours, by two radiologists (T. A. and M. O.) with 5 years and 25 years of experience in abdominal radiology, who were blinded to the clinical, surgical, and pathological results.The measurements from three ROIs were averaged and used as the patient results.The region of interest (ROI) was drawn to prevent a peripheral liver zone of <1 cm.Each radiologist independently placed the ROIs.Circular ROIs with a size of 10 mm were placed at the hilum level of the liver.The measurements were recorded and compared separately.

CT Volumetry Analysis for Spleen
SV was measured using a viewer workstation (Synapse 3D ® ver5.5, Fujifilm Medical) to perform 3D reconstruction using dynamic liver CT data.An abdominal radiologist (T.A) with 5 years of experience made CTV images of the spleens.Without recognition of other organs in close proximity, SV measurements (Figure 2) were performed at approximately 1 min using the SAI viewer ® (Fujifilm Medical), which uses Deep Learning technology.The SV was corrected using BSA (TLV/BSA, SV/BSA) [14].BSA was calculated using the Dubois formula (BSA [m 2 ] = 0.007184 × height [cm] 0.725 × weight [kg] 0.425 ) [14].

Pathology
From surgically resected specimens for all patients, the pathological evaluation for HF was performed by two pathologists using the New Inuyama Classification [15]: F0, no fibrosis; F1, fibrous portal expansion; F2, bridging fibrosis; F3, bridging fibrosis with architectural distortion; F4, cirrhosis.

Pathology
From surgically resected specimens for all patients, the pathological evaluation for HF was performed by two pathologists using the New Inuyama Classification [15]: F0, no fibrosis; F1, fibrous portal expansion; F2, bridging fibrosis; F3, bridging fibrosis with architectural distortion; F4, cirrhosis.

Statistical Analysis
The intraclass-correlation-coefficients (ICCs) were used for interrater reliability, with an ICC of <0.50 defined as poor; 0.50-0.74as moderate; 0.75-0.90as good; and ≥0.90, excellent [16].The mean and standard deviation of the IVIM and SV/BSA were calculated for each group.The correlation between the results of each imaging analysis and the HF stage was assessed using Spearman's rank correlation.Receiver operating characteristic curve analysis was performed to determine the accuracy of IVIM and SV/BSA in staging HF.The highest AUCs for IVIM and SV/BSA were compared for HF grading using the Delong test.SPSS Version 27.0 (IBM Corp., Armonk, NY, USA) was used for analysis.

Patients
This study included 88 patients who underwent hepatic surgery; however, 20 patients were excluded because the interval between IVIM and surgery was longer than the inclusion criteria or image deficiency obtained.One of the remaining 68 patients was excluded from the study because he had undergone a splenectomy.Finally, 67 patients were included in the study (Figure 3).The patient characteristics are shown in Table 3.

IVIM Analysis for Liver
The IVIM parameters, such as D, D*, ADC, and f, at each fibrosis grade determined by the two observers are summarised in Table 4.The hierarchisation for each HF stage was clear in the ADC for R1 and f for both R1 and R2.IVIM_f was significantly associated with the grading of HF by both observers (p < 0.001 and p = 0.007 by Spearman's rank correlation coefficient test).ADC was significantly correlated with HF (p = 0.034) in only one observer.Other parameters of IVIM showed no significant differences in HF grading.Box-and-whisker diagrams for each parameter measurement are shown in Figure 4.When the degree of agreement between the two observers was evaluated using the intra-class correlation coefficient, poor reproducibility of IVIM measurements between the two observers was observed due to an ICC of less than 0.7, as shown in Table 5.

IVIM Analysis for Liver
The IVIM parameters, such as D, D*, ADC, and f, at each fibrosis grade determined by the two observers are summarised in Table 4.The hierarchisation for each HF stage was clear in the ADC for R1 and f for both R1 and R2.IVIM_f was significantly associated with the grading of HF by both observers (p < 0.001 and p = 0.007 by Spearman's rank correlation coefficient test).ADC was significantly correlated with HF (p = 0.034) in only one observer.Other parameters of IVIM showed no significant differences in HF grading.Box-andwhisker diagrams for each parameter measurement are shown in Figure 4.When the degree of agreement between the two observers was evaluated using the intra-class correlation coefficient, poor reproducibility of IVIM measurements between the two observers was observed due to an ICC of less than 0.7, as shown in Table 5.  Spearman's correlation analysis was used to assess the correlation between each measurement and fibrosis stage.Data are persented as mean ± standard deviation.ρ, Spearman's correlation coefficient; IVIM intravoxel incoherent motion; D, molecular diffusion; D*, fast diffusion-coefficient related to perfusion in micro-vessels; ADC, apparent diffusion-coefficient; f, perfusion-related diffusion fraction; SV/BSA, ratio of splenic volume to body surface area; R1, Observer 1; R2, Observer 2.

CT Volumetry Analysis for Spleen
SV/BSA increased with the exacerbation of HF.The results of the SV/BSA measurements at each fibrosis grade are shown in Table 4. SV/BSA was significantly associated with HF grading (p = 0.015, Spearman's rank correlation coefficient test).
A box-and-whisker diagram for each parameter measurement is shown in Figure 4.The hierarchisation for each stage of HF was clear in the SV/BSA.

ROC Analysis
The ability of IVIM_f and SV/BSA to differentiate HF, which were both significantly correlated with the severity of HF in the assessment of the two observers in IVIM, was investigated using ROC analysis.

Discussion
Previous studies have shown that HF is associated with decreased D*, f, and ADC values [17,18].According to Dyvorne et al. [19], the IVIM parameters f and D decrease as the HF progresses.In our results, f was especially useful for estimating HF, especially for no or mild HF.Previous reports have shown that D*, f, and ADC values were reduced as HF increased [7,17,18].Thus, as to which parameter of IVIM best reflects HF remains controversial.Considering the ICC confidence intervals, the ICCs of ADC and f were moderate (below 0.50-0.74),whereas those of D and D* were poor (below 0.50).Manual registration of the ROI in IVIM is prone to interobserver variability and errors [20].
The AUC of SV/BSA was good (0.76-0.83) to estimate HF in a previous report [11], but the AUC was 0.628-0.698 in the present study was lower than that of the previous study.This may be attributed to the fact that portal hypertension was mild in most cases, and splenomegaly was sufficient to buffer portal blood flow because SV was associated with the degree of portal hypertension.
There were significant differences in patient background factors such as age, Plt, Alb, ALBI, APRI, and FIB-4.A previous report [11] showed lower SV and SV/BSA than the current report, although SV and SV/BSA did not show significant differences (Supplementary Table S1; this Table is a comparison using the original data from our previously published paper [11]).Categorical variables, such as sex, HBV, HCV, alcoholic liver disease, and Child-Pugh classification score, showed no significant differences between the previous [11] and the current study.However, the previous report showed significantly higher APRI and FIB-4 than the current report; thus, the previous study included patients with relatively poor liver function.
IVIM showed a large difference between the two measurements in two observers; the location of the ROI and different ROI sizes were likely to cause differences in measurements.This would result in a quantitative assessment with low reproducibility.CTVs, such as splenic volumetry, can be accurately assessed using artificial intelligence technology, the so-called SAI, because the imaging workstation enables semi-automatic and rapid reconstruction.
Spleen enlargement is caused by portal hypertension or cirrhosis [21].We believe that SV/BSA is suitable for estimating HF and can contribute to safe operative management because highly fibrotic livers are known to be at risk for severe complications after surgery [22].
In particular, SV/BSA is a better predictor of HF than IVIM as an imaging parameter that can predict severe HF (≥F3).This result is in line with the previous considerations that SV/BSA is superior to extracellular volume fraction (ECV) in estimating HF [11].The correction by BSA was used because SV may be affected by differences in patient body size, although SV is unrelated to patient height and weight [23].SV/BSA was useful for predicting severe HF (≥F3), with an AUC of 0.698.This is consistent with the results of previous reports [24,25].Without the need to use a liver biopsy or MR elastography, CT volume biomarkers can be obtained retrospectively with routine scans obtained for other

Figure 1 .
Figure 1.Intravoxel incoherent motion measurement.By measuring three regions of the liver (anterior and posterior segments of the right lobe and lateral segment of the left lobe), IVIM parameters were measured in this patient (F1).This case showed images with an ROI placed in the anterior region of the right lobe of the liver.A model of signal strength variation in IVIM analysis was made

Figure 1 .
Figure 1.Intravoxel incoherent motion measurement.By measuring three regions of the liver (anterior and posterior segments of the right lobe and lateral segment of the left lobe), IVIM parameters were measured in this patient (F1).This case showed images with an ROI placed in the anterior region of the right lobe of the liver.A model of signal strength variation in IVIM analysis was made from Synapse Vincent software.White line is actual data obtained.Red line is the IVIM nonlinear regression fit providing slow diffusion-coefficient related to molecular diffusion (true diffusion-coefficient) as D, fast diffusion-coefficient related to perfusion in micro-vessels (pseudodiffusion-coefficient) as D*.Yellow line is the monoexponential fit providing apparent diffusion-coefficient (ADC).

Figure 2 .
Figure 2. Three-dimensional reconstruction image of the spleen.Splenic hilum view; The volume of the spleen was 74 mL in this patient (F1).

Figure 2 .
Figure 2. Three-dimensional reconstruction image of the spleen.Splenic hilum view; The volume of the spleen was 74 mL in this patient (F1).

14 Figure 3 .
Figure 3. Patient flowchart.A total of 88 patients with liver surgery received both IVIM of MRI and liver dynamic CT.Twenty patients with image deficiency and long intervals between IVIM and surgery were excluded.One of the remaining 68 patients was excluded from the study because he had a splenectomy.Finally, 67 patients were included in the study.

Figure 3 .
Figure 3. Patient flowchart.A total of 88 patients with liver surgery received both IVIM of MRI and liver dynamic CT.Twenty patients with image deficiency and long intervals between IVIM and surgery were excluded.One of the remaining 68 patients was excluded from the study because he had a splenectomy.Finally, 67 patients were included in the study.

Figure 4 .
Figure 4. Box-and-whisker diagrams for each parameter measurement.Box-and-whisker diagrams for each parameter measurement of IVIM and SV/BSA were shown.R1, Observer 1; R2, Observer 2; D, molecular diffusion; D*, fast diffusion-coefficient related to perfusion in microvessels; ADC, apparent diffusion-coefficient; f, perfusion-related diffusion fraction; SV/BSA, ratio of splenic volume to body surface area; F grade, hepatic fibrotic grade.

Figure 4 .
Figure 4. Box-and-whisker diagrams for each parameter measurement.Box-and-whisker diagrams for each parameter measurement of IVIM and SV/BSA were shown.R1, Observer 1; R2, Observer 2; D, molecular diffusion; D*, fast diffusion-coefficient related to perfusion in microvessels; ADC, apparent diffusion-coefficient; f, perfusion-related diffusion fraction; SV/BSA, ratio of splenic volume to body surface area; F grade, hepatic fibrotic grade.

Table 4 .
IVIM and CT Volumetry for each liver fibrosis stage.

Table 4 .
IVIM and CT Volumetry for each liver fibrosis stage.