ESGAR 2019 Book of Abstracts

REVIEWING PANEL 6 O. Akhan, Ankara/TR


TABLE OF CONTENTS
Scientifi c Sessions, Thursday, June 6 (SS 1 -SS 5) 8-21 Scientifi c Sessions, Friday, June 7 (SS 6 -SS 10) 22-34 Scientifi c Sessions, Saturday, June 8  35-45 Authors' Index 47-51 Purpose: To investigate the effect of nodular regenerative hyperplasia (NRH) on the liver stiffness measurement (LSM) obtained with magnetic resonance elastography (MRE). Material and methods: This retrospective, IRB-approved study included 39 subjects with NRH (Group 1) and no or minimal fibrosis (≤F2), a control group (Group 2) made of 30 subjects with non-advanced fibrosis (F0-F2) and a control group (Group 3) made of 30 subjects with advanced fibrosis (F3-F4), all with available MRE. LSM was measured in each subject along with the assessment of hepatic morphological features of cirrhosis and signs of portal hypertension. The significance of the difference in mean LSM between Groups 1 and 2 and between Groups 1 and 3 was evaluated using the Mann-Whitney U test. The difference in the distribution of imaging features among groups was assessed using the Pearson χ 2 or Fisher exact test. Statistical significance was set at P<0.05.

SS 1.8
Evaluation of liver fibrosis and necro-inflammation using stiffness and dispersion slope from 2D-shear wave elastography D.H. Lee, E.S. Lee; Seoul/KR Purpose: To prospectively evaluate whether liver stiffness (LS) and shear wave (SW) dispersion slope obtained from 2D-shear wave elastography (SWE) could evaluate fibrosis stage and necro-inflammatory activity (NIA) using histopathology as a standard reference. Material and methods: We prospectively enrolled 114 patients who underwent percutaneous liver biopsy. All patients underwent SWE examination just before biopsy, and LS (kilopascal [kPa]) as well as SW dispersion slope ([m/s/]/kHz) were obtained. On histopathologic examination, fibrosis stage (F0-F4) and NIA (A0-A4) were assessed. Multivariate linear regression analysis was done to determine the significant affecting factors for LS and SW dispersion slope. Diagnostic performance of each variable in staging fibrosis and grading NIA was assessed using receiver operating characteristic (ROC) analysis. Results: Both LS and SW dispersion slope were significantly different among the different fibrosis stage and NIA grade (P<0.001). Only fibrosis stage was significantly associated with LS (P<0.001). However, both fibrosis stage (P<0.001) and NIA grade (P<0.001) were significant determinant factors for SW dispersion slope. LS provided significantly better diagnostic performance in detecting cirrhosis than SW dispersion slope (area under curve [AUC]: 0.950 vs. 0.812, P=0.020). The optimal cut-off LS were 5.4 kPa, 7.8 kPa, 9.4 kPa and 12.2 kPa, respectively, for F≥1, F≥2, F≥3 and F=4. Diagnostic performance of SW dispersion slope in detecting A≥3 of NIA was marginally higher than LS (AUC: 0.782 vs. 0.836, P=0.065). Conclusion: LS obtained from 2D-SWE provided quite a good diagnostic performance in evaluating liver fibrosis stage, and SW dispersion slope might help assess the NIA grade.
Results: All liver segments demonstrated statistically significant reductions in FF (p<0.01), with an average 12-week reduction across all liver segments in absolute FF of -5.9% (p=0.008) and relative FF of -28.2% (p=0.002). The average 12-week change in absolute and relative LIC across all liver segments was -0.87umol/g (p=0.1197) and -2.8% (p=0.3729), respectively.
Conclusion: MRI-PDFF demonstrated significant changes in absolute and relative liver FF 12 weeks after DMR. There was no significant change in LIC 12 weeks after DMR. Further longitudinal assessment of liver FF and LIC in this cohort will assess potential sustained reductions in liver FF, longitudinal effects on LIC and correlation with biochemical markers of diabetes.

SS 1.10
CT-based liver surface nodularity for the detection of clinically significant portal hypertension: defining measurement quality criteria R. Sartoris Material and methods: 77 consecutive cirrhotic patients (50 males, mean 62±6 years), including 30 with clinically significant portal hypertension (CSPH), 39%), underwent CT and hepatic venous portal gradient. Three independent readers performed 15 LSN measurements/patient on contrast-enhanced CT using a dedicated software. Quality criteria were identified to maximize LSN accuracy while minimizing inter-and intrareader variability. LSN was computed both as median and mean using 1 to 15 individual measurements. Accuracy of LSN for the diagnosis of CSPH was assessed using receiver operating characteristic (ROC) curve analysis. Variabilities were assessed by computing the intraclass correlation coefficient (ICC), coefficient of variation (CV) and Bland-Altman plot (BA). Results: Using 1 to 15 individual measurements, mean and median LSN showed similar diagnostic performance (area under (AU) ROCs from 0.79±0.05 to 0.91±0.04 and from 0.86±0.04 to 0.91±0.03, all pair-wise comparisons p>0.05). AUROCs of mean LSN increased from 1 to 8, and plateaued from 8 to 15 measurements. Using mean LSN values, inter-and intra-reader variability decreased when using 1 to 15 individual measurements, but improvement was marginal > 8 measurements. With 8 measurements, intra-and inter-observer variability was low   Purpose: To determine the performance of texture analysis (TA) in predicting tumoral response in colon-rectal cancer (CRC). Material and methods: Forty consecutive CRC patients were prospectively enrolled and underwent 3T-MRI pre-and post-chemo-radiotherapy (CRT). After surgery, the histological results were considered the reference standard. A region-of-interest was manually drawn around the tumour area on axial T2-w slice and analyzed with a dedicated software (TextRad) extrapolating the following parameters: skewness, kurtosis, entropy, and mean value of positive pixels (MPP). Non-parametric Mann-Whitney U test was used to compare TA parameters and the response rate among complete responders (CR), partial responders (PR), and non-responders (NR) before and after CRT. Receiver operating characteristic (ROC) curves were used to assess the discriminatory power of TA to predict complete response to CRT. Results: Thirteen patients (32.5%) showed CR, twenty-two patients (55%) showed PR and five patients (12.5%) were classified as NR. After CRT, CR showed significant reduction of entropy ( Purpose: To compare the diagnostic performance of MRI (T2-HR), diffusionweighted imaging (DWI) and their combination in determining complete response to neoadjuvant chemoradiotherapy (LCCRT) in patients with locally advanced rectal cancer. Material and methods: In this retrospective study of 312 patients (mean age 47.3+/-14.3 years, range 19-86 years, M:F= 204:108) with locally advanced rectal cancer treated with LCCRT and surgery, two radiologists blinded to the surgical histopathology, independently reviewed MRI performed pre-and post-LCCRT. Diagnostic performance of morphological assessment on T2-HR images, pattern on DWI and combination of both in determining complete (CR) versus incomplete (IR) response was assessed with pathological response as the reference standard. Either restoration of rectal wall morphology or thin band of fibrosis less than normal rectal wall thickness or complete absence of restricted diffusion was considered as MRI-CR. Level of confidence was recorded on a scale of 1 to 5, 5 being very confident.
Results: Rate of pathological CR was 16%. Pattern-based interpretation of DWI had the highest accuracy (AUC = 0.822), followed closely by combined interpretation of DWI and T2-HR images (AUC of 0.806), p<0.001. Interobserver agreement was substantial (k=0.688) for combined DWI + T2-HR images and moderate (k=0.402) for DWI alone with both observers exhibiting significantly higher confidence with the combined approach, p<0.05. Morphology-based assessment on T2-HR alone had poor sensitivity of 8% and accuracy (AUC= 0.532) and interobserver agreement (k=0.231), p>0.05. Conclusion: MR-complete response to neoadjuvant LCCRT can be determined with high accuracy and confidence with a combination of DWI signal pattern and morphology on T2-HR images.

SS 2.3
Artificial intelligence automatic identification of complete-and non-responders using texture analysis of rectal cancer 3T MR images performed before, during and after neoadjuvant chemoradiotherapy M. Zerunian 1 , R. Ferrari 1 , C. Mancini Terracciano 1 , C. Voena 1 , M. Rengo 2 , R. Paramatti 1 , D. Caruso 1 , R. Faccini 1 , A. Laghi 1 ; 1 Rome/IT, 2 Latina/IT Purpose: To evaluate an artificial intelligence (AI) model for classification of complete (CR) and non-responder (NR) patients with rectal cancer treated with neoadjuvant chemo-radiotherapy (CRT). AI input is based on texture analysis of high-resolution 3T MR images performed before, during and after CRT. Material and methods: 55 consecutive patients with rectal cancer were prospectively enrolled in this study. Patients underwent 3T T2-weighted (T2w) MRI before, during and after CRT; volumetric regions of interest (VOI) around the tumor were manually drawn. All patients underwent complete surgical total mesorectal excision and the gross specimen was used as reference standard. Two AI models were built with the most statistically significant features training a random forest (RF) classifier on 28 patients (training cohort). The model performances were estimated on 27 patients (validation cohort) using a ROC curve analysis and a decision curve analysis.
Results: Textural analysis points to a lower intra-tumor heterogeneity at the preand during-treatment stage for CR patients with a characteristic time evolution of some of the textural features. The obtained AI model shows good discriminatory ability with a ROC area under curve (AUC) of 0.94 (95% CI: 0.89,0.99) in the validation cohort. The discriminatory power of the AI model built for NR discrimination has a ROC AUC of 0.85 (95% CI: 0.80,0.90). Decision curve analysis confirms clinical usefulness of the models.
Conclusion: AI models based on textural parameters of MR images of patients with rectal cancer taken before, during and after CRT show good performances for stratification of response to therapy.

SS 2.4
18F-fluoro-2-deoxy-D-glucose-avid presacral soft tissue mass in previously treated rectal cancer: diagnostic outcome and additional value of MRI, including diffusion-weighted imaging J.P. Pennings Material and methods: This retrospective study included 38 patients who completed primary rectal cancer treatment and who presented with a suspicious FDG-avid presacral lesion on PET/CT. Twenty-seven patients also underwent MRI, of whom 24 with DWI. PPV of FDG-PET/CT and additional value of MRI, including DWI, for the diagnosis of recurrent presacral cancer were determined.
Results: The PPV of PET/CT with an FDG-avid presacral lesion for the diagnosis of locally recurrent rectal cancer was 58% (22/38). Air in the FDG-avid presacral lesion, as visible on the CT component of the PET/CT examination, favoured the diagnosis of benign presacral tissue with a sensitivity of 56.3% (9/16) and a specificity 81. 8% (18/22). Areas under the receiver operating characteristic curve (AUCs) of MRI without DWI for the diagnosis of locally recurrent rectal cancer in FDG-avid presacral tissue were 0.765 and 0.840, for observers 1 and 2. AUCs of MRI with DWI were 0.803 and 0.811, for observers 1 and 2. There were no significant differences among any of these AUCs (P=0.169 to 0.906).
Conclusion: FDG-PET/CT has a poor PPV for locally recurrent rectal cancer in the presacral space. The observation of air in the FDG-avid presacral lesion and additional MRI assessment are diagnostically helpful, without a significant additional value of DWI.

SS 2.5
Diagnostic value of dynamic perfusion MRI in patients with locally advanced rectal cancer in the assessment of chemo-radiation treatment: relation to tumor regression grade at histology P.P. Arcuri 1 , A.K. Sikora 1 , S. Roccia 1 , G. Fodero 1 , V. Aiello 1 , C. Bertucci 2 , E. Mazzei 1 , D. Laganà 1 ; 1 Catanzaro/IT, 2 Buckingham/UK Purpose: To investigate the value of dynamic contrast-enhanced perfusion MRI parameters in the evaluation of the response to chemo-radiation therapy in patients with locally advanced rectal cancer in comparison with histology. Material and methods: We evaluated, retrospectively, thirty-eight patients affected by rectal adenocarcinoma (confirmed by biopsy) studied with dynamic contrast-enhanced MR sequence, before (MR-T0) and after chemo-radiation therapy (MR-T1). The protocol included T1 gadolinium-enhanced THRIVE sequence. A region of Interest (ROI) was manually drawn on tumor tissue and normal rectal wall. The following parameters were calculated and statistically analyzed: maximum enhancement (ME), relative enhancement (RE), time to peak (TTP), wash-in rate (W-inR) and wash-out rate (W-outR). Data were expressed in terms of median value ± range. Kruskal-Wallis non-parametric test was performed. Sensitivity, specificity and accuracy were assessed. Fisher's exact test was used to evaluate statistical significance. A p value <0.05 was considered statistically significant. The biopsy of the masses was the gold standard of reference. Perfusion parameters were related to pathologic tumor regression grade (Mandard's criteria; TRG1=complete regression, TRG5=no regression).

Coimbra/PT
Purpose: To describe the process of developing a structured report tool in our center reporting MRI in patients at risk for HCC. Material and methods: First, a preliminary list with relevant questions was presented at the liver disease multidisciplinary team (MDT) meeting. After discussion, a final list of questions and answers was accepted by consensus. We agreed to use the Liver Imaging Reporting and Data System (LI-RADS) v2018 lexicon for reporting the MRI findings. Bioimaging engineers developed a webbased platform using Angular® with a Mongo Database to save data. All data were stored in an ancillary server located at the hospital. Results: The developed structured report tool includes mainly a customized checkbox for each item. Following sections were included: 1) clinical data (age, gender, Child-Pugh, model for end-stage liver disease (MELD), performance status test, etiology of chronic liver disease); 2) technical MRI aspects (magnetic field strength, UKCA used, quality of study); 3) chronic liver disease findings (morphologic cirrhosis changes, steatosis, inflammatory changes, iron deposit, fibrosis and portal hypertension findings); 4a) non-treated focal liver lesions (major and minor criteria based in LI-RADS v2018); 4b) treated focal liver lesion; 5) other findings not related to liver disease. After completing all items, the tool offers a final report with LI-RADS category of each lesion and suggests an EASL staging and treatment to start discussion at MDT meeting. Purpose: Studies have found a decrease in negative laparotomy and appendiceal perforation rate when CT imaging was used in selected patients with suspected appendicitis. The aim of our study was to determine the cost-effectiveness of performing diagnostic imaging as part of the preoperative assessment of patients with right iliac fossa pain (RIF). Material and methods: We have developed a local pathway for patients with RIF pain. On this pathway, if the examination is suggestive of appendicitis, and the C-reactive protein >10mg/L or white cell count >10.5 x 10 3 /uL, then an US focussed on the assessment of the appendix is performed. If this is equivocal, the patient automatically has a low-dose CT abdomen and pelvis.
We have calculated the cost of managing 1500 patients pre-pathway and 1500 post-pathway. Results: After introducing imaging assessment in this patient group, the negative appendicectomy rate decreased to <4% (from 36%) in our institution. The cost of an US scan is £32, of a CT scan is £118 with a laparoscopic appendicectomy costing £2358 (if complicated £3494, rate of complicated appendicitis/ complications 10-20%). Pre-pathway, we performed 800 US scans, 100 CT scans and 825 laparoscopies (of which 82 were complex), resulting in a cost of £2,075,902. During the pathway, 1500 US scans, 1000 CT scans and 450 laparoscopies (45 complex) were performed, resulting in a cost of £1,278,220. Conclusion: Introducing an imaging pathway for patients with suspected appendicitis has saved our institution ~£800,000 per annum and has significantly reduced the negative appendicectomy rates. Material and methods: This retrospective study was approved by our institutional review board, and informed consent was waived. Contrast-enhanced CT studies in 189 patients with adhesive SBO that was initially treated conservatively were reviewed independently by two radiologists to identify CT findings predicting failure of non-surgical treatment. The findings included the location of transition zone, number of beak sign, maximum diameter of bowel dilatation, C-loop, fecal sign, whirl sign, bowel wall thickening, mesenteric haziness, amount of mesenteric or peritoneal fluid and submucosal edema. These findings were statistically compared according to the success or failure of non-surgical treatment.
Results: Nonsurgical treatment succeeded in 144 patients (76.2%) and failed in 45 patients (23.8%). At univariate analysis, anterior location of adhesion and a fecal sign were associated with success of non-surgical treatment, whereas two beak signs or more, the presence of c-loop, mesenteric haziness and moderate amount of mesenteric fluid were associated with failure of non-surgical treatment. At multivariate analysis, lack of fecal sign, mesenteric haziness and moderate amount of mesenteric fluid were independent findings predicting failure of non-surgical treatment, with odds ratios of 5.08, 7.77 and 6.74, respectively. Conclusion: The lack of fecal sign, mesenteric haziness and moderate amount of mesenteric fluid are useful indicators to predict failure of non-surgical treatment in patients with non-strangulated adhesive SBO. Material and methods: We retrospectively evaluated CT scans of 146 patients that acceded our emergency department with acute (< 3 days) abdominal pain. Two radiologists (1 expert and 1 young) independently and blindly evaluated both unenhanced and enhanced (venous phase) CT scans, and for each of them they were asked to formulate a final diagnosis. These diagnoses were classified into 14 categories and used for calculating intra-observer agreement (Cohen's kappa), diagnostic accuracy, sensitivity, specificity, positive and negative predictive value.
Results: Intra-observer agreement between unenhanced and enhanced scans was approximately 0.8 (good/very good) both for the expert and the young radiologist. Diagnostic accuracy increased from 78.8% to 83.6% after contrast injection; however, if sensitivity increased from 78.4% to 90.7%, specificity unexpectedly decreased from 79.4% to 71.8%. Contrast medium incremental diagnostic value, defined as the percentage of cases in which the correct diagnosis was identified only on enhanced scans, was 11.6% for the expert radiologist but only 6.8% for the young. Predictive value stratified on the differential diagnosis was very variable. Diagnoses with the best predictive value (> 90%) were diverticulitis, ileus, cholecystitis, pancreatitis, obstructive nephropathy and muscular hematoma.
Conclusion: The level of agreement between unenhanced and enhanced scan diagnoses is high. In selected cases when the diagnosis on unenhanced scan matches the diagnostic suspect, contrast medium injection may not be necessary. Material and methods: We reviewed more than 7000 CT scan examinations performed with new 256 row CT scan (iCT Elite, Philips) combined with new IMR algorithm, in the abdominal district in patients referred to our ERD with different clinical emergency settings (trauma, abdominal pain, diverticulitis, etc). A control group of 5500 patients underwent standard dose abdominal CT protocol (100 kV; automated mAs) on 256 row CT scanner. For each exam, we analyzed the CT dose index (CTDI, expressed in mGray) and the dose length product (DLP, expressed in mGray.cm) and compared with the dose of our previous CT scan equipment.

Results:
The mean values of CTDI were lower with IMR compared with our previous CT equipment: abdomen 17.7 vs 6.3 (-64%) and whole body 16.8 vs 12.4 (-26%). In the same way, the total DLP were lower with IMR compared with standard CT: abdomen 911.0 vs 361.5 (-60%) and whole body 1118.0 vs 893.0 (-20%). The subjective image quality of low kV IMR-abdominal study was also higher compared with standard dose CT studies. Conclusion: Low-dose CT (80 kV and automated mAs modulation) abdominal study reconstructed with model-based algorithm represents a feasible protocol for the evaluation of abdominal disease in the emergency setting, achieving high image quality with extremely low noise and a significant dose reduction within adequate reconstruction times (less than 120 seconds).
In one case, we had incomplete sealing and splenic artery's hilar branch fissuration during the procedure and thus treated with coils and ruled out of the follow-up. We reported complete aneurysmal exclusion in 20/20 patients at follow-up. In 4 cases, we observed stent occlusion, although without organrelated ischemia; in 3 of these 4 cases, stent occlusion was associated with stent graft migration: one inside the gastric antrum at the 36th month during follow-up (confirmed with endoscopy) and two in the lesser omentum at the 12-and 60-month follow-up.
Conclusion: Endovascular treatment of visceral artery aneurysms and pseudoaneurysms with stent grafts is a safe and effective technique with good immediate-and long-term results. Extravascular migration of the stent graft after its occlusion is a possible late complication. According to our evidence, we recommend follow-up even in case the stent is occluded.

SS 4.10
Endovascular management of hepatic artery thrombosis on the first post-operative day after living donor liver transplantation O. Abdelaziz, S. Emadeldin; Cairo/EG Purpose: Our aim was to asses the feasibility and potential complications of endovascular intervention for the management of early hepatic artery thrombosis (HAT) on the first post-operative day after living-donor liver transplantation (LDLT). Material and methods: This is a retrospective review of 668 recipients who underwent LDLT between August 2001 and August 2016 at 3 centers. Endovascular interventions were performed using standard catheter techniques. Thrombolysis was performed using tPA or streptokinase, whereas angioplasty and stent placement were performed if there was an underlying stricture. Results: Early HAT within 2 weeks postoperatively occurred in 30 patients (4.5% Material and methods: Among 184 patients from the SARAH trial who received SIRT, 121 and 109 were included in dose/survival and dose/tumor response analyses, respectively; CT, technetium-99 m macroaggregated albumin -single-photon emission tomography/CT ( 99m Tc-MAA-SPECT/CT) and 90 YSPECT/ PET were centralized. Tumor-absorbed dose was computed using 99m Tc-MAA-SPECT/CT. Visual agreement between CT-MAA-90 Y (optimal, suboptimal, not optimal), overall survival (OS) and tumor response at 6-month follow-up CT (RECIST 1.1) was assessed.

SS 5.2
Does the tumoral density heterogeneity after selective internal radiation therapy predict local progression in colorectal hepatic metastases?
The purpose of the present study is to investigate whether the appearance of tumor density heterogeneity after selective internal radiation therapy (SIRT) on CT can predict local progression of colorectal cancer liver metastases (CRCLM). Material and methods: After approval of the IRB, forty-five CRCLM among nineteen patients treated with 90 Y-radioembolization and imaged with CT at the portal venous phase performed at baseline and 6-8 weeks after the treatment were retrospectively analyzed using TumourMetrics in-house software. The longest tumor diameter, the volume, the mean density, the standard deviation and the kurtosis of the density of the non-enhancing and enhancing portions of the 3 largest lesions were recorded at each time point. Adapted RECIST 1.1 criteria were applied to assess treatment response. Non-parametric tests were used to assess differences between responders and non-responders.
Results: 36 lesions showed no progression and 9 showed local tumor progression at first or second follow-up. In lesions showing progression, the decrease of the standard deviation of density in the non-enhancing portion of the tumor was more pronounced compared to the responding lesions (-27% vs -15%) (p.=.0.002). The threshold of -17% allows to obtain a sensitivity of 78% and specificity of 89% with an area under the curve of 83% and a Youden index of 67%. No other significant differences were found. Conclusion: The decrease of standard deviation of density in the non-enhancing tumor portion at CT follow-up after SIRT seems to be able to predict local progression.

SS 5.3
Yttrium 90-radioembolization in patients with HCC and portal vein invasion: external validation of the Milan prognostic score P. Scalise, I. Bargellini To evaluate the role of diffusion-weighted imaging (DWI) in assessing treatment response in patients with liver metastases of primary neuroendocrine tumors (NETs) following selective internal radiotherapy (SIRT) with 90Yttriummicrospheres. Material and methods: 43 patients with liver metastases of primary NET who underwent abdominal MRI with DWI 40 ± 27 days before and 77 ± 48 days after SIRT were included. Tumor size, intralesional minimal, maximal and mean apparent diffusion coefficient (ADCmin, ADCmax and ADCmean, respectively) were measured for maximal 3 target lesions per patient on baseline and postinterventional DWI. Tumor response to radioembolization was categorized according to Response Evaluation Criteria in Solid Tumors v1.1 (RECIST) on follow-up examination.
Results: A total of 120 metastases with a mean diameter of 3.04 ±1.59 cm was analyzed. 27 (22.5%) lesions were categorized as partial response (PR), 87 (72.5%) lesions as stable disease (SD) and 6 (5%) lesions as progressive disease (PD). ADC values (ADCmin, max, mean) increased significantly (p<0.005) after SIRT in the group of PR and SD whereas there was no significant change of ADC values in the group of PD. Between the group of PR and SD, there was a significant difference in percentage change of ADCmean (61% vs. 11%, respectively, p<0,05) pre-and post-interventional. Conclusion: ADC values, especially ADCmean changes on DWI, seem to represent a valuable marker for the evaluation of treatment response after radioembolization of hepatic metastases in patients with primary NET and may help in assessing further therapeutic strategies.

SS 5.6
Prediction of treatment response following transarterial chemoembolization in patients with HCC using dualtracer positron emission tomography W. Purpose: To use CT texture analysis (CTTA) to identify specific imaging biomarkers of hepatic metastases, able to predict patient's response to therapy and overall survival. Material and methods: We exploited the imaging dataset of HERACLES trial (NCT03225937): 23 patients with amplified HER2 mCRC were included in the study. All had received anti-HER2 treatment, and underwent CT examination every 8 weeks, until disease progression. CT scans were semi-automatically segmented to extract all liver metastases. CTTA was performed on each segmented area, computing for each lesion 34 quantitative parameters. Monoparametric and multi-parametric analyses were assessed to identify features correlated to therapy response. We also performed a correlative survival (OS) analysis, considering subjects with good survival those with OS>9 months.
Results: In 23 patients we found 124 metastases, 55 classified as responding and 69 as nonresponding. Nine parameters reached statistical significance in mono-parametric analysis (best AUC=0.67, p=0.001), while in multivariate regression, ten parameters were used in the model, achieving AUC equal to 0.82, sensitivity of 82% and specificity of 72%. For OS analysis, 12 patients were "good" and 11 "poor" survivors. In mono-parametric analysis "cluster prominence" and "sum entropy" predicted OS with AUC equal to 0.78 and 0.83, respectively. The regression model with two variables ("cluster prominence" and "dissimilarity") reached sensitivity of 83% and specificity of 82%. Conclusion: Our study demonstrated CTTA as a potential biomarker to predict response of hepatic metastases to target therapy, possibly saving patients predicted as non-responder from toxicity. Moreover, CTTA could give indications on patients OS.

SS 5.9
Portosystemic shunt surgery in the era of transjugular Purpose: To assess the impact of a dedicated "respiratory motion correction software" on contrast-enhanced cone beam CT angiography (CBCTa) during intra-arterial liver-directed therapy. Material and methods: From 2015 to 2017, two groups of patients undergoing intra-arterial liver-directed therapy with (breathing, n=30) or without (still, n=30) significant respiratory motion artifacts were retrospectively included. All CBCTa were processed with and without a dedicated motion correction software. For both reconstructions (with and without motion correction), four readers were asked to independently assess: 1) the overall image quality on a 5-point scale, and 2) the presence of per-procedural relevant information on tumor and vasculature (overall vessel geometry; visibility of extrahepatic vessels; target tumor conspicuity; visibility of tumor feeders) on a 3-stage scale (good/intermediate/poor). Results: In the breathing group, motion correction increased the average image quality from 2.0±0.9 to 2.9±1.0 (p<0.01). The visibility of vessel geometry, extrahepatic vessels, and tumor feeders were significantly improved for all 4 readers, and that of tumor conspicuity was improved for 3 readers. In the still group, the average image quality was not significantly different between reconstructions (with and without motion correction) for all readers (4.0±0.6 vs 4.2±0.6; p=0.12). Visibility of vessel geometry, extrahepatic vessels, tumor feeders and tumor conspicuity were not altered for all 4 readers using the correction software. Conclusion: Using dedicated motion correction software increases both the image quality and the visualization of the per-procedural relevant information on tumor and vasculature needed during intra-arterial liver-directed procedures of motion-corrupted CBCTa, while maintaining that of still CBCTa unaltered. Material and methods: We obtained ethical approval by the institutional review board and informed consent for this prospective study. Between July 2015 and February 2018, high-risk patients with suspected liver lesions were consecutively enrolled and underwent gadoxetic acid-enhanced MRI. We constructed the following modified LI-RADS models: model 1 with "restricted diffusion" upgraded, model 2 with "hepatobiliary phase hypointensity" upgraded and model 3 with both features upgraded as major features. All images were reviewed by two independent radiologists blindly. The diagnostic accuracies were determined on their sensitivity and specificity and were compared with the McNemar test.
Conclusion: The modified LI-RADS models were no more accurate than the original v2018 LI-RADS, and the LR4/LR5/LR5-TIV combination showed potential in improving the diagnostic sensitivity with no substantial loss of specificity.

SS 6.6
Inter-reader, intra-reader agreement, and correlation with pathology of CT/MRI Liver Imaging To compare CT -texture analysis and CT features of liver metastases in pancreatic neuroendocrine tumours (PNETs) and in non-pancreatic neuroendocrine tumours (NPNETs) according to tumour grading. Material and methods: Contrast-enhanced CT images of liver metastases in 23 patients with PNETs and in 25 patients with NPNETs were analysed with 3D CT texture analysis (parameters evaluated: mean attenuation, standard deviation, skewness, kurtosis, entropy, mean of positive pixels and Tx_sigma) in arterial and portal phase; delta enhancement of the lesions was also calculated. The CT exams were performed before the beginning of any medical treatment. All patients presented a well-differentiated tumour according to WHO classification (G1 and G2). Data were analysed with Mann-Whitney and Chi-squared tests.
Results: Among CT texture analysis, in a comprehensive comparison between PNETs and NPNETs, the parameter "Skewness" was significantly higher in NPNETs (p value<0.05). These data were confirmed in subgroup comparisons evaluated in portal phase (NPNETs G1 vs PNETs G1, NPNETs G2 vs PNETs G2, and NPNETs G1+G2 vs PNETs G1+ G2). The parameter "Mean" was significantly higher in PNETs in comparison to NPNETs (p value=0,0066). Among CT features the "delta enhancement" was significantly higher (p value<0.05) in PNETs in two comparisons: PNETs G1 Vs NPNETs G1 and PNET Vs NPNETs. Conclusion: These results demonstrate significant differences between CT texture parameters of liver metastases in PNETs and NPNETs, and highlight the diversity of the two groups of NETs. These findings, in future, could be used as an "imaging biomarker" to predict therapy response. The purpose of this study is to investigate the benefit of synergistic analysis of positron emission tomography (PET) and MRI data using deep learning for automatic detection and segmentation of HCC. Material and methods: The micro-PET/MRI data were retrospectively collected from an animal study with orthotropic HCC tumor model conducted in our institution. Totally thirty-eight sets of coronal F18-FDG and corresponding T2WI images were selected for preliminary test. The labeled images were generated by drawing the ROI of tumors on T2WI images. Afterward, 28 and 10 sets of images were, respectively, used for network training and validation. The SegNet network architecture was selected for this implementation with a multi-channel data input and was pre-trained with BRATS data (brain data with 600,000 steps). For micro-PET/MRI data, the training steps were 10,000 with 6 images for each step. Three types of inputs (both F18-FDG and T2WI, F18-FDG only, and T2WI only) were tested for the efficacy of tumor detection using SegNet.
Results: The combination of PET and MRI information (F18-FDG and T2WI) provided a best mean dice coefficient (0.69), compared to either using PET data (0.47) or MRI data (0.58) for tumor detection and segmentation.
Conclusion: This preliminary study shows that the synergistic analysis of PET and MRI data using deep learning is feasible for automatic detection and segmentation of HCC, and provides better performance than using either individual PET or MRI data. Additionally, the transfer learning can ensure a proper training of the network, even though with limited amount of training data.

SS 7.9
Machine learning-based automated image registration improves reader confidence and lesion colocalization in cross-sectional liver studies Purpose: Variability in liver morphology, patient positioning and motion can impact lesion colocalization across different series or studies. We applied a fully automated registration algorithm to liver imaging studies to determine its impact on reader confidence and lesion colocalization. Material and methods: This is a retrospective, cross-sectional study. From surveillance gadoxetate-enhanced liver MRIs, we randomly selected 100 withinpatient inter-exam pairs (hepatobiliary phase series; baseline and follow-up), and independently applied manual image registration performed by expert readers and a fully automated algorithm, comprising both a machine learning-based liver segmentation and a 3D affine transformation network. Reader confidence on image feature similarities was analyzed using summary statistics. Colocalization was assessed through the distance of lesion centers on overlapping baseline and follow-up images using percentiles and paired t tests.
Results: Reader confidence improved across 49 pairs for reader 1 and 32 pairs for reader 2 after applying the automated registration. 84 and 73 pairs were classified with the highest confidence score when registered using the automated algorithm, compared to 40 and 46 using manual registration. Lesion colocalization significantly improved in comparison to manual registrations. Compared to reader 1 manual registration, mean lesion distance reduced by 6.47 mm (95% CI 3.87, 9.06; p<0.001), whereas for reader 2 mean reduction was 7.60 mm (95% CI 4.86, 10.34; p<0.001). Readers were not significantly different from each other across performance metrics. Conclusion: Automated liver focused image registration improves reader confidence and lesion colocalization in comparison to manual registration, potentially improving clinical care when visual comparison is the standard as in surveillance or treatment response assessment.

SS 7.10
Withdrawn by the authors  Purpose: To identify useful morphological and texture analysis CT features for the differential diagnosis between pancreatic ductal adenocarcinoma (PDAC) and focal autoimmune pancreatitis (AIP). Material and methods: Were reviewed the MDCTs performed to characterize solid pancreatic masses in 60 patients (30 focal AIP and 30 PDAC). Were evaluated lesion size and margins, main pancreatic duct (MPD) intralesional stenosis and upstream dilatation, upstream chronic pancreatitis and biliary dilatation. Attenuation was measured in the lesion and the unaffected parenchyma in the baseline, late-arterial, venous and delayed phases. Texture-analysis was performed using LIFEX software. Statistical analysis (t test, univariate and multivariate regression) was performed using SPSS. Results: Significant differences were observed between AIP and PDAC (p<0,05) in lesion attenuation in the unenhanced, late-arterial and venous phase, in normal pancreas unenhanced attenuation, in biliary dilatation and in texture entropy in the delayed phase. In univariate logistic regression analysis, lesion attenuation in the unenhanced, late-arterial and venous phases was significant (p<0.05). "AIP risk features", namely upstream MPD dilatation, biliary dilatation and chronic pancreatitis were significant "PDAC risk features". In multivariate regression analysis, upstream chronic pancreatitis was a statistically significant predictor of PDAC. The whole statistic model showed a significant predictive value for the diagnosis of PDAC, with a COX-SELL R2=0.422; receiver operating characteristic (ROC) curve confirmed its accuracy (AUC=0.918, p=0.037).

Conclusion:
We have identified several features that can improve the differential diagnosis between AIP and PDAC: higher lesion attenuation in the baseline, arterial and venous phases increase the probability of AIP, while upstream MPD dilatation, biliary dilatation and chronic pancreatitis increase the probability of PDAC. Results: A significant correlation between the TRG defined on DWMRI after CRT and pathological TRG was found (Spearman's rank correlation test: p value <0.001). The mean post-CRT ADC and ∆ADC in responder patients were significantly higher compared to non-responder ones (Student's T test: p value <0.001). By qualitative analysis, responders were correctly identified in 85% of cases. CR was identified in 56% of cases; integrating quantitative analysis using a cut-off value of ADC post-CRT of 1.23 x 10 -3 mm 2 /s sensitivity and specificity were 65% and 80%, respectively. Conclusion: Through both qualitative and quantitative analyses of DWMRI, MRI enables LARC response assessment after CRT, also resulting in a valid tool in CR prediction.

SS 9.2
Polyp detection rate as a quality measure in CT colonography: analysis of the performance of a CT colonography service using a reduced bowel preparation without dietary restriction S. Conclusion: New-generation CT, using high-resolution MPR images, represents a diagnostic tool in the assessment of loco-regional and whole-body staging of advanced rectal cancer, especially in patients with MRI contraindications.

SS 9.9
Analysis of anatomic variants of superior mesenteric artery and vein using MDCT S. Pashapoor 1 , K. Atasoy 2 ; 1 Bursa/TR, 2 Ankara/TR Purpose: To give a preoperative description of the arterial and venous anatomy of the colon on MDCT. Material and methods: Colic branches of the superior mesenteric artery (SMA) and vein (SMV) were studied in 100 patients who underwent abdominal MDCT for various clinical reasons. To reduce radiation burden, a single CT scanning was employed in each patient which allowed opacification of both the arterial and venous structures owing to a peculiar technique where the contrast medium was injected in two separate boluses.
Results: The incidence of colic arteries arising from the SMA was ileocolic artery, 100%; right colic artery, 25%; middle colic artery, 97%; and accessory middle colic artery, 20%. All patients had a single ileocolic vein, which drained into the SMV in 95 cases, into the gastroomento-pancreaticoduodeno-colic trunk (GPCT) in 3 cases, into ileal trunk in 1 case and into accessory MCV in 1 case. The GPCT was detected in 57 cases with several forms of the origin of the respective branches: the gastroomento-pancreaticoduodenal trunk (GPT) was detected in 34 cases, and the classic GCT was in 8 cases. IMV joined to the splenic vein in 50% of patients. Conclusion: Although the vascular anatomy of the colon is variable and complex, preoperative 3D-CT is informative and very helpful for surgeons in colonic resections. Both the arteries and veins of the colon can be opacified with a single CT scanning via injecting contrast medium in two separate boluses, without increasing patient radiation. 30.9±27.6 months) underwent hydrostatic reduction. Patients were grouped as those in whom the reduction was successful (success group) or failed (failure group). Percentages were given with 95% confidence intervals (CIs). Regarding the patient age, the groups were compared using Student's t test. Regarding the patient sex and the presence of an organic cause, the groups were compared using Fisher's exact test. p values smaller than 0,05 were considered statistically significant. Results: Regarding the patient age, there was no statistically significant difference between the success (n=51; 35.8±8.6 months) and failure groups (n=13; 29.7±3.8 months; p=0.521). Regarding the gender, although there was no statistically significant difference, the female ratio was higher in the failure group (success: 36 male,15 female; failure: 6 male, 7 female; p=0.114). In six patients, there was an organic cause (Meckel diverticulum (n=3), appendicitis (n=2), Burkitt lymphoma (n=1)). The reduction failed in all six patients with an organic cause (6/6; 100% [95%CI=60.9%-100%]) and in seven of 58 patients without (7/58; 12,1% [95%CI=5.9%-22.8%]; p<0.00001).

Conclusion:
In hydrostatic reduction of intussusception in pediatric patients, presence of an organic cause is an important factor that negatively affects the outcome. Thus, a detailed pre-reduction ultrasound examination is needed to rule out it. Although it was statistically not significant in this study, female sex may contribute to the failure rate. Material and methods: We retrospectively reviewed our institution's medical records of 338 patients diagnosed with liver metastases of GEP-NETs, from 2000 to 2018 identifying 43 patients with pancreatic (n = 22) and ileal (n =21) origin who underwent at least two gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA) enhanced MR examinations that included 20-min delayed hepatobiliary phase imaging. Two radiologists independently evaluated two sets of MRI exam of each of the patient-basic and follow-up exam using RECIST 1.1 categorization and classified them into one of the three categories: response, stable and progressive disease. For the follow-up exam, one radiologist reviewed the complete exam, the second-one disposed only of Gd-EOB-DTPA-enhanced 20-min hepatobiliary phase images. We compared RECIST evaluation of the two readers.
Results: There was a good agreement between the two reviewers (97%) with only 1 discordance in RECIST categorization (stable vs. progressive disease).
Of the 43 patients included in the study, 5 were excluded because of the incomplete examination. Of the 38 evaluated patients, 15/16 were categorized as progressive, 22/21 as stable disease and 1 as a partial response. Conclusion: In conclusion, hepatobiliary phase images obtained after gadoxetic acid-enhanced dynamic MRI have a good diagnostic accuracy in assessment of GEP-NETs' hepatic metastases. Gd-EOB-DTPA hepatobiliary phase has a potential value as a single-phase short MRI protocol for the mid-term follow-up in this group of patients.

SS 10.2
The value of diffusion-weighted imaging in dynamic monitoring to chemotherapy effectiveness of advanced gastric carcinoma J. Xu; Xi'an/CN Purpose: To evaluate the value of diffusion-weighted imaging (DWI) and apparent diffusion coefficient (ADC) value in monitoring the chemotherapy effectiveness of advanced gastric carcinoma dynamically. Material and methods: 42 advanced gastric carcinoma patients who were histopathologically confirmed underwent T2WI and DWI examinations at prechemotherapy, post-chemotherapy 3 d, 7 d, 30 d and 60 d, respectively. The longest diameters of tumor pre-chemotherapy and post-chemotherapy 60 d were measured on axial T2WI. Meanwhile, the ADC values at different time points were calculated. The mean ADC value among pre-and post-chemotherapy of each group (PR and SD) was compared.
Results: The ADC value of PR group increased gradually. The mean ADC value before therapy was statistically lower than those at different time points postchemotherapy (P < 0.05). The ADC value of SD group increased gradually from pre-chemotherapy to post-chemotherapy 30 d, and then the ADC value decreased at post-chemotherapy 60 d. The differences in the mean ADC values at different time points were statistically significant (P < 0.05).
Conclusion: DWI and ADC value can dynamically, quantitatively and early detect and monitor the chemotherapy response of advanced gastric carcinoma. This study examined the size, countable number and extent of nodal and distant metastatic sites on CT images obtained at initial diagnosis in stage IV GC patients and analyzed if and how they can predict overall survival. Material and methods: This retrospective study included 55 subjects diagnosed with stage IV GC. In addition to the retrieval of demographical, clinical and up-to-date survival data, baseline CT images obtained at the time of initial admission were reviewed for the extent of metastatic disease and prognostic factors on survival were analyzed by Cox proportional hazard models.
Results: None of the patient or tumor characteristics including age, gender, histological type, differentiation, location, and C-erbB status was found to significantly influence overall survival (p>0.05 for all). Multivariate analysis identified the number of perigastric and para-aortocaval metastatic lymph nodes as significant independent predictors of poor overall survival. Presence of metastasis in 7 to 14 perigastric lymph nodes (OR 4.0; 95% CI: 1.5-11.1; p=0.007) and presence of more than two metastatic para-aortocaval lymph nodes (OR 2.4; 95% CI: 1.1-5.5; p=0.034) were associated with significantly higher mortality risk.
Conclusion: The number of enlarged perigastric and paraaortic lymph nodes at initial CT examination seems to predict overall survival in patients with metastatic GC. However, large-scale studies are needed to generalize such a conclusion, which may provide the basis for alternative therapies other than palliation for those with a better outcome expectancy.

SS 10.4
Risk assessment for pancreatic fistula after pancreaticoduodenectomy with preoperative CT R. Menghini, G.A. Zamboni, A. Cybulski, R. Valletta, G. Mansueto; Verona/IT Purpose: To evaluate the predictive value of preoperative CT features for the risk of postoperative pancreatic fistula. Material and methods: We included 74 patients who underwent preoperative MDCT and pancreaticoduodenectomy in our institution. The patients were divided into 2 groups according to clinical data: 37 patients with postoperative pancreatic fistula (POPF) (group A) and 37 patients without POPF (group B). One reader reviewed the CTs and measured at the planned resection plane the main pancreatic duct (MPD) diameter, the density of the parenchyma in the different enhancement phases and the parenchymal thickness. The difference in attenuation between the venous and arterial phase was calculated as a surrogate for parenchymal fibrosis. Body composition was analysed by calculating visceral adipose tissue area (VAT), subcutaneous adipose tissue area (SAT), and skeletal muscle area at the L2 level using ImageJ software. Retrorenal fat thickness and psoas density were also measured. Fisher's exact test was used for categorical variables and Student's t test for continuous variables. Results: MPD diameter was 2.8±2.9 mm in group A and 6.5±3.2 mm in group B (P<0,0001). The mean attenuation difference between venous and arterial phase was -15.30 hounsfield units (HU) in group B and 0.70 HU in group A (P=0.0015). SAT was 17497.5 mm 2 in group A and 13292 mm 2 in group B (P=0.046). No significant difference was observed for the other parameters. Conclusion: Fibrosis, expressed by increasing enhancement of the normal pancreatic parenchyma at the planned resection plane, MPD diameter and increased SAT may express increased risk for pancreatic fistula after pancreaticoduodenectomy.

Conclusion:
The commonest s-MRCP finding in patients with BPH was mild/ moderate chronic pancreatitis (65.7%); the other most frequent anomalies were SOD (31.7%) and acinar filling (9.5%). Based on our experience, s-MRCP is to be recommended in the diagnostic workup of patients with BPH.

SS 10.6
Evaluation of pancreatic fibrosis with gadolinium ethoxybenzyl diethylenetriaminepentaacetic acidenhanced MRI D. Yunaiyama, H. Yamaguchi, Y. Nagakawa, T.L. Harada, T. Nagao, K. Saito; Tokyo/JP Purpose: To evaluate radiological-pathological correlation in pancreatic fibrosis between gadolinium ethoxybenzyl diethylenetriaminepentaacetic acid (Gd-EOB-DTPA)-enhanced MRI and resected non-tumorous tissue. Material and methods: From January 1,2011 to April 30, 2018, those who underwent pancreatic surgery and Gd-EOB-DTPA-enhanced MRI in our hospital were analyzed. Poor study because of breath-hold discipline, discrepancy of TR/TE between each dynamic study phase, severe susceptibility artifacts and patients who suffered from pancreatitis at the examination were excluded. Patient's age, sex, with or without diabetes mellitus were analyzed. The value of apparent diffusion coefficient (ADC) and intensity of pre-dynamic study, portal phase, venous phase, hepatobiliary phase and each ratio with pre-dynamic study were analyzed by 2 diagnostic radiologists trained 8 years and 6 years, respectively. The degree of pancreatic fibrosis was categorized into 4 grades: none, mild, moderate and severe by a well-trained pathologist. Results: Totally, 142 patients who underwent pancreatic surgery were analyzed in this study. The median age was 69 years (range: 40-88). Ninety-six patients were male and 46 were female, respectively. The degree of pancreatic fibrosis was associated with the intensity of pre-contrast (p=0.037), portal pre-contrast ratio (p=0.001) and venous pre-contrast ratio (p=0.001) with Kruskal-Wallis test; however, was not associated with the value of ADC (p=0.09). The degree of pancreatic fibrosis was prone to be associated with the necessity of percutaneous drainage; however, there was no statistical significance. Conclusion: Dynamic MRI using Gd-EOB-DTPA can predict the degree of pancreatic fibrosis with the intensity of pre-contrast, portal pre-contrast ratio and venous pre-contrast ratio. Results: Statistically significant difference (p<0.0001) was between nodules >5mm with enhancement and signal restriction in DWI and tumor dysplasia, between size and number of nodules detected by MRI and the grade of dysplasia and between dilatation of the Wirsung duct and grade of dysplasia. ADC map showed statistically significant results (p<0.0001) in the correlation between the entropy parameter and the grade of dysplasia. No statistically significant difference was between nodules <5mm detected by MRI and the pathological analysis and between nodular dimensions and tumor dysplasia.
Conclusion: MRI with dynamic study and DWI sequences is an accurate method in the identification of solid nodule with cut-off of 5 mm. DWI sequences and dynamic study with contrast media can be useful in the identification of lesional malignancy. Entropy could be used as a predictive parameter of lesional malignancy.

SS 10.9
Preoperative imaging evaluation after downstaging of pancreatic ductal adenocarcinoma: a multi-center study A. Beleù, G. Rizzo, N. Bellini, A. Grecchi, I. Testa, G. Giannotti, N. Cardobi, R. De Robertis, M. D'Onofrio; Verona/IT Purpose: Evaluation of pancreatic ductal adenocarcinoma (PDAC) after chemoradiotherapy downstaging is challenging due to CT overestimation of tumor extension and residual vascular involvement. With this study, we wanted to assess which radiological findings are most reliable at pre-operative imaging to achieve complete resection. Material and methods: We retrospectively enrolled 71 patients with locally advanced and borderline-resectable PDAC who underwent neoadjuvant chemoradiotherapy. Pre-operative CT or MR has been evaluated by three radiologists to assess major qualitative and quantitative parameters of lesions. Accuracy, sensibility and specificity compared to anatomopathological results were evaluated. Cohen's K coefficient has been calculated to evaluate inter-observer agreement (IOA). Different dimensional cut-offs were tested and compared to anatomopathological diameter, tumor persistence and margin infiltration.
Results: 25 mm cut-off was 67% sensitive, 90% specific and 77% accurate in assessing real tumor dimension. Sensitivity and specificity for 25 mm cut-off were, respectively, 23% and 94% for margin infiltration, and 51% and 67% for tumor persistence. The imaging presence of perivascular cuff reported a low accuracy in determining tumor persistence and margin infiltrations. Lesion enhancement and pattern homogeneity were not accurate in determining tumor persistence. IOA was generally poor to fair, except for 25 mm cut-off classification where IOA was moderate. Diagnostic accuracy is superior in consensus lecture rather than single lecture. Conclusion: Imaging methods tend to underestimate PDAC resectability after neoadjuvant-CRT. IOA is poor to fair in evaluating most of the qualitative parameters of downstaged pancreatic adenocarcinoma. Surgery should be considered for downstaged borderline resectable PDACs, independently of perivascular cuff presence, especially for tumors smaller than 25 mm. Material and methods: We included in our IRB-approved retrospective study 56 consecutive patients with histologically proven pancreatic adenocarcinoma, pancreatic neuroendocrine tumors (pNET) or pancreatic cystic neoplasm who underwent pre-operative contrast-enhanced computed tomography (CECT) before pancreaticoduodenectomy. 28 patients developed a pancreatic fistula after surgery and 28 had no post-surgery complications. CT texture analysis was used to quantify the heterogeneity of the pancreatic parenchyma on pre-operative CECT of both groups and the results were compared with the Mann-Whitney U test. Results: Several texture features were significantly different between patients who developed CR-POPF and those who did not: kurtosis (p = 0.011 ), entropy log 2 (p = 0.019) and energy (p = 0.038) as for the first-order features, all the gray-level co-occurrence matrix (GLCM) features (homogeneity, energy, contrast, correlation, entropy and dissimilarity) (p<0.05), several of the gray-level run length matrix (GLRLM) (SRE, LRE, HGRE, SRHGE, LRHGE, RP) and of the gray-level zone length matrix (GLZLM) features (SZE, HGZE, SZHGE, LZLGE, GLNU, ZLNU, ZP) (p<0.05).
Conclusion: Texture analysis of the pancreatic parenchyma on preoperative CECT was able to predict the onset of a post-operative pancreatic fistula. If confirmed on larger series, texture analysis could provide an additional means of preoperative risk stratification and possibly modify the management of these patients.

Rochester, MN/US
Purpose: MR fingerprinting (MRF) provides quantitative mapping of T 1 and T 2 relaxation times of a tissue from a single acquisition. The purpose of this study was to evaluate the feasibility and utility of MRF-based relaxometry in the bowel for quantitative evaluation of disease activity in Crohn's disease. Material and methods: 52 patients (27:25 M:F) undergoing MR enterography exams at 1.5T were included in this IRB-approved study. 3 axial and 3 coronal 2D-MRF slices were acquired through the bowel using a single breath-hold MRF-fast imaging with steady-state precession sequence. Regions of interest (ROIs) were drawn in the wall of small bowel and colon to assess T 1 and T 2 relaxation times of unaffected, active (wall edema, stratified or layered enhancement, adjacent mesenteric edema) and/or chronic inflamed segments (fibrotic wall thickening ± fatty infiltration, fat wrapping), for each patient. Information from weighted clinical images and endoscopy was used to identify affected segments. Mann-Whitney U and Kruskal-Wallis test with Dunn-Bonferroni post hoc tests were used to assess differences in T 1 and T 2 values between unaffected, acutely, and chronically inflamed bowel.
Results: There were 17 segments with chronic disease and 20 with acute inflammation. T 1 relaxation times were significantly longer in unaffected segments (1428±327 ms, n=76) versus inflamed segments (active 1309±322 ms and chronic 12138±301 ms), p = 0.004. T 2 relaxation times were longer in segments with active inflammation (67±20 ms) versus chronic fibrotic segments (44±19 ms), p = 0.003. Conclusion: Initial application of MRF in bowel imaging is presented, with promising results for quantitative differentiation of unaffected, actively inflamed and chronically diseased bowel.

SS 11.3
Evaluation of a new CT enterography score for inflammatory activity in Crohn's disease: correlation with laboratory and endoscopic findings, and interreader agreement A. Agostini The proposed structured report for Crohn's disease aims standardisation, comparability and completeness in the quality and accessibility of the radiologic information, providing integration and combination of radiology data elements with other key clinical parameters in a structured database for further exploitation.

SS 11.6
MR enterography for the assessment of postoperative GI function: technique and patient tolerability S.J. Chapman 1 , J.A. Helliwell 1 , A. Menys 2 , D.G. Jayne 1 , D. Tolan 1 ; 1 Leeds/UK, 2 London/UK Purpose: Ileus is common after colorectal surgery, but its diagnosis is an unmet clinical challenge. Traditional measures of gut function (such as passage of flatus and stool) are limited by patient and assessor bias. We explored the feasibility of magnetic resonance enterography (MRE) performed three days after surgery for the assessment of GI function. Material and methods: A technical description of the MRE protocol and patient tolerability are described in a study population of patients undergoing laparoscopic colorectal resection and receiving MRE within an ongoing, double-blinded, randomised controlled trial. The MRE protocol was devised and iteratively refined prior to study initiation by a multi-disciplinary group of GI radiologists, radiographers, colorectal surgeons, and patient representatives.
Results: MRE was performed on the third postoperative day in 18 patients. An enteral challenge of 2 glasses of water was administered 30 minutes prior to the scan. A series of 60-second, coronal, balanced gradient-echo motility scans were performed (20-second breath-hold and 40-second free-breathing). Scan parameters included 2cm slice gap, 1cm slice thickness, and 1 imageper-second temporal resolution (total scan time 20-30 minutes). Global small bowel motility was quantified using dedicated software. Seventeen (n=17/18; 94.4%) patients completed the MRE protocol. One patient terminated the scan early because of abdominal discomfort. There were no adverse events attributed to the MRE protocol.
Conclusion: A comprehensive MRE protocol assessing postoperative GI function was feasible and well tolerated on postoperative day 3 after laparoscopic colorectal resection. Further evaluation is required to determine the clinical applicability of this form of assessment.

SS 11.7
Qualitative and quantitative analyses of virtual noncontrast images in CT enterography with a 3rdgeneration dual-source dual-energy CT A. Agostini Purpose: To quantitatively compare the gastric motility between obese and normal weight people with cine-MRI. Material and methods: In this non-randomized prospective single-center study, obese patients candidated for sleeve gastrectomy (OB) and normal weight volunteers (NW) were included. Cine-MRI (1.5T) was performed using 2D-TRUEfisp on a plane parallel to the long axis of the antrum. Images were acquired before (T0), immediately after the end of a liquid meal intake (T1) and every 20 minutes (T2-T5) for a total exam time of 100 minutes. Each sequence lasted 60 seconds, with 120 images acquired. Two radiologists evaluated images on a dedicated software. Antral diameters (AD) and antral contraction waves' width (ACWw) and amplitude (ACWa) were assessed on each image of the entire sequence. The results obtained in OB and NW were compared, using a non-parametric test. The inter-reader and intra-reader agreement were evaluated. Results: Our final population consisted of 50 patients, 25 for each group.
Cine-MRI showed that, both during fasting and postprandial period, in OB all parameters (AD/ACWs width/amplitude) were significantly lower (p val-ue=0.002/0.003/0.001, respectively). A good inter-reader and a very good intra-reader agreement were observed. Conclusion: The cine-MRI showed a significantly more frequent altered motility in OB compared to NW. The obesity-related cause of the impaired antral motility, whatever it is, could become a new potential therapeutic target.

SS 11.10
Is there a relationship between the number of middle colic artery and transverse colon length? A study of radiological anatomy with 3D CT S. Pashapoor; Bursa/TR Purpose: Preoperative knowledge of the normal pattern and variations of the mesenteric arteries is substantially important in the successful surgery. In this study, we identified the relationship between middle colic artery (MCA) variations and the transverse colon length using 3D multidetector CT.

Clichy/FR
Purpose: To evaluate if iso-or hyperintensity of hepatocellular adenomas (HCAs) on hepatobiliary phase (HBP) is systematically related to high uptake of hepatospecific contrast agent, using a quantitative approach. Material and methods: This bi-centric retrospective study included all patients with histologically confirmed and subtyped HCA from 2009 to 2017 who underwent MRI with HBP after gadobenate disodium (Gd-BOPTA) injection and who showed iso-or hyperintensity on HBP. The signal intensity of tumors on pre-and post-contrast images, and the presence of hepatic steatosis were noted. Contrast uptake on HBP was quantified using the liver-to-lesion-contrast-enhancement ratio (LLCER). Results: 24 HCA showed iso-or hyperintensity on HBP, specifically 17 inflammatory (IHCA) and 7 ß-catenin HCAs (BHCA). Eighteen HCAs (75%) [17 IHCA and 1 BHCA] had a LLCER<0% (median -13.6% %, group 1), of which 94% were hyperintense on pre-contrast T1-w images, with background hepatic steatosis. Six HCAs (25%) had LLCER≥0% (median 2.9%, group 2), all were BHCA. A LLCER ≥1.6% was associated with the diagnosis of BHCA with a sensitivity of 86% and a specificity of 100%. Conclusion: Iso-or hyperintensity of HCAs on HBP does not necessarily correspond to a higher hepatospecific contrast agent uptake when compared to that of the surrounding liver. In inflammatory HCA, spontaneous tumor hyperintensity on pre-contrast images and underlying steatosis likely explain the iso-or hyperintensity of these tumors, which do show reduced HBP contrast agent uptake. On the other hand, marked contrast uptake can be observed, especially in BHCA. This emphasizes the importance of quantitative analysis of contrast uptake for tumor subtyping.

SS 12.6
Hepatic localization of extramedullary hematopoiesis in beta-thalassemia patients: diagnostic accuracy of T2*, apparent diffusion coefficient and gadolinium-enhanced dynamic MRI P.P. Arcuri 1 , A.

SS 12.10
Inter-observer reproducibility of liver stiffness measurement using MR elastography V. Phou, P. To assess the inter-observer reproducibility of liver stiffness measurement using magnetic resonance elastography (MRE) in a large panel of observers, including technicians. Material and methods: Fifty-five consecutive patients with liver MRE acquisition and pathological analysis of the liver parenchyma were included in this retrospective study (F0-2: N=26/F3-F4: N=29). MRE was acquired on a 1.5T Siemens system, using a gradient-recalled echo (GRE) MRE sequence. Eight observers (5 physicians and 3 technicians) were first asked to draw in the liver 2 elliptical regions of interest (ROI) (250-500mm²) and then one free-hand ROI, as large as possible based on the confidence map and the anatomy. Inter-observer reproducibility was assessed using the intra-class correlation coefficient (ICC) and Bland-Altman analysis. To investigate the sensitivity of morphologic criteria for the detection of cirrhosis in alcoholic liver disease (ALD), hepatitis C (HCV), and non-alcoholic steatohepatitis (NASH).
Material and methods: This is a retrospective study of 100 patients (53M, 48F) with different etiologies of chronic liver disease (CLD) including NASH (n=41), HCV (n=39), and ALD (n=20) with different degrees of fibrosis on liver biopsy. Two readers (R1:3 years' experience in advanced liver imaging; R2: junior radiology resident) independently analyzed the CT exams performed within 6 months of biopsy for the presence of morphologic changes of cirrhosis (surface nodularity, fissural/periportal widening, increased caudate-right lobe ratio) and portal hypertension (splenomegaly, varices, ascites). Each reader assigned an overall score for cirrhosis and portal hypertension. Sensitivity/specificity of morphologic criteria in each etiology group was calculated for each reader and compared using Chi-squared test. Frequencies of different morphological changes in each group were compared. Results: Using morphologic criteria, sensitivity for the diagnosis of cirrhosis was significantly lower in NASH (R1: 81%; R2: 63%) compared to ALD (R1: 95%; R2: 90%) and HCV (R1: 85%; R2: 77%) for both readers (p<0.001). The reader with advanced training outperformed the other reader in NASH (p<0.001). Surface nodularity was less common in NASH cirrhosis (p<0.001). Sensitivity for the diagnosis of NASH cirrhosis significantly decreased when only patients without portal hypertension were included (58%, p=0.003). Conclusion: Use of morphologic criteria for the diagnosis of cirrhosis had significantly lower performance in patients with NASH. Liver morphology changes in NASH cirrhosis are more subtle and degree of fibrosis could be underestimated when traditional morphologic criteria are used. Purpose: This study hypothesized that the liver vein to cava attenuation (LVCA) on portal venous abdominal CT scans is a helpful add-on to liver volumetry and the caudate-right lobe ratio (crl-r) to detect clinically significant liver fibrosis. Material and methods: Fifty consecutive patients with portal venous phase abdominal CT scans and gradient-echo-based MR elastography within 3 months without portal vein thrombosis or prior liver surgery were included. One patient was excluded because of insufficient MR elastography quality. Thirty-six patients had a liver stiffness ≤3.5kPa, while twelve patients had a stiffness >3.5kPa, consistent with clinically significant liver fibrosis (corresponding to a fibrosis stage ≥f2). Liver segmental volume ratio (LSVR), defined as Couinaud segments I-III to segments IV to VIII, as well as LVCA (1-3: liver vein attenuation higher, equal and lower than vena cava, 4: liver veins not contrasted) were calculated. LSVR-A was defined as LSVR*LVCA, while LIMA-FS (liver imaging morphology and attenuation-based fibrosis score) was defined as crl-r*LVCA.
Conclusion: LVCA is a helpful add-on to volumetry on portal venous abdominal CT scans. LIMA-FS, a combination of LVCA and crl-r, allows a just slightly inferior prediction of clinically significant liver fibrosis than volumetry without time-consuming image post-processing. Material and methods: A total of 149 patients with esophageal varices (EV) caused by liver cirrhosis were enrolled in this study in the last 2 years of our hospital, all patients underwent MSCTP and endoscopy within 4 weeks, the patients were divided into 68 hemorrhage group and 81 non-hemorrhage group according to whether they were bleeding evaluated by clinical manifestation and endoscopy. Test MSCTP left gastric vein (LGV) diameter, esophageal varices (EV) diameter and the vascular area, calculating sensitivity, specificity, ROC and AUC. All patients were followed up for 6 months, 30 patients were re-admitted due to EVB, and the predictive value of new indicators was verified on their last MSCT appearance.
Results: The diameter of LGV, EV and the vascular area of the bleeding group were, respectively (7.81±6.72)mm, (6.78±2.90)mm, (1.72±1.30)cm 2 , and the non-bleeding group were (6.22±4.61)mm, (5.92±3.05)mm, (1.05±1.26)cm 2 , the critical points were 5.90mm, 5.15mm, 1.04cm 2 , AUCs were 0.75, 0.69, 0.80, respectively. In the last CT findings of 30 patients readmitted with EBV, logistic regression analysis showed that the diameter of LGV >5.90mm (OR=2.136, P=0.007), and the vascular area>1.04cm 2 (OR=2.338, P=0.013), which were independent risk factors for hemorrhage. Conclusion: MSCTP can be used as an alternative to predict EVB in cirrhotic patients, the LGV diameter and the vascular area on MSCTP can be used as new effective indicators to predict the occurrence of EVB in patients with liver cirrhosis.

SS 13.5
Acute alcoholic hepatitis, towards a radiological diagnosis? F. Grillet, P. Calame, J.P. Cervoni, E. Delabrousse; Besançon/FR Purpose: To study the radiological signs of acute alcoholic hepatitis (AAH), and their regression. Material and methods: For this monocentric observational study, we retrospectively established a list of 104 consecutive patients with a histological diagnosis of AAH, between January 2008 and June 2018. We included 57 patients who had undergone an injected cross-sectional imaging up to 30 days before the liver sampling. The radiological signs we analysed were based on the histological pattern of AAH: the presence of heterogeneous steatosis, hepatic perfusion disorders, and signs of liver dysmorphia. We evaluated the regression of radiological signs observed on the control cross-sectional imaging when performed. We created two control groups of patients with histologically proven cirrhosis (respectively, alcohol-induced without AAH and metabolic induced) and a prior cross-sectional imaging. We then paired patients between the AAH group and each control groups. Results: In the 57 included patients, 54 (94.7%) had a heterogeneous steatosis, 39 (78%) hepatic perfusion disorders and 48 (84.1%) signs of liver dysmorphia. When analysing the paired patients, the association of those 3 signs showed a 100% specificity and positive predictive value towards the AAH diagnosis. For the 30 patients who underwent a follow-up imaging, a regression of both steatosis and perfusion disorders was observed in 21 of them (70%) and the liver's volume decreased by a mean 38%. Conclusion: In our study, the association of heterogeneous steatosis, hepatic perfusion disorders and signs of liver dysmorphia was a specific radiological pattern for the diagnosis of acute alcoholic hepatitis.

SS 13.6
Predicting different stages of liver fibrosis with 2D shear-wave elastography: histopathological correlation study M. Aksakal, S. Özhan Oktar, H. Şendur, G. Esendağlı Yılmaz, S. Özenirler, M. Cindoruk, K. Hızel, F.N. Baran Aksakal, C. Yücel; Ankara/TR Purpose: In this study, we investigated the accuracy of 2D shear-wave elastography (2D SWE) in detecting liver fibrosis using histopathological analysis as the reference method. Material and methods: Our single-center prospective study included 80 consecutive adult patients who had liver biopsy within 14 months of elastographic examination. A real-time SD SWE evaluation was performed using LOGIQ E9 system (GE Medical Systems, Wisconsin, USA). The median values of 10 valid liver stiffness measurements in kPa for each patient were obtained and were compared with the METAVIR scores obtained from the liver biopsy, using Kendall's rank correlation test. The diagnostic performance of real-time 2D SWE was assessed and cut-off values were set by ROC curve analysis.
Results: A significant correlation was found between liver stiffness kPa values and degree of liver fibrosis (Kendall's tau=0.56, p=0.0001). Our cut-off values for different stages of fibrosis were (F1-4); 5.92 kPa (sensitivity 75%, specificity 70%) and 6.68 kPa (sensitivity 95%, specificity 85%), 9.20 kPa (sensitivity 90%, specificity 96%) and 12.8 kPa (sensitivity 80%, specificity 96%), respectively. Conclusion: Our findings suggest that 2D-SWE can be used for the assessment of liver fibrosis. It should be noted the best sensitivity value (%95) obtained in our study for identifying patients with no or minimum/mild fibrosis and those with severe fibrosis and cirrhosis (F0-1 versus F2-3-4). This is an important finding since patients with severe fibrosis need to be closely supervised. Studies on larger groups of biopsied patients are required to establish the most appropriate cut-off values for each particular stage of liver fibrosis.

SS 13.7
Withdrawn by the authors SS 13.8 Revisiting the role of US surveillance for HCC in chronic liver disease J.H. Kim, S.J. Ahn, J.K. Han; Seoul/KR Purpose: We investigated the role of US surveillance for HCC in chronic liver disease and analyzed the factors which affect US detection of HCC Material and methods: Among 16986 patients with chronic liver disease who underwent surveillance US, 1901 patients who underwent both surveillance US and CT within 6 months were included in this study. We retrospective evaluated patient characteristics, US findings, and laboratory assessment. We also assessed the factor which affects the development of HCC within 2 years. Purpose: In multi-institutional clinical trials, liver MR elastography (MRE) is considered analyzable if there are > 2,000 acceptable pixels across the four acquired slices, marginally analyzable if there are < 2,000 pixels but > 500-700 pixels, and not analyzable if there are < 500-700 pixels. However, little objective data supports these cutoffs. We performed a simulation study to analyze these cutoffs using data from a multi-center drug development clinical trial of adults with nonalcoholic steatohepatitis (NCT02854605). Material and methods: Twenty-five cases were randomly selected from the aforementioned clinical trial. Liver stiffness values for every pixel in every subject were recorded. Average liver stiffness values were iteratively recalculated by randomly removing ten pixels at a time, and repeating each iteration 100 times. The range of liver stiffness values, across all 100 simulations, at 500, 700, and 2,000 pixels was captured, and means and standard deviations (SDs) of these ranges were calculated.
Results: Using all pixels, the mean(SD; range) liver stiffness value and ROI area were 3.44(1.54; 1.99-8.55) kPa and 4,272(2,276; 1,115-11,163) pixels, respectively. The mean(SD) of the range of simulated liver stiffness values for cutoffs of 500, 700, and 2,000 pixels were 0.19(0.14), 0.16(0.11), and 0.07(0.07) kPa, respectively. Conclusion: The ranges of simulated liver stiffnesses for all three pixel cutoffs are small, and for most contexts of use are likely acceptable for drug development clinical trials, and probably also for clinical care. Hence, at a proof-ofconcept level, subject to validation in larger cohorts, these MRE liver stiffness analyzability cutoffs are reasonable. To determine if magnetic resonance elastography (MRE) data of the spleen are usable in case of liver MRE acquisition and to assess the determinants of non-usable spleen MRE data. Material and methods: Eighty-six consecutive MRE examinations (gradientrecalled-echo), acquired on a 1.5T system, were retrospectively evaluated. Clinical data were recorded: chronic liver disease, gender, age, weight, BMI. Imaging data were recorded by one observer: degree of ascites, liver and spleen T2*, spleen largest diameter, body diameters (transverse and anteroposterior), stomach status (collapsed/half-filled/distended), liver MRE failure, liver MRE quality score based on the coverage on the confidence map. Spleen MRE non-usable data were defined as no pixel with confidence index higher than 95% and/or no apparent shear waves imaged and were assessed by 2 observers and a 3 rd one in case of disagreement. Logistic regression analysis was performed to assess the link between spleen MRE success and potential predictive factors of failure. Results: Spleen data were not usable in 20 cases (33%). On univariate analysis, higher BMI, larger transverse and anteroposterior body diameters, lower liver and spleen T2*, and lower liver MRE quality score were all significantly associated with non-usable spleen MRE data (P<0.016); while on multivariable analysis, only higher BMI, lower spleen T2* and lower liver MRE quality score were significantly associated with non-usable spleen MRE data (P<0.023).
Conclusion: Spleen MRE data on a liver acquisition are usable in 2/3 of the cases. Successful liver MRE acquisition, lower BMI and lower spleen iron content (higher T2*) are predictive factors of usable spleen MRE data.