ESGAR 2021 Book of Abstracts

REVIEWING PANEL 6 O. Akhan, Ankara/TR C. Aubé, Angers/FR M.A. Bali, Brussels/BE I. Bargellini, Pisa/IT T.V. Bartolotta, Palermo/IT A. Ba-Ssalamah, Vienna/AT N. Bastati-Huber, Vienna/AT R.G.H. Beets-Tan, Amsterdam/NL E. Biscaldi, Genoa/IT I. Boulay Coletta, Paris/FR G. Brancatelli, Palermo/IT R. Cannella, Palermo/IT K. Cortis, Msida/MT N. Courcoutsakis, Alexandroupolis/GR L. Crocetti, Pisa/IT L. Curvo-Semedo, Coimbra/PT G. D’Assignies, La Roche-sur-Yon/FR M. Dioguardi Burgio, Clichy/FR


TABLE OF CONTENTS
Scientific Sessions, Wednesday, June 16 (SSL 1.1 -SSL 1.6) 8-9 Scientific Sessions, Thursday, June 17 (SSL 2.1 -SSL 2.7) 10-11 Scientific Sessions, Friday, June 18 (SSL 3.1 -SSL 3.7) 12-13 Scientific Sessions, On Demand (SSD 1.1 -SSD 6.10) 14-29 Authors' Index 31-33 Purpose: To study the prognostic significance of MRI-identified extramural vascular invasion (EMVI), tumour deposits (TD), significant mesorectal nodes (LN) and pelvic sidewall disease (PSW) in locally advanced rectal cancer. Material and methods: This is an institutional review board-approved study of patients with non-metastatic locally advanced rectal adenocarcinoma treated with neo-adjuvant long-course chemoradiotherapy and surgery. Staging and restaging MRI was reviewed for EMVI, TD, LN and PSW by a radiologist blinded to the patient's outcome. Interobserver agreement (IOA) between four radiologists was studied in an anonymised subset. Outcome data were obtained from a prospectively maintained database. Prognostic significance of imaging findings was assessed using Kaplan-Meier analysis and Cox proportional hazard model. To identify any potential problem areas in the applicability of the TNM8 classification for the radiological staging of rectal cancer. Material and methods: A web-based survey including 52 cases/questions related to various staging TNM8 staging scenarios for (colo)rectal cancer was circulated among radiologists/radiology residents and clinical colleagues worldwide. Each case was based on a single MRI (or CT) image + schematic representation and a detailed description of the imaging findings. Responses were analyzed using descriptive statistics and correlated with nature of practice (academic/non-academic/cancer-referral center), profession (radiologists/nonradiologists), and experience level. Items with <80% consensus were classified as potential problem areas. Results: To date, the survey has been completed by 311 respondents (from 31 countries), of which 82% are radiologists/radiology residents with an experience level in rectal imaging of >10 years (32%), 5-10 years (32%) and <5 years (36%), respectively. Respondents originated from academic/non-academic/ cancer-referral centers in 48%/29%/23%, respectively. Main problem areas (<80% consensus) included the following: 1. Effect of sphincter invasion on T-stage (39-68% consensus). 2. Mesorectal fascia versus peritoneal invasion in tumours above the peritoneal reflection (52% consensus). 3.Differentiation between lymph nodes and tumour deposits (43% consensus). 4.Differentiation of regional (N) versus non-regional (M) nodes (57-68% consensus). Items 1 and 3 showed significant correlations with experience level and/or profession (p=0.01). Conclusion: With this survey, we have identified several potential problem areas that can serve as a basis for discussion to produce more practical guidelines on the radiological application of the TNM8 specified for rectal cancer ultimately aiming to avoid inconsistencies in radiological reporting in the future.

SSL 2.3
Before fifty: imaging findings in early age-onset colorectal cancer A. Perez Birmingham/UK, 4 Northwood/UK Purpose: Perfusion CT imaging relies on analysis software, linked to the scanner acquisition, to provide vascular measurements of the tissue of interest. In a multicentre setting, different scanners, acquisition/reconstruction parameters, software/mathematical analysis models and/or readers potentially contribute to measurement variability. We assessed the variability of perfusion CT measurements in the PROSPECT trial. Material and methods: Adult participants with suspected or proven primary colorectal cancer were recruited prospectively from 13 hospitals (ISRCTN 95037585). Exclusions were metastatic disease at staging, prior cancer(s), and contraindications to contrast agents. Perfusion CT at local sites was analysed by 25 radiologists using software based on deconvolution, distributed parameter, slope or Patlak analysis (GE, Siemens, Phillips or Toshiba). Imaging was reviewed centrally by 3 experienced radiologists. Bland-Altman methods determined limits of agreement between local and central review measurements. Potential sources of variation including scanner and region of interest were assessed graphically. Results: 291/303 (96%) participants imaged successfully were included in this analysis as technical issues precluded central review in 14/303 (4%). Mean difference [95% limits of agreement] for blood flow (mL/min/100mL), blood volume (mL/100mL), and permeability surface area product (mL/min/100mL) were -9.5 [-71.2, +52.3], -2.6 [-13.4, +8.19], and 3.1 [-27, +33.1], respectively. Plotted data indicated differences in the region of interest between local and central reviews reflecting observer variation, but this was not a major contributor to variance. Similarly, scanner type did not impact substantially the variance. Conclusion: Variation in derived measurements, particularly vascular permeability, highlights ongoing challenges for quantitative imaging in a multicentre setting.

SSL 2.6
Assessment of lymph node status and tumor response after chemoradiation therapy in locally advanced rectal cancer: comparison of three methods of region of interest for intravoxel incoherent motion parameters Y. Yuan, H. Pu, X.-L. Chen, H. Li; Chengdu/CN Purpose: To assess the diagnostic performance of region of interest (ROI) methods of intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) for determining lymph node metastases (LNM) and tumor response after chemoradiation therapy (CRT) in locally advanced rectal cancer (LARC). Material and methods: 79 patients underwent preoperative IVIM-DWI before and after CRT. IVIM-DWI parameters apparent diffusion coefficient (ADC), slow diffusion coefficient (D), fast diffusion coefficient (D*), perfusion-related diffusion fraction (f) and their percentage changes [Δ%] were obtained according to three ROI protocols: whole volume, single slice and small samples. Risk factors were evaluated through logistic regression analyses. Areas under the receiver operating characteristic curves (AUCs) were calculated to evaluate diagnostic performance. Disease-free survival was estimated using Kaplan-Meier survival curves. Results: Interobserver agreement was good for pre-and post-CRT wholevolume ROI and single-slice ROI (intraclass correlation coefficient [ICC], 0.775-0.953), and moderate for small samples ROI (ICC, 0.581-0.905). As for LNM, pCR and good response, AUCs for whole-volume ROI-derived Δ%D were higher than that of the other IVIM-DWI parameters (AUC,0.810,0.851,. In multivariate analysis, whole-volume ROI-derived Δ%D was an independent risk factor for discriminating LNM, pCR and good response (odds ratio, 0.947, 0.952, 0.805; p=0.001, <0.001, <0.001). After CRT, patients with LNM or poor response showed earlier recurrence (hazard ratio, 3.408, 3.498; 95% confidence interval, 1.289-9.012, 1.018-12.021; p=0.013, 0.047, respectively). Conclusion: Whole-volume ROI-derived Δ%D provided high diagnostic performance and was an independent factor for evaluating LNM, pCR and good response. Furthermore, patients with LNM or poor response may have worse outcomes regarding recurrence.

SSL 2.7
Vascular but not metabolic phenotype is associated with neoadjuvant therapy response in primary esophageal/ esophagogastric cancer S.J. Withey 1 , K. Owczarczyk 1 , M. Grzeda 1 , C. Yip 2 , R. Neji 3 , A. Green 1 , J. Bell 1 , H. Deere 1 , M. Green 1 , G. Cook 1 , V. Goh 1 ; 1 London/UK, 2 Singapore/SG, 3 Frimley/UK Purpose: The current standard-of-care for potentially curable esophageal/ esophagogastric cancer is neoadjuvant therapy prior to surgery. Imaging biomarkers that can predict response to neoadjuvant treatment may help to deliver individualized care. We aimed to investigate whether the tumor vascular or metabolic phenotype, assessed with dynamic contrast-enhanced MRI (DCE-MRI) and 18F-fluorodeoxyglucose positron emission tomography/CT (18F-FDG PET/ CT), provides predictive and/or prognostic information beyond current staging. Material and methods: Following ethical approval, participants with potentially resectable esophageal/esophagogastric cancer underwent DCE-MRI and 18F-FDG PET/CT prior to neoadjuvant therapy. Vascular (K trans , v e , k ep , peak enhancement integral (PEI)) and metabolic (SUV max , SUV mean , metabolic tumour volume, total lesion glycolysis) parameters were generated from MRI and PET, respectively, and compared with a pathological response (Mandard tumour regression grade (TRG)) and recurrence-free or overall survival using logistic regression modelling. Results: 39 participants (30 male; median 65 years, range: 54-72) underwent successful imaging. In multivariable analysis (adjusted for age, gender, T/N stage), lower PEI and higher K trans were potentially predictive of response to neoadjuvant therapy (TRG 1-2). For PEI, odds of response decreased by 5% for each 0.010 increase in PEI (OR = 0.95; 95% CI 0.90-1.00; p=0.03); for K trans , odds of response increased by 13% for 0.010 increase in K trans (OR = 1.13; 95% CI 1.00-1.28; p=0.05). PET parameters were not predictive of response. No relationships between any imaging parameters and recurrence-free or overall survival were identified. Three texture analysis (TA) parameters assessed the homogeneity of the luminal content, with ratios calculated between the TI and 1) the SB and 2) the ascending colon. Four TI motility metrics were derived. Ascending colon diameter was measured. Comparison between HCs and IBS-C patients was performed independently for 1) the three TA parameters, 2) four TI motility metrics, and 3) ascending colon diameter. Results: Compared to HCs, IBS-C patients had TI:colon ratios higher for TA contrast (P < 0.001), decreased TI motility [lower mean motility (P = 0.04), spatial motility variation (P = 0.03) and area of motile TI (P = 0.03)], and increased ascending colon diameter (P = 0.001). Conclusion: IBS-C patients show reduced TI motility and differences in luminal content compared to healthy controls. This potentially indicates reflux of colonic contents or delayed clearance of the TI which alongside increased ascending colon diameter may contribute to symptoms of constipation and bloating.

SSL 3.7
Intestinal malrotation in adults: prevalence and findings based on CTC A. Perez, P.J. Pickhardt; Madison, WI/US Purpose: Intestinal malrotation, or nonrotation, is largely a pediatric diagnosis, but initial detection can be made in adulthood. CTC provides an ideal means for adult diagnosis and estimating prevalence. The purpose of this study was to evaluate the prevalence and imaging findings of intestinal malrotation in asymptomatic adults at CTC screening, as well as incomplete optical colonoscopy (OC) referral.

Material and methods:
The CTC database of a single academic institution was searched for cases of intestinal malrotation. Prevalence was estimated from 9,844 adults undergoing initial CTC screening and 1,332 referred for incomplete OC. Demographic, clinical, and imaging data were reviewed. Results: 27 cases of malrotation were confirmed (mean age, 62±9 years; 15M/12F), including 17 from the CTC screening cohort (0.17% prevalence) and 10 from incomplete OC (0.75% prevalence; p<0.001). Most cases (56%; 15/27) were initially diagnosed at CTC. In 67% (12/18), the presence of malrotation was missed on at least one relevant abdominal imaging examination. At least 22% (6/27) had a history of unexplained, chronic intermittent abdominal pain. At CTC, the SMA-SMV relationship was normal in only 11% (3/27). The ileocecal valve was located in the RLQ in only 22% (6/27). Two patients (7%) had associated findings of heterotaxy (polysplenia).

Conclusion:
The prevalence of intestinal malrotation was over four times greater for patients referred from incomplete OC compared with primary screening CTC, likely related to anatomic challenges at endoscopy. Malrotation was frequently missed at other abdominal imaging examinations. CTC can uncover unexpected cases of malrotation in adults, which may be relevant in potential future complications.

SSD 1.1
Can preoperative CT scan and bioimpedance vector analysis help to predict the development of pancreatic fistula? C. Maino 1 , D. Ippolito 2 , M. Ragusi 2 , D. Gandola 2 , C. Talei Franzesi 2 , T.P. Giandola 2 , S. Sironi 1 ; 1 Milan/IT, 2 Monza/IT Purpose: To evaluate the accuracy of the bioimpedance vector analysis (BIVA) in predicting pancreatic steatosis (PS) and development of postoperative pancreatic fistula (POPF), compared with preoperative CT scan and the pathologic specimen. Material and methods: A total of 75 patients who underwent pancreatic resection having a preoperative CT staging for pancreatic cancer were prospectively enrolled. All CTs were analyzed to determine the overall mean attenuation value of the pancreas, excluding focal lesions, expressed as Hounsfield unit (HU). Moreover, a radiologist drew three different regions of interest (ROI), located in the head, body, and tail, to calculate the mean attenuation value at these levels. BIVA was performed the day before surgery and pancreatic steatosis was assessed with the fat mass index (FMI). Spearman correlation and ROC analysis were used to analyze and compare the techniques.

P.M.; Chennai/IN
Purpose: To assess the relationship between CT attenuation indexes and impaired glucose metabolism in a clinical setting. Material and methods: This retrospective study was done from October 2019 to March 2020; 80 patients (47 men and 33 women; age range 35-60 years) who underwent CT were included in the study. Attenuation was measured in three regions of interest in the pancreas and the spleen on nonenhanced CT images. The difference between pancreatic and splenic attenuation and the pancreasto-spleen attenuation ratio were calculated. A multivariate logistic regression model was used to determine whether CT attenuation indexes correlated with impaired glucose metabolism (i.e., impaired glucose tolerance, impaired fasting glucose, or presence of diabetes). Results: Patient with impaired glucose metabolism was significantly correlated with the difference between pancreatic and splenic attenuation (r = -0.621, P < .01) and the pancreas-to-spleen attenuation ratio (r = -0.611, P < .01). The CT attenuation indexes were significant and independent variables predictive of impaired glucose metabolism. Conclusion: Pancreatic CT attenuation indexes that are applied to the quantification of pancreatic fat are significantly associated with clinical assessment of impaired glucose metabolism. Accumulation of fat in the pancreas is the commonest cause of pancreatic dysfunction, leading to diabetes mellitus. remains poor despite the progress in treatment over the last decades. Stromal deposition in PDAC is thought to play a crucial role in preventing chemotherapy efficacy. In this study, the tumour and its stroma were targeted using chemotherapy (LDE225 and gemcitabine+nab-paclitaxel). The goal of this research was to assess PDAC's response with intravoxel incoherent motion (IVIM) MRI and to correlate IVIM parameters with overall survival (OS). Material and methods: 34 patients with metastatic PDAC underwent a baseline MRI scan, of whom 21 underwent an additional scan after 8 weeks of treatment. Diffusion-weighted imaging (DWI) was performed at 3.0T using single-shot 2D echo-planar imaging, TR/TE:2145/46ms and 12 b-values (0-600). An IVIM model was fitted using bi-exponential fitting, obtaining diffusion (D), pseudodiffusion (D*) and perfusion fraction (f) maps. Mean values inside the region of interest were used to determine the overall effect of chemotherapy on the tumour, the correlation with OS and the predictive value of these parameters. Results: A significant increase in D was seen during chemotherapy. Furthermore, significant correlations between OS and baseline f were found. The area under the receiver operating characteristic curve (AUC) was the highest for baseline f (AUC=0.835, sensitivity=93%, specificity=77%) to distinguish long (>220 days) from short OS.

Conclusion:
The increase in D during treatment can be explained by necrosis and a decrease in stroma. Lower cellularity leads to higher D. A positive correlation was found between OS and the change in perfusion. IVIM MRI can be used for the evaluation of chemotherapy response on PDAC micro-environment.
Purpose: To assess the diagnostic accuracy of contrast-enhanced US (CEUS) in grading the severity of acute pancreatitis and to correlate CEUS findings with clinical outcome variables such as hospital stay, need for intervention and BISAP score. Material and methods: 56 patients with acute pancreatitis referred for CECT between January 2019 and August 2020 were included in the study, and Bmode USG and CEUS were performed in all these patients. Parameters such as pancreatic size and enhancement, peripancreatic fluid collections and extrapancreatic complications were recorded in CEUS and compared with CECT. Ultrasound severity index (USSI) and modified ultrasound severity index were calculated for each patient and compared with CT severity index and modified CT severity index, respectively.

Results:
The sensitivity and specificity of CEUS in differentiating acute interstitial pancreatitis from acute necrotizing pancreatitis were 93.1% and 96.3%. The sensitivity of CEUS in diagnosing splenic vein thrombosis and peripancreatic fluid collections was 87.9% and 76.9%, whereas the specificity was 100% in both. The agreement between USSI and CTSI was calculated as 0.86 (Cohen's kappa coefficient) and between modified USSI and modified CTSI as 0.85, indicating an almost perfect agreement. No significant differences were noted between USSI and modified USSI in grading the severity of acute pancreatitis and both indices showed a good correlation with clinical outcome variables.
Conclusion: CEUS has a good diagnostic accuracy to detect necrosis and grade the severity of acute pancreatitis and can be used as an alternative in patients where CECT is contraindicated. USSI and modified USSI are equally good indicators to predict clinical outcome variables. Purpose: Pancreatic MR elastography (MRE) could allow for non-invasive estimation of tissue viscoelastic properties and has the potential to better characterize pancreatic inflammation and cancer. However, it remains challenging and requires careful optimization. We aimed to optimize a rapid interleaved multi-slice (Ristretto) gradient echo (GRE) acquisition for pancreatic MRE. Material and methods: Acquisitions were performed with informed consent in 10 healthy volunteers at 3.0T using a gravitational transducer at 3x3x3mm 3 voxel-size (FOV: 384x256x27mm), SENSE2 and four consecutive breath-holds. Vibration frequency (40/50Hz), number of measured wave-phase offsets (4/5), and TE (6.9/9.2ms) were adjusted for optimization. Post-processing was performed using dedicated software. Analysis was performed on shear-wave amplitudes (SWA), deviation from a sinusoid (nonlinearity) and shear-wave speed (SWS) by averaging manually drawn regions-of-interest in the pancreatic head. Repeated measures analysis of variance and pairwise comparison were used to determine intra-subject variability for all quality parameters. Results: The SWA was significantly higher at 40Hz than at 50Hz (40Hz=90+/-28μm; 50Hz=65+/-20μm; F(3,27)=11, p=0.00 , 26 (32.2%) and 2 (2.5%) observations progressed to LR-5 or LR-M categories, respectively. Overall, cumulative incidence of progression to malignant categories was 18.5% at six months, 20.0% at one year, and 22.5% at two years. Cumulative incidence was significantly higher in LR-4 vs LR-3 (logrank P<0.001) and LR-4 vs LR-2 (log-rank P=0.001). Initial LI-RADS category (hazard ratio 4.16, 95% confidence interval 2.04-8.50, P<0.001) was the only independent predictor of progression to malignant categories by multivariate Cox analysis. Conclusion: Initial LI-RADS category is an independent predictor of progression to malignancy categories in indeterminate observations occurring after DAA therapy.

SSD 2.2
Outcome of LR-3 and LR-4 observations without arterial phase hyperenhancement at gadoxetic acid-enhanced MRI follow-up I. Viola, F. Agnello, L. Rabiolo, F. Midiri, L. La Grutta, M. Galia; Palermo/IT Purpose: The aim is to retrospectively evaluate the outcome of LR-3 and LR-4 observations without arterial phase hyperenhancement (APHE) and identify which features could predict LR-5 progression at gadoxetic acid (Gd-EOB-DTPA)-enhanced MRI follow-up. Material and methods: 49 cirrhotic patients (55 LR-3 and 19 LR-4) without APHE were enrolled. LR-3 and LR-4 were classified as decreased, stable or increased in category at follow-up. The presence or absence of major and ancillary LI-RADS features, LI-RADS category and observation size was evaluated. Chi-square and Fisher's exact test were used to assess if baseline LI-RADS features and diameter (<10/ ≥10 mm) were associated with LR-5 progression of LR-3 and LR-4. P < .05 was considered statistically significant. Results: Of 55 LR-3, 17 (31%) progressed to LR-5, 3 (6%) progressed to LR-4, 6 (11%) remained stable in category and 29 (52%) decreased in category. Of 19 LR-4, 8 (42%) progressed to LR-5, 4 (21%) remained stable in category and 7 (37%) decreased to LR-3. Major and ancillary features were not significantly different among LR-3 and LR-4 that progressed to LR-5 and those that remained stable or decreased in category. A diameter ≥ 10 mm significantly increased LR-5 progression risk of LR-3 (OR = 6.07; 95% CI: 0.12; 60.28]; P < .001); LR-4 with a diameter ≥ 10 mm more likely become LR-5 at follow-up (OR = 8.95; 95% CI: 0.73; 111.8; P = .083). Conclusion: LR-3 and LR-4 without APHE were often downgraded or remained stable in category at MRI follow-up. LR-3 and LR-4 with a diameter ≥ 10 mm had an increased risk of LR-5 progression. To investigate the effect of saline-diluted gadoxetic acid, done for arterial-phase (AP) artifact reduction, on signal intensity (SI), and hence focal lesion conspicuity on MRI. Material and methods: We examined 118 patients who each had two serial gadoxetic acid-enhanced liver MRI performed at 1ml/sec, first with non-diluted, then with 1:1 saline-diluted contrast. Two blinded readers independently analyzed the artifacts and graded dynamic images using a 5-scale score. The absolute SI of liver parenchyma, focal liver lesions (if present), aorta, portal vein at the level of the celiac trunk, and paraspinal muscle, as well as the standard deviation of background noise, were measured in all phases. The normalized SI with muscle (SI_norm) and air signal-to-noise ratio (SNR) of these vascular structures and contrast-to-noise ratio (CNR) of focal liver lesions were calculated. Results: AP artifacts were observed in 15 non-diluted (12.7%) and 6 diluted (5.1%) patients. A non-diagnostic score was assigned only in the non-diluted group in three patients (2.5%) and none in the diluted group. Both SI_norm and SNR values in the diluted group were significantly higher in the arterial phase for the aorta and hepatobiliary phase for the liver. The CNR was significantly higher for the lesion in the portal-venous and transitional phases. The inter-rater correlation coefficient was excellent (0.99). Conclusion: Dilution of gadoxetic acid with saline 1:1, administered at an injection rate of 1ml/sec produces images with significantly fewer artifacts and equal or even higher SNR and CNR compared to standard non-diluted contrast. Material and methods: Twenty-seven patients with HCC were prospectively involved undergoing conventional acquired DWI at b-values of 50, 800, 1000, 1200, and 1400 s/mm 2 and synthetic DWI, of which images were calculated at b=1000, 1200, and 1400 s/mm 2 from acquired images using b-values of 200 and 600 s/mm 2 . Two readers reviewed image sets of both synthetic and conventional acquired images of 1000, 1200, and 1400 s/mm 2 independently and rated image quality (4-point scale) and lesion conspicuity (5-point scale). Quantitative analysis includes signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), SIR, and cancer-to-parenchyma contrast ratio. Statistical comparisons were performed using the Friedman test, analysis of variance, and intraclass correlation coefficient (ICC). Results: No significant differences in overall image quality and lesion conspicuity at both synthetic and conventional acquired DWI sets (p>0.05) were found, whereas synDWI 1400 showed lower overall image quality than other DWI sets (p<0.001). SNR, CNR, and SIR showed a continuous decrease as b-values increased, with significant differences in each set (P<0.001), while comparison between synDWI 1000 and conDWI 1000 , synDWI 1200 and conDWI 1200 , synDWI 1400 and conDWI 1400 presented no significant differences (p>0.05). Cancer-to-parenchyma contrast ratio almost unchanging showed no significant difference (p>0.05 To assess the performance of CT texture analysis (CTTA) to predict response to therapy and overall survival in patients with advanced gastric cancer treated with neoadjuvant chemotherapy (n-ChT). Material and methods: Twenty patients with advanced gastric cancer were retrospectively enrolled between December 2015 and February 2019. All participants underwent contrast-enhanced CT at baseline and after n-ChT (FLOT scheme); histology of the specimen after surgical resection was used as the reference standard. An expert radiologist manually draws a volumetric region of interest of the whole tumor before and after n-ChT on CT venous phase at an axial plane. Using a dedicated software (Slicer Radiomics), 109 texture parameters of the first and second order were extracted. Performance of CTTA in the prediction of response to therapy was assessed with receiver operating characteristic (ROC) curve correlated with baseline CT and histologic report; performance of CTTA to predict the overall survival (OS) was tested with ROC curve between CTTA delta (post-nChT and pre-nCHT) and OS. Area under the curve (AUC) with a P<0.05 were considered significant. Results: showed complete response in twelve patients (60%) while partial or non-response was observed in eight patients (40%

SSD 3.7
Local staging of colon tumors with MRI L. Soydan; Istanbul/TR Purpose: Accurate preoperative staging of colon cancer is important to distinguish between patients who can proceed to upfront surgery and patients who may benefit from neoadjuvant chemotherapy. We evaluated the diagnostic accuracy of preoperative MRI in identifying locally advanced colon cancer, extramural venous invasion (EMVI), a parameter indicating poor prognosis and inter-observer variation of the tumor apparent diffusion coefficient (ADC) values of diffusion-weighted imaging (DWI). Material and methods: 40 patients with colon cancer were evaluated using 1.5 T MRI with T2-weighted imaging, DWI, and contrast enhancement. T-stage, N-stage, EMVI and ADC values of the tumors were assessed. Early tumors were defined as T1 to T3ab (<5mm pericolic invasion) and advanced tumors as T3cd (>5mm extramural invasion) or T4. N+ were nodes with short axis >10mm or >5mm with an irregular border/inhomogeneous signal intensity. EMVI + was tumor extension to pericolic vessels. MR findings of two readers who were blinded to pathological findings were compared with postoperative histopathological examination which served as reference.
Results: Diagnostic accuracy of the two radiologists in staging early versus advanced tumors, N-stage, and detection of EMVI was 88% vs /80%, 60% vs 56%, and 66% vs 60% with an inter-observer agreement of i.

Scientific Session On Demand SSD 4
Diffuse liver diseases

SSD 4.1
Automated deep learning CT-based liver volume segmentation: defining normal and hepatomegaly for clinical practice A. Perez 1 , V. Noe-Kim 1 , M. Lubner 1 , R. Summers 2 , P.J. Pickhardt 1 ; 1 Madison, WI/US, 2 Bethesda, MD/US Purpose: Imaging assessment for hepatomegaly is not well defined and currently utilizes suboptimal unidimensional measures. Liver volume provides a more direct measure for organ enlargement. We applied a validated deep learning artificial intelligence (AI) tool that automatically segments the liver for organ volume and sought to establish thresholds for hepatomegaly. Material and methods: Hepatic volumes were derived for 3065 asymptomatic adults (mean age 54.3 years; 1426M/1639F) who underwent MDCT for colorectal screening (n=1960) or renal donor evaluation (n=1104). Linear regression analysis was utilized to assess major patient-specific determinate(s) of liver volume amongst age/sex/height/weight/BSA. The threshold for hepatomegaly was set at two standard deviations above the mean. Accuracy of craniocaudal and maximal 3D linear measures was assessed. Manual liver volume was compared with automated results in 189 patients. Results: Mean standardized automated liver volume was 1533±375 ml and demonstrated a normal distribution. Patient weight was the major determinant of liver volume, with a linear relationship. From this, a linear weight-based upper limit of normal results in hepatomegaly threshold volume (ml)=14.0(Wt)+979. Linear measures demonstrated only moderate performance for identifying volumedefined hepatomegaly; a craniocaudal threshold of 19 cm was 71% sensitive and 86% specific for hepatomegaly, and a maximal 3D linear threshold of 24 cm was 78% sensitive and 66% specific. For the subset (n=189) with manual versus automated comparison, mean difference in hepatic volume was 2.8% (41 ml).

Conclusion:
We derived a simple, weight-based threshold for hepatomegaly using an automated liver volume tool. If further validated in larger healthy and diseased cohorts, this approach could provide a more objective measurement of liver size.

SSD 4.2
CT portography with esophageal variceal measurements may be used for the evaluation of esophageal variceal severity S. Wan, B. Song, X. Zhang, Y. He; Chengdu/CN Purpose: To evaluate the severity of esophageal varices (EV) based on CT portography (CTP) measurements of EV in the distal esophagus and to assess whether CTP can be used as a complementary method for endoscopy. Material and methods: A total of 136 EV patients with clinicopathologically confirmed liver cirrhosis were evaluated. All were examined by CTP within 4 weeks of upper endoscopy, patients were divided into a non-conspicuous EV group (mild-to-moderate EV, n=30) and a conspicuous EV group (severe EV, n=106) according to standard endoscopy. The EVD (EV diameter), CSA (crosssectional surface area of EV), EVV (EV volume), SV (spleen volume) and DLGV (diameter of left gastric vein) were measured independently using 3D-slicer (Boston, USA). These indicators' predictive performances were studied using receiver operating characteristic (ROC) curve analysis, and the area under the curve (AUC), sensitivity and specificity were calculated to distinguish mild-tomoderate from severe EV. Data between the two groups were analyzed by T test or the Mann-Whitney test, a p<0.05 (two-tailed) was accepted as statistically significant for all tests. Results: In those indices, EVD, CSA, EVV and DLGV were larger in the conspicuous group than the non-conspicuous group. The difference between the two groups was statistically significant (p≤0.01 To evaluate the accuracy in the quantification of liver steatosis with 3rd-generation dual-source dual-energy CT (dsDECT) with pathology on the surgical specimen as the reference standard. Material and methods: Patients >18 years old, undergoing liver resection between Jan 2018 and Jan 2020, with an abdominal contrast-enhanced dsDECT within 1 month before resection, were retrospectively included. The exclusion criteria were the lack of contrast-enhanced dsDECT or pathological data. The dsDECT examinations were performed at 80-100/150Sn kV and modulated mA, with a triphasic contrast-enhanced protocol (370 mg[I]/ml, 1,3 ml/kg body weight). Liver steatosis was estimated with three-material decomposition algorithm on a dedicated workstation from arterial and venous datasets. Liver attenuations on virtual non-contrast (VNC) from the same datasets were also recorded. Surgical specimens were reviewed by two experienced pathologists in consensus. The estimates of steatosis from dsDECT datasets were correlated with pathology with receiver characteristic curves (ROC) analysis (endpoint: pathological steatosis ≥30%  To assess the feasibility of a manual segmentation approach in MR elastography (MRE) using a spin-echo echo-planar imaging (EPI) prototype sequence in patients with various hepatic diseases. Material and methods: In total, 111 (46 female, mean age 61.0±13.6 years) individuals were examined at 3T (Magnetom PRISMAfit, Siemens Healthineers, Erlangen, Germany) using a prototype spin-echo EPI sequence. The derived MRE images were independently evaluated by two independent readers, both qualitatively and quantitatively (in kPa) with strict adherence to the QIBA Consensus Profile for MRE of the liver of the Radiological Society of North America.

Results:
The wave propagation was graded as disorganized in 11 and 9 cases, suboptimal in 34 and 28 cases and optimal in 66 and 74 cases, by both readers, respectively. Mean stiffness was 3.2±1.8 kPa for reader 1 and 3.4±1.8 kPa for reader 2. The agreement of both readers was excellent with an ICC of 0.969 (95%CI 0.950-0.980) with only a minimal bias of -0.14 kPa (95%CI -0,228023 to -0,0532536). There was no correlation between disagreement and increasing kPa (r sp =0.019, p=0.856). Also, the mean difference between both readers did not differ between groups with better or worse wave propagation (p=0.549 and p=0.584). Conclusion: Based upon the presented data, assessing the agreement of manual stiffness assessment in MRE at 3T using a spin-echo EPI prototype sequence, the interobserver agreement was excellent. However, evaluating the cause-specific sources for suboptimal wave propagation is warranted.

SSD 4.5
Interpretable machine learning for predicting stereotactic body radiation therapy early response in liver colorectal cancer metastasis S. Waktola, D. Van Der Velden, F. Castagnoli, R.G.H. Beets-Tan; Amsterdam/NL Purpose: To develop a framework for automated early prediction of treatment response after liver stereotactic body radiation therapy (SBRT). Material and methods: 65 patients with pathology-confirmed diagnosis of liver colorectal cancer metastasis and treated with SBRT within our institute, between 2008 and 2020, were retrospectively analysed. The dataset consists of baseline CT (before the start of therapy) and follow-up CT scans (delivered after 4 weeks of therapy). First, all the liver lesions were manually delineated by a radiologist and then categorized into two main classes based on SBRT response: complete response (disappearance of all target lesions), and not responsive (progressive disease). We developed an interpretable machine-learning model by extracting significant features using radiomics and then predicting early response (i.e. before the patient goes to any radiation therapy). To interpret the model predictions, we used model-agnostic explanations (LIME) and Shapley Additive exPlanations (SHAP). Results: During model training, 5-fold cross-validation was performed. Both LIME and SHAP models were able to predict the early SBRT response with a mean AUC of 0.72 and Mann-Whitney U test of P < 0.003.

Conclusion:
These results indicate that radiomics-based machine-learning models could potentially provide non-invasive biomarkers for early prediction of SBRT treatment response and improving patient stratification for personalized medicine. Interpretable models can pave for better acceptance of machine-learning techniques and to be practically adopted in real clinical settings. However, their evaluation with large and multicenter datasets needs to be researched further. To investigate the accuracy of spleen volume for predicting hepatic decompensation in patients with chronic liver disease (CLD). Material and methods: 402 patients with CLD who had undergone gadoxetic acid-enhanced liver MRI were included. Spleen volumetry was measured either using Syngo-Via software or a 2D-volume-approximation method using only axial and cranio-caudal maximal diameter. Patients were stratified into three groups according to fibrosis stage and present or past hepatic decompensation: non-advanced CLD, compensated-advanced CLD (cACLD), and decompensated-advanced CLD (dACLD). Pearson correlation coefficient assessed the relationship between spleen volume measured by the two above-mentioned methods. The predictive value of spleen volume for first or further hepatic decompensation was investigated using Kaplan-Meier analysis, log-rank tests, and Cox regression analysis. Results: In a subset of 238 patients, we found a strong positive correlation of spleen volume between the 3D-segmentation method and 2D-volume approximation method (R=0.910; p<0.0001). We, therefore, used the simpler and faster approximation method for further analysis. In cACLD patients (n=197)  were given a preliminary diagnosis of DS versus non-DS on T2-weighted MRCP and later HBP GA-T1-MRC by two independent readers, blinded to all patient data. The final diagnosis of DS versus non-DS was based on LFTs, ERCP, and histology, if available. Using societal guidelines, DS was diagnosed on T2-weighted MRCP if common bile duct (CBD) and right or left hepatic main duct diameters were ≤ 1.5 mm and ≤ 1 mm, respectively. We diagnosed DS on GA-T1-MRC if no GA excretion was seen on 20-minute HBP images. The differences between DS and non-DS in both groups and correlation with LFTs and splenic volume were calculated using Mann-Whitney and Spearman tests, respectively.

Results:
The inter-reader agreement was poor (k< 0.2) for the diagnosis of DS onT2-weighted MRCP, but excellent (k=0.9) for DS diagnosis on HBP-T1-MRC. LFTs and splenic volume were significantly higher in patients lacking excretion compared to those with timely GA excretion on HBP (p=0.001). Our definition of DS in PSC correlated well with functional definition based upon HBP GA-T1-MRC and LFTs (p=0.001), but not with societal guideline-defined measurements (p>0.05). Conclusion: Punctual contrast excretion on HBP GA-T1-MRC is reproducible and correlated well with the presence of DS and LFTs.

Scientific Session On Demand SSD 5
Abdominal Oncology Birmingham/UK, 4 Northwood/UK Purpose: Hypoxia and angiogenesis are recognised as important drivers of colorectal tumour growth and dissemination and occur early in the adenomacarcinoma sequence. We explored the biological associations between primary colorectal cancer perfusion CT-derived vascular measurements and immunohistochemistry markers of angiogenesis and hypoxia in the PROSPECT trial. Material and methods: Adult participants with suspected or proven primary colorectal cancer were recruited prospectively from 13 hospitals (ISRCTN 95037585). Exclusions were metastatic disease at staging, prior cancer(s), and contraindications to contrast agents. In addition to standard staging and pathology investigations, participants underwent perfusion CT (blood flow, blood volume) and immunohistochemistry for angiogenesis (CD105, VEGF) and hypoxia (HIF-1, GLUT-1) markers. For CD105 expression, the two areas of highest vascularization (hot spots) were averaged and given as a count per mm2. Scores for VEGF (0-3), Glut-1 (0-8), and HIF-1α (0-6) were based on the intensity and percentage of staining. Correlations between imaging and immunohistochemistry were assessed.

SSD 5.2
Prognostic value of pre-operative CT lymph node features and location in stage III colon cancer patients E.K. Hong 1 , F. Landolfi 2 , F. Castagnoli 3 , R.G.H. Beets-Tan 1 ; 1 Amsterdam/NL, 2 Rome/IT, 3 Brescia/IT Purpose: To evaluate the prognostic value of preoperative CT lymph node location and features in stage III colon cancer patients. Material and methods: A total of 176 consecutive stage III colon cancer patients who underwent curative surgery without pre-operative treatment from January 2011 to December 2017 were retrospectively included. The size of the largest lymph node, presence of at least one lymph node with round shape, internal heterogeneity and irregular outer border were assessed on a patient level and the anatomical location (peritumoral, mesenteric and apical) of the lymph node with each imaging feature was recorded and analyzed for prediction of prognosis. Results: Cox regression analysis showed short and long diameters of the largest lymph node, locations of the largest and round-shaped lymph node to be prognostic factors for tumor recurrence (P = 0.004, 0.001, 0.012 and 0.046, respectively). From Kaplan-Meier survival analysis, the 3-year recurrence-free survival (RFS) rates were lower in patients with specific imaging features of lymph nodes than the patients without (66% vs. 90%, P < 0.001 for internal heterogeneity and 77% vs. 90%, P = 0.014 for irregular outer border). The 3-year RFS rate was significantly lower in the group with the largest lymph node located in the apical region than other regions (87%, 78% and 57% for peritumoral, mesenteric and apical, respectively, P = 0.028). Conclusion: Pre-operative radiologic imaging features and location of lymph nodes are predictors of recurrence-free survival in stage III colon cancer patients. . Several studies have shown an additional value for neoadjuvant chemotherapy (nCHT) before CRS-HIPEC. If CAIRO6, a large multicenter study investigating the added value of nCHT, confirms this, it could affect current guidelines. However, the performance of MRI after nCHT for staging PM in CRC patients has never been investigated. Therefore, the aim is to determine whether DW-MRI can accurately select CRS-HIPEC candidates after nCHT. Material and methods: Patients with PM from CRC and appendiceal origin who received nCHT followed by a diagnostic laparoscopy (DLS) or CRS-HIPEC from January 2016 to August 2020 were eligible. Two radiologists assessed the PCI on DW-MRI (MRI-PCI) after nCHT. The surgical PCI (S-PCI) was subtracted from patient files. The reference standard was histology PCI (H-PCI). The main outcome was the accuracy of DW-MRI after nCHT in predicting whether patients were eligible for CRS-HIPEC. Results: Thirty-three patients were included. Both readers detected all 24 patients (24/33) with resectable disease. Seven out of nine patients with unresectable disease during staging surgery were detected with MRI. The intraclass correlation (ICC) between both readers was excellent (0.89 (0.75 to 0.95)). ICC between S-PCI and MRI-PCI was 0.87 (0.66 to 0.94). S-PCI had similar correlation with H-PCI (0.90 (0.76 to 0.96)) than MRI-PCI (0.89 (0.74 to 0.95)). Conclusion: MRI is a promising tool to re-assess the PCI after neoadjuvant chemotherapy to guide patient selection. Patients were excluded if they had undergone treatment, local or systemic, between clinical diagnosis and the MRI scan. Patient characteristics, like primary tumor location and the peritoneal cancer index (PCI), were retrieved from patient records and radiological reports. The frequency of affected PCI regions was assessed and compared between tumor sidedness. Results: 126 patients were included with a median age of 65 (IQR: 56-72). 46% percent were male. No difference in mean PCI or number of regions affected was found between patients with a right-sided or left-sided colon tumor (p=0.30 and p=0.44, respectively). PM were found most frequently in close proximity to the primary tumor. In right-sided tumors, small bowel and upper regions were significantly more affected (both p=0.04 ) than in left-sided tumors. Conclusion: Colorectal peritoneal metastases seem to spread from the close proximity of the primary tumor to further abdominal sites via known peritoneal fluid flows. Known unfavorable sites for cytoreductive surgery (small bowel and upper abdominal sites) are, therefore, more frequently reached from rightsided tumors.

SSD 5.5
Diagnostic impact of CT slice thickness in detecting high-risk colon cancer E.K. Hong 1 , S.J. Park 2 , J.M. Lee 2 , R.G.H. Beets-Tan 1 ; 1 Amsterdam/NL, 2 Seoul/KR Purpose: To evaluate the diagnostic impact of different slice thickness of CT in optimizing detection of high-risk colon cancer. Material and methods: A total of 42 patients who underwent curative surgical resection without pre-operative treatment of colon cancer was retrospectively included in this study. The portal phase scan of each patient was reconstructed into 1-mm-, 3-mm-and 5-mm-thick images and utilized for analysis. Tumor and lymph node staging of colon cancer were independently performed by 2 radiologists on each set of images. Sensitivity, specificity and area under the receiver operating characteristic curve (AUC) were calculated and compared between the reconstructed images with different slice thicknesses. Results: The mean AUC, sensitivity and specificity of distinguishing tumor invasion beyond muscularis propria were 0.733, 83.3 and 63.3 for 1-mm-, 0.725, 91.7 and 53.3 for 3-mm-and 0.667, 66.7 and 67.4 for 5-mm-thick images. There was a statistically significant difference in diagnostic performances between using 1-mm-, 3-mm-and 5-mm-thick images in detecting over pT3 stage tumors (P = 0.003). For identifying the lymph node involvement of colon cancer, the mean AUC, sensitivity and specificity were 0.550, 56.2 and 53.8 for 1-mm-, 0.526, 43.7 and 61.5 for 3-mm-and 0.538, 50.0 and 58.7 for 5-mm-thick images. There was no statistically significant difference in diagnostic performances for detecting lymph node involvement when using different slice thicknesses.

Conclusion:
The difference in CT slice thickness resulted in a significant difference in detecting high-T stage colon cancer. However, there was little or no advantage in reducing slice thickness in detecting lymph node involvement of colon cancer.

SSD 5.6
Role of CT texture analysis in predicting peritoneal carcinomatosis in patients with gastric cancer G.M. Masci, F. Iafrate, F. Ciccarelli, D. Grasso, A. Laghi, C. Catalano; Rome/IT Purpose: The aim of the study was to perform CT texture analysis in patients with gastric cancer (GC) to investigate the potential role of radiomics in predicting the occurrence of peritoneal carcinomatosis (PC).

Material and methods:
In this single-center retrospective analysis, GC patients with and without synchronous PC (group PC and group non-PC, respectively) were enrolled, based on surgically confirmed degree of peritoneal involvement. Pre-operative contrast-enhanced CT examinations were evaluated. Texture analysis was performed on portal phase images: the region of interest was manually drawn along the margins of the primitive lesion on each slice and the volume of interest of the whole tumor was obtained. A total of 38 texture parameters were extracted and analyzed. ROC curves were performed on significant texture features. Multiple logistic regression was conducted on features with the best AUC to identify differentiating variables for both groups. Results: Ninety patients with GC (group PC, n=45; group non-PC, n=45) were included. T1/T2 tumors were prevalent in group non-PC, whilst T3 was significantly associated with group PC. Significant differences were observed for 22/38 texture parameters. Features with the highest AUC in ROC curve analysis were volume and GLRLM_LRHGE (0.737 and 0.734, respectively), which were found to be independent differentiating variables of group PC in the multiple regression analysis (OR 8.44 The greater omentum is a common site of malignant peritoneal spread and is a useful target for percutaneous biopsy. We evaluated the diagnostic efficacy of percutaneous US-guided omental biopsy, correlated with pre-biopsy CT findings. Material and methods: Inclusion criteria consisted of US-guided biopsy with pre-biopsy CT available for review at a single academic institution. Demographics, clinical, and imaging findings at pre-biopsy CT and US biopsy were evaluated. Cases of CT-guided omental biopsy were also assessed. Results: 163 patients who underwent US-guided omental biopsy were included (mean age, 65±12 years; 120F/43M; mean BMI, 28.9±7.9). On pre-biopsy CT, omental disease appeared infiltrative in 127 (78%) and mass-forming in 36 (22%) cases. Infiltrative soft-tissue component varied from 10% to 100% relative to fat, correlating with mean attenuation (r=0.83; p<0.01), and with hypoechoic (n=105) versus hyperechoic (n=58) US appearance (p<0.01). Mean omental thickness was 2.6±1.2 cm at the biopsy site (range, 0.7-6.7 cm). Biopsies (156 cases 18-gauge core, 7 cases FNA; mean passes 2.5) were diagnostic in 155 (95%) cases. Gynecologic (n=82; 50%) and gastrointestinal (n=45; 28%) malignancies were most common. 106 (65%) cases had extra-omental disease potentially amenable to biopsy. Only eight CT-guided omental biopsies were performed over this period, six (75%) of which followed planned US biopsy (non-visualization>non-diagnostic). No complications of biopsy were reported. Conclusion: US-guided core biopsy of suspected omental disease provides a safe and effective means for tissue diagnosis. Although the typical pattern of omental infiltration often lacks a discrete target at US, tissue sampling of the thickened structure nonetheless yields a diagnosis in the vast majority of cases.

SSD 5.11
Reporting colon cancer staging using a template M.R. Pedersen, C. Dam, M. Loft, S. Rafaelsen; Vejle/DK Purpose: The purpose of this study was to evaluate the effect of completeness of the radiological reports in primary local staging colon cancer when using a template.

Material and methods:
The study used primary staging reports retrieved from the departments RIS/PACS. Five key tumour descriptors were evaluated within each report: tumour morphology (polypoid or annular), information on tumour breach of the colon wall (≥ T3), tumour out-growth in mm, nodal status and TNM in conclusion. The failure to provide a description of the presence or absence of a feature in a report counted as 'not reported'. To allow comparisons between reporting styles, the template or free-text style of reporting was also recorded. Results: During a two-year period, a total of 666 patient CT reports were evaluated at the colorectal center multidisciplinary team (MDT) conference. In 200 of these reports, a template was used. Information on tumour morphology (polypoid or annular) was present in 81% of the template reports vs 9% in freetext style. The figures in percentage for information on tumour breach of the colon wall (≥ T3) were 93% vs 48%, tumour out-growth in mm: 51% vs 17%, nodal status: 99% vs 86% and TNM in conclusion: 98% vs 51%, P < 0.0001.

Conclusion:
The present study provides additional support for the routine use of template reports to improve imaging reporting standards in colon cancer.

SSD 5.12
CT Purpose: This systematic review will determine the risk of gallbladder cancer (GBC) in patients with gallbladder polyps (GBP) detected on trans-abdominal ultrasound (TAUS). If possible, meta-analysis will estimate the effect size of identified prognostic factors for the development of GBC. Material and methods: Primary studies reporting prognostic factors in patients with GBP will be systematically searched from databases including MEDLINE, Embase and Cochrane Library. This review will include observational cohort, cross-sectional and case-control studies. Included studies will have aimed to determine the natural history of gallbladder polyps and/or evaluated the risk of gallbladder malignancy by monitoring a defined cohort of patients. Studies unavailable in the English language will be excluded. Risk of bias will be assessed using the Quality in Prognostic Factor Studies (QUIPS) tool. The strength of the overall weight of evidence will be judged using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group methodology. This review has been registered with PROSPERO (CRD42020223629).

Results:
This systematic review will be reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidance. After abstract screening, data from full-text articles meeting the inclusion criteria will be extracted by two independent observers. If heterogeneity is considerable (I 2 > 75%), then quantitative data synthesis will not be performed. Conclusion: GBP are a common finding on TAUS, but the risk of developing GBC is low. This systematic review will identify important prognostic factors for GBC and will inform the update of ESGAR joint society GBP guidelines.

Scientific Session On Demand SSD 6
Acute abdominal conditions and miscellaneous We aimed to assess the prevalence and types of variations of coeliac and hepatic arteries and their relationship with surgical complications in patients undergoing pancreatoduodenectomy.

Material and methods:
In this single-center study, we retrospectively reviewed the contrast-enhanced abdominal computed tomography scans of the patients with periampullary cancer imaged before surgery. The coeliac artery branching with a particular focus on hepatic artery origin was assessed on arterial phase images. Results: A total of 66 patients [39 (59.1%) males] with a mean age of 64±13 were included. The most frequent tumor origin was pancreatic head in 41 (62.1%) cases, followed by ampulla in 15 (22.7%) cases. Arterial variations were observed in 15 (22.7%) cases. These included 11 cases with hepatic artery (HA) origin variations (right HA originating from superior mesenteric artery in 8, right HA originating from aorta in 1, left HA originating from aorta in 1, both left and right HA originating from coeliac trunk in 1 case) and 4 coeliac trunk variations (one each case from Uflacker types 1, 2, 3 and 5). Total perioperative complications occurred in 14 cases (13 fistulae and 1 hematoma). The higher rate of complications in the variation group [5 (33.3%)] as compared to no-variation [9 (17.6%)] group did not reach statistical significance (p=0.279).

Conclusion:
The numbers of coeliac and hepatic arterial variations are quite common but do not seem to significantly increase complications in patients undergoing pancreatoduodenectomy.

SSD 6.6
Automated segmentation of the stomach on MRI using a MultiRes Unet T.J. Wright, A. Bard, H. Fitzke, S.A. Taylor; London/UK Purpose: Quantitation of GI motion using MRI shows considerable promise. However, a current limitation is the need for time-intensive manual segmentation of images. We examined the performance of two neural network architectures to determine the feasibility of fully automated gastric segmentation. Material and methods: Datasets from 196 patients undergoing MR enterography and prepared with oral mannitol were collated from an anonymous database. A single frame from a 20-second dynamic balanced gradient echo sequence was automatically selected based on the shortest Euclidean distance to the pixel-wise median over time. The reader manually segmented the stomach wall in the selected frame using Horos (Annapolis, MD USA) that was reviewed by an expert with >8 years' experience and became the ground truth. The performance of two convolutional neural network architectures (Unet and MultiRes Unet) was assessed with three different loss functions (binary crossentropy, binary crossentropy weight balance, and Dice coefficient loss). Two measures of agreement (intersection over union, IOU; Dice coefficient) and two measures of distance (Hausdorff distance, HD; mean contour distance, MCD) were used to test segmentation accuracy against the ground truth. Results: The best performing architecture was a MultiRes Unet with a Dice coefficient loss function, generating metrics of 0.69 IOU, 0.76 Dice, 30.4mm HD, and 5.25mm MCD. However, 15% of individual overlap scores were 0, indicating complete segmentation failure. Conclusion: This method has potential for the automatic segmentation of gastric volumes, but requires optimisation. Larger training datasets, higher quality images, a tailored acquisition protocol, and a user-defined 'seed' point may reduce the number of failed segmentations.

SSD 6.7
Patients with small gallbladder polyps: a long-term follow-up study M.R. Pedersen, P.O. Otto, S.R. Rafaelsen; Vejle/DK Purpose: Gallbladder polyps are lesions that protrude from the inside of the gallbladder wall into the cavity and are generally detected by abdominal ultrasonography. The knowledge about the growth rate of gallbladder polyps < 6 mm is limited, especially in patients without risk factors. The aim of this study was to examine the long-term follow-up growth in gallbladder polyps < 6 mm in size and to explore the risk of developing gallbladder cancer. Material and methods: Abdominal ultrasonography reports from 2007 to 2009 were reviewed, including reports on patients diagnosed with a gallbladder polyp (polyp size < 6 mm) during the 2007-2009 period. The patients were invited to a final follow-up ultrasonography of the gallbladder conducted during October 2019 to February 2020. A total of 154 patients were included (100 women and 54 men). Results: In 53 patients (34.4%), the polyp was not visible at the ultrasonography follow-up. Gallbladder polyps were found in 101 (65.6%) patients. Single polyp was found in 49 patients (31.8%) and 52 (33.8%) patients had multiple polyps. The median polyp size was 4 mm (range: 2.0-5.9 mm) at baseline compared with 4 mm (range: 1.7-15.0 mm) at the follow-up. A total of 15 patients experienced polyp growth of 2 mm or more. None developed gallbladder cancer. Conclusion: Our study showed that gallbladder polyps < 6 mm had a low probability of increasing in size. No cases of gallbladder cancer were observed among the patients. The need for follow-up in patients with small gallbladder polyps is questionable.

Scientific Sessions
Authors' Index SCIENTIFIC SESSIONS / ON DEMAND 29 SSD 6.8 Imaging manifestations of acute GI bleeding: a casebased review with management U. Gupta, S. Chinnappan, R. Ramachandran, V.S. P.M.; Chennai/IN Purpose: To retrospectively analyse the unexplained causes of hypotension in patients presenting with hematochezia or hematemesis in a non-trauma setting using CT angiography and correlate with digital subtraction angiography (DSA) and endoscopy. Material and methods: A series of cases of acute GI bleed presenting with hypotension in non-traumatic patients were chosen from a period of five years from 2015 to 2020 whose CT angiogram was done in Philips brilliance 16 slice scanner and GE revolution 128 slice CT scanner. The imaging findings were reviewed and correlated with DSA, intraoperative findings and endoluminal procedures like endoscopy and colonoscopy guided ligation. Results: A total of 16 cases were identified and reviewed: spontaneous gastric perforation (n=1); aortogastric fistula (n=1); intramural hematoma of oesophagus following thrombolysis (n=1); oesophageal and gastric varices (n=2); perforated Meckel's diverticulum (n=2); Henoch-Schonlein pupura vasculitis involving small bowel loops (n=3); angiodysplasia of caecum, distal ileum and rectum (n=3); bleeding caecal diverticulum (n=1); graft versus host disease (n=1); arteriovenous malformation of rectum (n=1). The management included medical management (4 out of 16), surgery (3/16), DSA and embolization (3/16) and colonoscopy/endoscopy-guided ligation of bleeder (3/16). 3 patients succumbed to hypovolemic shock before initiation of any definite treatment. Conclusion: Imaging plays an important role in the diagnosis and management of patients with acute GI bleeding. CT angiography is one of the most important diagnostic modalities with easy availability. By diagnosing the cause of acute GI bleeding, we could help guide endoscopic, endovascular or surgical management and appropriate treatment. Radiologist plays a significant role in accurate diagnosis and guides the clinician for definitive management.

SSD 6.9
Influence Clichy/FR Purpose: Acute mesenteric ischemia (AMI) may be underdiagnosed when not clinically suspected before CT is performed. We assessed the influence of clinical suspicion of AMI on CT accuracy. Material and methods: This retrospective single-centre study included patients who underwent CT in 2014-2019 and had clinically suspected AMI and/or confirmed AMI. CT protocols were adapted based on each patient's presentation and on findings from unenhanced images. The CT protocol was considered optimal for AMI when it included arterial and portal venous phases. CT protocols, accuracy of reports, and outcomes were compared between the groups with and without suspected AMI before CT. Results: Of the 375 events, 337 (90%) were suspected AMI and 66 (18%) were AMI, including 28 (42%) with and 38 without suspected AMI. These two groups did not differ significantly regarding the medical history, clinical presentation, or laboratory tests. The CT protocol was more often optimal for AMI in the group with suspected AMI (

SSD 6.10
The Material and methods: 30 patients (23 women, 7 men) with clinical symptoms of abdominal angina and suspected atherosclerotic etiology were included in the study. Each patient was evaluated by US (B-mode, color and spectral Doppler) and CT angiography (CTA), as a reference method. Based on the results of a diagnostic imaging workup, patients were referred for endovascular treatment. FU Doppler US was performed at 4-6 weeks after the intervention. Results: Doppler US examination revealed significant stenoses of the visceral arteries in 27 patients; 15 superior mesenteric arteries (SMAs) and 12 co-existing celiac trunk and SMA narrowings. Doppler US failed to detect visceral stenoses in 3 cases (1 SMA and 2 celiac trunk and SMA stenoses), for which CTA indicated significant stenosis. Each of these patients underwent endovascular treatment (angioplasty with/without stenting). FU Doppler US was performed at 4-6 weeks after intervention and confirmed undisturbed visceral blood flow in 27 patients; 3 patients required CTA due to failure to visualize visceral arteries. Conclusion: Doppler US evaluation constitutes a valuable alternative for CTA both in primary diagnostics and FU of patients with abdominal angina. Based on the study results, Doppler US is characterized by high diagnostic efficacy, and CTA should be required only in individual cases.