ESGAR 2017 Book of Abstracts

REVIEWING PANEL S590 O. Akhan, Ankara/TR C. Aubé, Angers/FR M.A. Bali, London/GB I. Bargellini, Pisa/IT T.V. Bartolotta, Palermo/IT A. Ba-Ssalamah, Vienna/AT R.G.H. Beets-Tan, Amsterdam/NL E. Biscaldi, Genoa/IT A. Blachar, Tel Aviv/IL P. Boraschi, Pisa/IT R. Bouzas, Vigo/ES G. Brancatelli, Palermo/IT D.J. Breen, Southampton/GB V. Cantisani, Rome/IT F. Caseiro Alves, Coimbra/PT J. Cazejust, Paris/FR N. Courcoutsakis, Alexandroupolis/GR L. Crocetti, Pisa/IT L. Curvo-Semedo, Coimbra/PT C. De Cecco, Charleston, SC/US R. Dondelinger, Liège/BE M. D’Onofrio, Verona/IT N. Elmas, Izmir/TR H. Fenlon, Dublin/IE A. Filippone, Chieti/IT N. Flor, Milan/IT B. Fox, Plymouth/GB A.H. Freeman, Cambridge/GB A. Furlan, Pittsburgh, PA/US Y. Gandon, Rennes/FR V. Goh, London/GB M. Gollub, New York, NY/US S. Gourtsoyianni, London/GB S. Gryspeerdt, Roeselare/BE L. Guimaraes, Porto/PT J.A. Guthrie, Leeds/GB A. Hatzidakis, Heraklion/GR T. Helmberger, Munich/DE P. Huppert, Darmstadt/DE


A T H E N S G R E E C E
E u r o p e a n S o c ie t y o f G a s t r o in t e s t in a l a n d A b d o m in a l R a d io lo g y  Purpose: Recent evidence indicates that longer pre-operative imaging to surgery time interval (ISI) for pancreatic ductal adenocarcinoma (PDAC) increases the risk of unexpected progression (UP) at laparotomy. We assessed the relationship between ISI and survival. Material and methods: Retrospective review of all patients who underwent attempted resection of PDAC at our institution between January 2010 and December 2015. Patients who received neoadjuvant therapy were excluded. All were followed up until 31st June 2016. Survival was determined from the national death register. The population was divided based upon ISI ≥ or < 25 days. Kaplan Meier survival analysis was performed.

588
Results: 240 patients were classi ed as resectable on pre-operative abdominal CT and underwent laparotomy. UP was found in 30 (13%) cases. The ISI was longer for patients with UP compared to those without (51.7 vs 35.1 days, p<0.05). When intention-to-treat analysis was performed, there was no difference in survival between patients with ISI ≥ 25 (median 19.8 months) vs < 25 (12.6 months), however, in those who underwent resection, an ISI ≥25 days was associated with longer survival (25.2 vs 19.6 months, p<0.05).
Conclusion: Performing surgery for PDAC within 25 days of pre-operative abdominal CT reduces risk of UP, however, this does not improve survival. In fact, longer ISI is associated with survival advantage for those who are resected. Lack of progression over a longer pre-operative ISI likely indicates more indolent disease. Peschiera del Garda/ IT Purpose: To evaluate the added value of CT texture analysis in the study of neuroendocrine pancreatic neoplasms. Material and methods: 24 patients with neuroendocrine pancreatic neoplasms were included in this study. All tumors were pathologically diagnosed after resection or by means of biopsy. Histological grade of these neoplasms was available in all cases. There were 7 G1, 7 G2, and 10 G3 neoplasms with presence of metastases in 21/24 (87%) cases. Three-dimensional CT texture analysis of the primary tumor was performed comparing the results with the tumor grading using Wilcoxon correlation test. Results: Pancreatic neoplasms were located in the head in 7/24 (29%) cases; in the body-tail in 17/24 (71%) cases. G1 neoplasms show mean dimension of 41 mm (range 9-112 mm). G2 neoplasms show mean dimension of 42 mm (range 21-73 mm). G3 neoplasms show mean dimension of 50 mm (range 16-111 mm). CT texture analysis was nalized in all cases. There was no statistical signi cant difference regarding all the parameters of CT texture analysis among G1, G2, and G3 neoplasms. Higher value of kurtosis was found in G3 (Kurtosis median=2,1499000) in respect to G1 (Kurtosis median=0,1128400) and G2 (Kurtosis median=1,3848000) neoplasms. Grouping together G1 and G2 neoplasms, statistical signi cant difference (p< 0,05) was found regarding the only kurtosis CT texture analysis parameter. Material and methods: This study included 35 surgically proven HCCs in 30 patients, who were examined by gadoxetic acid-enhanced MRI using differential sub-sampling with Cartesian ordering (DISCO). Multiphasic HAP imaging was started at 30s after the injection and completed within a single breathhold (acquisition time=22-26s, temporal resolution=~4s). The HCCs were classi ed per their enhancement patterns: ring-like enhancement in all phases (group A; n=7) and total enhancement in any phase (group B; n=28). A single pathologist evaluated the ratio of poorly differentiated component and necrosis, and thickness of the capsule for all nodules.

Conclusion:
The HCCs represented by ring-like enhancement on all phases of DISCO had more necrosis and thicker capsules compared to classical HCCs.

SS 2.4
Association between non-hypervascular hypointense nodules on gadoxetic acid-enhanced MRI and hepatic fibrosis or hepatocellular carcinoma J.A. Hwang, T.W. Kang, J.E. Lee, Y.K. Kim, S.H. Kim; Seoul/ KR Purpose: The risk factors for the occurrence of non-hypervascular hypointense nodules (NHHN) on gadoxetic acid-enhanced magnetic resonance imaging (MRI) are still unclear. We assessed the association between these nodules and hepatic brosis or hepatocellular carcinoma (HCC) in patients with chronic liver disease, and analyzed their progression to overt HCC. Material and methods: Between August 2012 and March 2016, 714 consecutive patients who had undergone transient elastography for liver stiffness (LS) measurement and gadoxetic acid-enhanced MRI were investigated. The association between the presence of NHHN on the hepatobiliary phase and LS, and the patient's HCC status was assessed using multivariate logistic analysis. In patients with these nodules, cumulative progression rates of nodules to overt HCC were compared with rates of new HCC development in other parts of the liver using the strati ed log-rank test.
Results: The prevalence of NHHN was 16.8% (120/714). The presence of NHHN was signi cantly associated with the log LS (Odds ratio [OR], 1.48, p = 0.002) and hepatitis B virus infection (OR, 3.14, p = 0.017). The two-year cumulative progression rate of overt HCC from NHHN and rate of progression to HCC in other parts of the liver were 34.1% and 18.3%, respectively (p = 0.071).
Conclusion: The presence of NHHN was not associated with the patient's HCC status but was associated with hepatic brosis and hepatitis B virus infection. Furthermore, these lesions frequently progressed to overt HCC.

SS 2.5
Adding intratumoral pathomorphologic ancillary features to conventional enhancement patterns of hepatocellular carcinoma on gadoxetic acid-enhanced MR imaging improves diagnostic performance Y.K. Kim, J.H. Min, W.K. Jeong; Seoul/ KR Purpose: To assess the added value of intratumoral ancillary features to enhancement pattern-based diagnosis of hepatocellular carcinoma on gadoxetic acid-enhanced MRI. Material and methods: A total of 773 consecutive patients with surgically resected 773 primary hepatic tumors (699 HCCs, 63 intrahepatic cholangiocarcinomas, and 11 benign nodules) who underwent gadoxetic acid MRI were retrospectively identi ed. Enhancement patterns and three ancillary features of capsule, septum and T2 spotty hyperintensity were assessed by two radiologists. Performance of enhancement pattern-based diagnosis of HCC was compared to diagnosis with enhancement pattern plus ancillary features.
Results: Enhancement patterns were positive (arterial hyperenhancement with washout) for 562 (72.7%) tumors, negative (no arterial hyperenhancement and no washout) for 75 (9.7%), and inconclusive (either no arterial hyperenhancement or no washout) for 136 (17.6%). Capsule was observed in 498 (64.4%) tumors, septum in 521 (67.3%), and T2 hyperintensity in 107 (13.8%). The accuracy and sensitivity of HCC diagnosis was improved signi cantly after adding at least one ancillary feature compared with enhancement patternbased diagnosis of HCCs ( Purpose: To evaluate the value of gadoxetic acid-enhanced and diffusionweighted (DW) MR imaging for distinguishing malignant from benign hyperintense nodules on unenhanced T1-weighted images (T1WIs) in patients with chronic liver disease. Material and methods: Forty-two patients with 37 malignant and 41 benign hyperintense nodules on unenhanced T1WIs who underwent gadoxetic acidenhanced and DW MR imaging, followed by histopathological examination, were included. Qualitative (signal intensity [SI], arterial enhancement, washout) and quantitative (size, contrast enhancement index, tumor-to-liver SI ratio, mean and minimum apparent diffusion coef cient [ADCmean and ADCmin] values) analyses were conducted by two radiologists in consensus. Signi cant imaging ndings on univariate and multivariate analyses were identi ed and their diagnostic performances were analyzed for predicting malignant hyperintense nodules. Results: In univariate and multivariate analyses of qualitative and quantitative variables, hyperintensity on T2WI (OR, 13.58; P = 0.02), arterial enhancement (OR,8.21; P = 0.002), ADC min ≤ 0.83 x 10 -3 mm 2 /s (OR, 6.88; P = 0.008) were independently signi cant factors for differentiating malignant from benign hyperintense nodules. When two of these three criteria were combined, 75.7% (28/37) of malignant nodules were identi ed with a speci city of 92.7%, and all three criteria were satis ed, the speci city was 97.6%. Conclusion: Gadoxetic acid-enhanced and DW MR imaging may be helpful for differentiating malignant from benign hyperintense nodules on unenhanced T1WI.
SCIENTIFIC SESSIONS / WEDNESDAY, JUNE 21, 2017 S595 SS 2.7 Comparison of the accuracy of AASLD and LI-RADS criteria for the non-invasive diagnosis of small HCC O. Fouque 1 , M. Ronot 1 , M. Esvan 2 , J. Lebigot 3 , C. Aubé 3 , V. Vilgrain 1 ; 1 Clichy/ FR, 2 Paris/ FR, 3 Angers/ FR Purpose: To prospectively compare the diagnostic accuracy of the AASLD and LI-RADS criteria for the non-invasive diagnosis of small HCC. Material and methods: Between April 2009 and April 2012, patients with cirrhosis and one to three 10-30 mm nodules were enrolled and underwent CT and MR imaging. The diagnostic accuracy of the combination of the AASLD and the LI-RADS criteria were determined based on their sensitivity, speci city, positive (PPV) and negative predictive values (NPV) with 95% con dence intervals. Conclusion: The current version of the LI-RADS is not more accurate than the AASLD score for the non-invasive diagnosis of small HCC in high-risk patients but it provides important information on the probability of having HCC in highrisk patients allowing possible changes in management in these patients.

SS 2.8
Differentiating hypervascular hyperplastic nodule from hypervascular hepatocellular carcinoma in patients with alcoholic liver cirrhosis: value of gadoxetic acidenhanced and diffusion-weighted MR imaging S.S. Kim 1 , S.H. Kim 2 , K.D. Song 2 ; 1 Cheonan-si, Chungcheongnam-do/ KR, 2 Seoul/ KR Purpose: To evaluate the value of gadoxetic acid-enhanced and diffusionweighted (DW) magnetic resonance (MR) imaging for differentiating hypervascular hyperplastic nodule (HHN) from hypervascular hepatocellular carcinoma (HCC) in patients with alcoholic liver cirrhosis (LC). Material and methods: Among 310 patients with alcoholic LC who underwent gadoxetic acid-enhanced and DW MR imaging, 11 patients with 65 HHNs and 23 patients with 32 hypervascular HCCs were included. Qualitative and quantitative analyses were conducted. Signi cant MR imaging ndings on univariate and multivariate analyses were assessed and their diagnostic performances for predicting HHN were analyzed. Results: In univariate and multivariate analysis, lesion size (≤14 mm) (odds ratio [OR] = 145.65) and high SI on unenhanced T1-weighted image (T1WI) (OR = 59.18) were independently signi cant imaging ndings for predicting HHN (P < 0.001 and P = 0.001, respectively). The speci city of their combination for predicting HHNs was 100%. In quantitative analysis, mean SI on HBP image and lesion-to-liver SI ratio on HBP of HHNs were signi cantly higher than those of hypervascular HCCs (P < 0.001, respectively). Mean ADC values and lesion-to-liver ADC ratios between HHNs and HCCs were not signi cantly different (P = 0.163 and P = 0.531, respectively). Conclusion: Gadoxetic acid-enhanced and DW MR imaging may be helpful for differentiating HHN from hypervascular HCC in patients with alcoholic LC and lesion size ≤ 14mm and high SI on unenhanced-T1WI were signi cant imaging ndings for predicting HHN.
Material and methods: This study had institutional review board approval; the requirement for informed consent was waived. One hundred and thirty-seven patients with 143 lesions (45 HCAs and 98 HCCs) were included in the study. There were 37 female patients and 100 male patients (mean age, 58 years). Two radiologists evaluated morphologic features, signal intensity (SI) of tumors on MR images including DW (b=800) imaging and dynamic enhancement pattern in consensus. For quantitative analysis, tumor-to-liver SI ratio and contrast enhancement index (CEI) on unenhanced, dynamic, and hepatobiliary phase images and apparent diffusion coef cient (ADC) maps were calculated. Statistically signi cant imaging ndings were identi ed through univariate and multivariate analyses, and their diagnostic performance for predicting HCA was analyzed.
Results: In univariate and multivariate analyses, high SI on portal phase images (p=0.0009), lower tumor-to-liver SI ratio on T2-weighted images (<1.647, p=0.0009), and higher tumor-to-liver SI ratio on T1-weighted images (≥0.807, p=0.0068) and tumor-to-liver SI ratio on ADC maps (≥0.841, p<0.001) were independently signi cant factors for predicting HCA. When three of these four criteria were combined, 68.9% (31/45) of HCA were identi ed with specicity of 92.9%. When all four criteria were satis ed, speci city was 100%. Conclusion: Gadoxetic acid-enhanced and DW MR imaging may be helpful for differentiating HCA from HCC.

SS 2.10
Comprasion of intravoxel incoherent motion and conventional-diffusion weighted imaging parameters for hepatacellular carcinoma grading at explant B.K. Sokmen, A. Oz, S. Server, M. Dayangac, N. Inan, G. Bulbul Dogusoy, C. Balci; Istanbul/ TR Purpose: To compare the diagnostic value of IVIM and conventional DWI parameters for estimation the HCC grading in explant livers. Material and methods: Twenty-three patients with histopathologically diagnosed HCC at explant were included in this retrospective study. All patients were examined by 1.5T MRI within 90 days before liver transplant. IVIM (16 different b factors of 0-1300 s/mm 2 ) and conventional DWI were obtained. Two different ADC maps reconstructed from conventional DWI (ADC con ) and IVIM (ADC ivim ). The mean D, D* and f values also calculated from IVIM. Histopathologically, HCC was classi ed as low (grade 1, 2) and high (grade 3, 4) grade. Quantitatively, ADC con , ADC ivim , D, D* and f values were compared by student's t test. The relationship between the parameters and grade was analyzed by Spearman correlation test. To evaluate the diagnostic performance of the parameters, ROC analysis was performed. Results: The ADC con and ADC ivim values of high-grade HCCs were signi cantly lower than those of low grade ones (p<0.005, respectively). Both ADC con and ADC ivim values were inversely correlated with grade (r=-519, p=0.011; r=-421, p=0,046, respectively). The f value of the high-grade HCC was signi cantly higher than low grade (p=0.005). The f values were positive correlated with grade (r=0.548, p=0.007). The best discriminative parameter was the f value (mean f value was 26% for low grade, 38% for high grade). MRI+DWI+endoscopy were re-evaluated by an expert radiologist and expert surgeon, who scored a con dence level for likelihood of CR and residual N+ disease and described the morphology/aspect of MRI & endoscopy after CRT. Updated selection criteria for CR were used. Two groups were formed based on the re-evaluation of images: (1) residual tumour at reassessment (non-CR) and (2) (near) CR. Results: 33 patients had ypT0N0, the remaining 9 had ypT0N+. Main reasons for missing CR were heterogeneous T2-signal, residual focal/massive diffusion-restriction, ycN+ disease and mucosal abnormalities at endoscopy. These imaging factors were signi cantly more frequently found in patients who were still deemed to have residual tumour compared to those who were deemed a CR upon reassessment. 8 patients were overstaged as ycN+, of which 7 had irregular nodes. Also, nodal irregularity was found in 6/9 patients with ypN+ disease. Normalized nodes all were ypN0. Conclusion: Missing a CR after CRT is usually due to heterogeneous T2-signal, extensive brosis, residual diffusion-restriction, ycN+ disease and residual mucosal abnormalities. Nodes are overstaged in approximately half of the cases. Awareness of these pitfalls can help in the selection of patients for a watch-and-wait strategy after CRT for rectal cancer.

SS 3.2
Rectal cancer: comparison of MR-TRG, volume ratio and signal intensity decrease for the identification of complete responders after radiochemotherapy S. Picchia 1 , M. Rengo 2 , D. Caruso 1 , D. Bellini 1 , D. De Santis 1 , A. Laghi 1 ; 1 Latina/ IT, 2 Rome/ IT Purpose: To compare three methods to identify complete responders after chemoradiotherapy (CRT) in a population of patients with locally advanced rectal cancer. Material and methods: 65 patients, diagnosed with locally advanced rectal cancer were prospectively enrolled in the study. All patients underwent MRI on a 3-Tesla before, during and after CRT. All patients underwent total mesorectal excision (TME). MR-TRG, volume reduction ratio (VR) and signal intensity percentage decrease (SI) were compared. Measurements were performed on the entire tumour volume using a dedicated software. MR-TRG, VR and SI were compared with histology. DFS, OS and ROC analysis were evaluated and compared among the three groups. Patients were strati ed according to the histology result in complete responders (CR) and partial or non-responders (PNR  (60), (AUC of time to peak (TTP), maximum enhancement, bolus arrival time (BAT), initial signal excess (ISE), mean transit time (MTT), wash-in&washout parameters) were compared between patients with complete (CR;ypT0) or good response (GR;ypT0-2) and non-responders. Additionally, heterogeneity of DCE parameters (measured by coef cient of variance (CoV:sd/mean)) was compared.
Results: 8/20 (40%) had CR. Almost all DCE-parameters were lower in CR&GR (p>0.05). MTT and BAT were higher in CR, with BAT signi cantly higher (p=0.015). In GR the BAT was also signi cantly higher (p=0.028). Heterogeneity was higher in CR for almost all DCE-parameters, except for BAT and washout, which were lower in CR (p>0.05). In GR AUC60 and TTP were signicantly more heterogeneous than in poor responders (p=0.024-0.041), while BAT was signi cantly less heterogeneous in GR. Conclusion: Bolus arrival time is longer and less heterogeneous in complete & good responders to CRT, potentially re ecting slower but more homogeneous ow. This could represent longer exposure to chemotherapy and reduced hypoxia and thus higher response. Most DCE parameters were lower in responders re ecting lower angiogenetic activity.

SS 3.4
Experience of the MRI tumour-regression-grading (TRG) system for anal cancer: interobserver agreement and impact of diffusion-weighted imaging D. Prezzi 1 , S. Gourtsoyianni 1 , K. Owczarczyk 1 , A. Gaya 1 , M. Leslie 1 , A. Qureshi 1 , R. Glynne-Jones 2 , V. Goh 1 ; 1 London/ GB, 2 Northwood/ GB Purpose: To evaluate independently a recently proposed MRI TRG system for anal cancer post chemoradiation. To assess interobserver agreement based on multiplanar high-resolution T2-weighted (T2-w) imaging. To measure the impact of additional diffusion-weighted imaging (DWI) on TRG scores and interobserver agreement. Material and methods: Forty-three patients with biopsy-proven anal squamous-cell-carcinoma (ASCC) underwent staging pelvic MRI at baseline and 3-6 months after chemoradiation, including multiplanar high-resolution T2weighted sequences and DWI (b=0;800 s/mm 2 ). Two radiologists scored posttreatment scans independently in two separate sessions: rst using high-resolution T2-w sequences; a month later, combining T2-w with DWI. Upon statistical advice, interobserver agreement weighted-Kappa was calculated. Results: No cases were assigned an initial TRG score of '1' or '5'. TRG scores of '2' (excellent response with brosis) or '3' (indeterminate response with heterogeneous signal at the tumour site) accounted for most cases (39 cases/43) and determined substantial interobserver discordance based on T2-w sequences alone (15 cases/39). Interobserver agreement improved from 'fair ' [Kappa=0.36] to 'moderate' [Kappa=0.55] with the inclusion of DWI; 6 cases of discordant-indeterminate response (TRG '2/3') resolved in favour of excellent response (TRG '2'). Conclusion: A 5-category TRG system may be redundant in ASCC. Interobserver agreement in this independent cohort is lower than previously reported, possibly due to misinterpretation of category descriptors. DWI appears valuable for staging ASCC post-chemoradiation.

S597
SS 3.5 MR T2 mapping imaging for quantitatively evaluating tumor response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer F. Li, Z. Zhou, W. Cao, Z. Li, J. Zhou, R. Malla, A. Mohamoud, X. Wang, J. Gong; Guangzhou/ CN Purpose: To explore the potential value of MR T2 mapping imaging in evaluating the response to neoadjuvant chemoradiotherapy (CRT) in locally advanced rectal cancer. Material and methods: 36 patients with locally advanced rectal cancer underwent MRI before and after CRT. MRI protocols included axial FSE T2 weighted imaging and T2 mapping imaging. The whole tumor tissue on T2 maps was included by manually drawing the region of interests (ROIs). T2 values before and after CRT were measured. T2 changes (∆T2), T2 reduction rate before and after CRT were calculated, respectively. Patients were divided into good response (GR) group and poor response (PR) group according to the tumor regression grade. Parameters were compared by using independent-samples T test. Receiver operating characteristic curve (ROC) and the corresponding areas under the ROC curve (AUC), accuracy, sensitivity, speci city, positive predictive value (PPV) and negative predictive value (NPV) were computed to assess the diagnostic performance of ∆T2 and T2 reduction rate. Results: 26 patients were classi ed as GR and 10 patients had a PR. ∆T2 and T2 reduction rates were signi cantly different (P<0.05) between GR and PR group. ∆T2 and T2 reduction rate had a AUC of 0.95 and 0.96, PPV of both 93% for the GR, respectively, and with a NPV of both ∆T2 and T2 reduction rate were 89% for the GR, respectively. Conclusion: High ∆T2 and T2 reduction rate correspond to pathological good response.

SS 3.6
DCE-MRI has a value in predicting and assessing response to chemoradiation for rectal cancer R.A.P. Dijkhoff, R.G.H. Beets-Tan, D.M.J. Lambregts, G.L. Beets, M. Maas; Amsterdam/ NL Purpose: To perform a systematic review evaluating the clinical value of DCE-MRI in rectal cancer. Material and methods: A search was performed on Pubmed, Embase and the Cochrane library. Studies on DCE-MRI for tumour aggressiveness and primary (re)staging after chemoradiation (CRT) were included. Information on population, DCE-technique, DCE-parameters and outcome were extracted. Results: 19 studies were identi ed; 10 evaluated quantitative analyses, 6 semiquantitative analyses and 3 evaluated both. 8 studies evaluated correlation between DCE-parameters and angiogenesis or tumour aggressiveness. 11 evaluated response prediction pre-and post-CRT. Several semiquantitative parameters showed a signi cantly positive correlation with angiogenesis (mostly wash-in parameters), for quantitative analyses con icting results were found. Con icting results were also reported for the correlation between DCEparameters and tumour aggressiveness. Both higher and lower vascularity in more aggressive tumours are reported, while some studies report no correlation. Six studies showed a predictive value of Ktrans for response. Higher Ktrans pre-CRT was signi cantly correlated with a complete/good response, but reported pre-CRT Ktrans varied substantially (0.36-1.93). Post-CRT reduction in Ktrans of 32%-36% was signi cantly associated with response. For semiquantitative analyses pre-CRT late-slope was reported to be signi cantly lower in good responders, however, only few studies exist on semiquantitative analyses of post-CRT DCE-MRI. Conclusion: DCE-MRI in rectal cancer is promising for response prediction and assessment of CRT, where a high pre-CRT Ktrans and a decrease in Ktrans are signi cantly predictive for response. Purpose: To perform fully automated segmentation of locally advanced rectal carcinomas (LARCs) via a supervised learning technique (SLT) trained and validated using delineations by expert-readers. Material and methods: We selected the MRI scans (1.5T, T2-weighted and b1000 DWI) of 140 LARC-patients. Manual whole tumor volume delineation, performed by an expert radiologist (on b1000 DWI, was used as the standard reference. From each image, 500 pixels within the tumor and 500 pixels outside the tumourr were randomly sampled. For each pixel, we extracted the surrounding patch of 25x25 in all MRI sequences, resulting in 1000 patches per patient. By tting the resulting three patches as color channels, a multiparametric representation of the image was obtained. The architecture consisted of three convolutional layers. A fully connected layer of 256 neurons was placed between the convolutional layers and the output layer. Both classes were balanced due to randomized sampling procedure. Patients were randomly and equally divided between training and testing. Results: The algorithm was trained until it reached a stable condition. The trained algorithm was then applied on the test set, reaching an AUC of 0.94 based on the probability of being tumor for each pixel. Conclusion: Automatic segmentation using an SLT is able to reproduce manual expert segmentations with an AUC of >0.90, suggesting that it can be a time-ef cient solution to help delineation in daily practice. SCIENTIFIC SESSIONS / WEDNESDAY, JUNE 21, 2017 S598 SS 3.9 Magnetic resonance imaging features of rectal signet ring cell carcinoma Z. Zhou, W. Cao, J. Gong, X. Wang, X. Meng, J. Zhou; Guangzhou/ CN Purpose: To investigate the magnetic resonance (MR) characteristics of primary rectal signet ring cell carcinoma (SRCC), compared with conventional rectal adenocarcinoma (AC). Material and methods: 28 cases of rectal SRCC and 31 cases of AC conrmed by pathology with primary rectal MR data were retrospectively analyzed. Compare SRCC with AC about the TNM stage, the basic imaging characteristics and tumor signal from T2WI sequence. Results: The distant metastasis rate of SRCC was 25% (7/28), 10.7% (3/28) in peritoneal metastasis. The ratio of SRCC circumferential in ltration (>1/2 of the periphery) was higher than that of AC, 92.9% (26/28) vs. 61.3% (19/31), p=0.004. The tumor/fat, tumor/muscle and tumor/urine signal ratios of SRCC from T2WI sequence were higher than that of AC. There were no signi cant differences in the transverse maximum thickness, longitudinal diameter, and the distance the lower edge of the tumor from anal edge between SRCC and AC. In terms of the distance tumor breakthrough from the intrinsic muscle layer of the rectum, AC was higher than SRCC (6.03 mm vs. 4.03 mm, P=0.044). Conclusion: The rectum invasion circumferential diameter range of SRCC was larger than rectal conventional adenocarcinoma, but the breakthrough of muscular layer distance was relatively limited. The MR T2WI signal of rectal SRCC was often higher than that of conventional adenocarcinoma, which may provide a certain prediction basis for rectal cancer multidisciplinary therapy.

Maastricht/ NL
Purpose: Diffusion-weighted imaging (DWI) using single-shot echo planar imaging (EPI) is increasingly included in standard rectal MRI-protocols. EPI-DWI is prone to susceptibility artefacts, mainly caused by air in the rectal lumen. Aim was to assess whether application of a micro-enema can reduce these artefacts. Material and methods: 50 patients were included who each underwent multiple sequential DWI-MRIs (1.5T; highest b-value b1000) during follow-up for a wait-and-see approach after chemoradiotherapy. Until March 2014 DWI-MRIs were acquired without bowel preparation, thereafter a micro-enema (Microlax ® ;5 ml) was routinely applied shortly prior to acquisition. Two readers scored the presence/severity of air artefacts per scan, ranging from 0 (no artefact) to 5 (severe artefact). A score >= 3 (moderate-severe) was considered a clinically signi cant artefact. Scores were compared between DWI-scans with/without a micro-enema. Potential confounding factors (age/gender, acquisition parameters, MRI-hardware, endoscopy prior to MRI) were taken into account. Results: In total 335 DWI-MRIs were assessed. Signi cant air artefacts were seen in 24.3% (no micro-enema) vs. 3.7% (with micro-enema). Using binary logistic regression with samples clustered by patient, the odds ratio between the use of a micro-enema and presence of signi cant artifacts was 0.12 (95% CI 0.04-0.40), p = 0.0005. None of the assessed potential confounders signicantly altered this effect. Conclusion: The use of a micro-enema prior to rectal EPI-DWI examinations signi cantly reduces air artefacts, compared to examinations without bowel preparation.

SS 4.4
Skin-liver distance and interquartile-median ratio as determinants of inter-operator concordance in acoustic radiation force impulse (ARFI) imaging S. Su, W. Wang, D. Nadebaum, A. Nicoll, S. Sood, A. Gorelik, J. Lai, R. Gibson; Melbourne, VIC/ AU Purpose: The accuracy of ARFI ultrasound compared to liver biopsy is higher when there is concordance between F-scores of two or more operators. We hypothesised that when the rst operator interquartile range/median-velocity ratio (IMR) is <0.3 and skin-liver distance (SLD) is ≤2.5cm, there is greater interoperator concordance and a second operator is not necessary. Material and methods: Two-operator ARFI ultrasound (Siemens S2000) measurements (F-score, SLD and IMR) were recorded for 927 consecutive patients. Concordance was de ned as F-scores in same or adjacent stages. Chisquared testing (SPSSv17) compared concordance for SLD≤2.5 cm versus SLD>2.5 cm; and IMR <0.3 versus IMR≥0.3 when SLD≤2.5cm, in each of the F-score groups of 0/1, 2, 3 and 4. Results: Statistically signi cant differences were demonstrated between SLD≤2.5cm and SLD>2.5cm groups when F-score was 0/1 or 4 (p=0.005), and when F-score was 2 or 3 (p<0.0005). Concordance, when SLD≤2.5cm, was more than 85% for all F-score groups. Chi-squared tests comparing higher skin-liver distances were not performed as correlation with biopsy reduces when SLD>2.5cm. In the SLD≤2.5cm group, concordance fell below 85% when IMR≥0.

São Paulo/ BR
Purpose: Fat distribution may have prognostic value in the evaluation of nonalcoholic fatty liver disease. This study was conducted to evaluate associations of MRI-measured abdominal fat areas with steatosis, steatohepatitis, and brosis, assessed histopathologically, in patients with type 2 diabetes. Material and methods: This prospective study included 66 patients with type 2 diabetes (12 males, 54 females, age 26-68 years), without chronic liver disease of other causes. Axial dual-echo MR images were acquired. Visceral, subcutaneous, and preperitoneal fat areas were measured using Osirix software. Liver biopsy specimens were obtained from all patients and examined histopathologically to evaluate steatosis, steatohepatitis, and brosis. Linear (for steatosis) and logistic (for steatohepatitis and brosis) regression models were tted for the outcomes. R2 was used as a measure of how much model variance the predictors explained, and to compare different predictors of the same outcome. Results: Visceral and preperitoneal fat areas correlated well with histopathologically determined liver steatosis grade (both p = 0.004) and liver brosis (p = 0.008 and p = 0.037, respectively). All fat areas correlated well with steatohepatitis (p ≤ 0.002). Preperitoneal and visceral fat areas were the best predictors of steatohepatitis (R2 = 0.379) and brosis (R2 = 0.181), respectively. Conclusion: Preperitoneal fat area was the best predictor of steatohepatitis and is a potential new noninvasive marker for use in the screening of these patients to detect more aggressive forms of NAFLD. Purpose: To evaluate the current liver ultrasound ndings in patients with Wilson's disease (WD) in correlation with clinical-laboratory data. Material and methods: In a case series we have analysed 11 diagnosed cases of WD patients, evaluating for multiparametric ultrasound (US) ndings: baseline US, point shear wave elastography (pSWE) and contrast-enhanced ultrasonography (CEUS) of liver in correlation with clinical and laboratory ndings.
Results: There were genetically proved 11 patients, 8 of them had non-differentiated liver disease in history. Baseline ultrasound revealed diffuse parenchymal changes with increased echogenicity (n=7), coarse-grained texture (n=6), multiple hypoechogenic nodules (n=3), liver contour irregularity suggesting cirrhosis (n=4) and perihepatic fat thickening (n=3). Non-speci c contrast enhancement (n=3) and nodular lesions with late washout (n=2) was proved by CEUS. Fibrosis was detected by pSWE measurements (F1 (n=2), F4 (n=1)). Decreased plasmatic ceruloplasmine (n=5) levels were associated with nodular liver changes on US and Kayser-Fleischer rings (n=3).All patients had elevated copper level in 24h urine sample, however liver functional analysis did not correlate with structural changes. Conclusion: There are some speci c ndings: hypoechoic nodules and perihepatic fat thickening in liver US and nodularity on CEUS with late wash-out pattern with variable brosis grade on elastography, that can indicate possible Wilson's disease. Multiparametric ultrasound can be used as rst modality to assess liver involvement as well as during progression of the Wilson's disease.

SS 4.8
The usefulness of native T1 of the liver obtained by SMART1Map (saturation method using adaptive recovery times for cardiac T1 mapping) in the evaluation of the severity of the chronic liver diseases T. Nonaka 1 , A. Yamada 1 , Y. Kitou 1 , Y. Iwadate 2 , A. Nozaki 2 , G.S. Slavin 3 , Y. Fujinaga 1 , M. Kadoya 1 ; 1 Matsumoto/ JP, 2 Hino/ JP, 3 Milwaukee, WI/ US Purpose: SMART1Map is a novel method for true T1 relaxometry using saturation recovery sequence. The purpose of this study was to clarify the usefulness of native T1 (nT1) of the liver obtained by SMART1Map comparing to other MR imaging biomarkers in the evaluation of the severity of the chronic liver diseases according to the liver stiffness (LS) measured by MR elastography (MRE). Material and methods: We evaluated consecutive 41 patients who underwent magnetic MR imaging including MRE, SMART1Map, IDEAL-IQ, diffusion weighted imaging (DWI) using 1.5-T MR scanner for the evaluation of chronic liver diseases from January to December 2016 in our hospital. The several MR imaging biomarkers of the liver, such as LS, nT1, fat fraction (FF), R2*, and apparent diffusion coef cient (ADC) were obtained. The correlation between LS and the other imaging biomarkers were statistically analyzed including stepwise linear regression analysis on LS using the other imaging biomarkers as explanatory factors. Results: Correlation coef cient and its P value between LS and the other imaging biomarkers were as follows: nT1, r=0.7156 (P<0.01); FF, r=0.1996 (P=0.2363); R2*, r=-0.2955 (P=0.0717); ADC, r=0.2995 (P=0.0605). Stepwise linear regression analysis revealed that nT1 was statistically signi cant explanatory factor for LS, and that the other factors were not signi cant. Conclusion: nT1 of the liver obtained by SMART1Map is signi cant predictor of the LS independent from fat deposition and iron overload of the liver.

SS 4.9
The negative hepatic arterial buffer response: pre and post-prandial changes in total liver and hepatic arterial blood flow measured using caval subtraction phase contrast MRI in normal volunteers M. Chouhan, A. Bainbridge, M. Lythgoe, R. Mookerjee, S. Halligan, S.A. Taylor; London/ GB Purpose: Non-invasive measurements of total liver blood ow (TLBF) and hepatic arterial ow (HAF) would be valuable in the assessment of portal hypertension. The purpose of this study was to study changes in TLBF and HAF following prandial stress, known to increase portal venous ow (PVF). Material and methods: Fasted healthy volunteers (n=13), underwent PV, proximal and distal IVC breath-hold,cardiac-gated 2D cine-phase-contrast MRI(PCMRI, 5 mm slice thickness, =10˚, 256x256(FExPE)) at 3T, with velocity encoding settings of 40,60 and 80cm/s. TLBF was estimated by subtracting proximal IVC ow (above renal but below hepatic venous inlets) from distal IVC ow (above hepatic venous inlets,but below the IVC-right atrial junction). HAF was estimated by subtracting PV ow from estimated TLBF. Subjects then ingested a 700 calorie nutrient milkshake with a second scan 45-60 minutes later. Data was analysed using paired Student t-tests. Results: As expected, a signi cant rise in post-prandial PVF was observed (56.9±3.6 vs 109.1±4.6 ml/min/100g; p<0.0001). A signi cant rise in postprandial estimated TLBF was also observed (72.0±4.8 vs 117.4±5.0 ml/ min/100g; p<0.0001). Reductions in post-prandial estimated HAF (12.0±4.8 vs 3.8±3.4 ml/min/100g; p=0.0592) were non-signi cant, but reductions in estimated HA fraction (20.2±2.5 vs 8.2±1.3%; p=0.0003) were signi cant. Conclusion: Caval subtraction PCMRI can be used to measure changes in TLBF and HAF after haemodynamic stress. A negative relative hepatic arterial buffer response (reduction in estimated HA fraction in response to the rise in PVF) was observed following prandial stress.

SS 4.10
Non-invasive evaluation of clinically significant portal hypertension: validation of two novel algorithms combining liver and spleen stiffness using shear-wave elastography in a large independent cohort of patients with cirrhosis M. Ronot, L. Elkrief, M. Dioguardi Burgio, M. Zappa, L. Castera, V. Vilgrain, P.-E. Rautou; Clichy/ FR Purpose: To externaly validate the accuracy of two recent algorithms using sequential measurements of liver (L-SWE) and spleen stiffness (S-SWE) for ruling-out or ruling-in clinically signi cant portal hypertension (CSPH) in a large independent cohort of patients with cirrhosis. Material and methods: 163 consecutive patients with stable cirrhosis (median 55 years; 76% male; Child-Pugh A 38%, B 26% and C 34%) who underwent L-SWE and S-SWE at the time of HVPG measurement were studied. The diagnostic performances of the 2 algorithms were assessed using sensitivity (Se), speci city (Sp), positive predictive value (PPV) and negative predictive value (NPV), as well as the proportion of correctly classi ed patients. Results: The median HVPG was 16 mmHg, and 80% of the patients had CSPH. The rst published algorithm, using L-SWE<16.0kPa and then S-SWE<26.6kPa could ruled-out CSPH with a NPV of 96.3%. Diagnostic performance was lower in our population with a NPV of 60%. The second published algorithm, using L-SWE>38.0kPa, or L-SWE≤38.0kPa but S-SWE>27.9kPa, could ruled-in CSPH with a PPV of 95.7%. Again diagnostic performance in our population was lower with a PPV of 87%. The diagnostic performances of these algorithms did not improve when restricting the analyses to the subset of 68 patients without previous decompensation of cirrhosis. Conclusion: Diagnostic accuracies of algorithms based on sequential measurements of liver and spleen stiffness using SWE are acceptable, but not good enough to replace HVPG measurement or to base clinical decisions. Material and methods: Thirty-six patients with focal liver lesions were prospectively enrolled. For biopsy planning, real-time imaging fusion of CT/MRI with USG (hereafter USG-Fusion) was performed and subsequently real-time SonoVue-enhanced USG was fused with CT/MRI (hereafter CEUS-Fusion) in all patients. Biopsy operator was evaluating lesion visibility, necrotic degree, con dence level of technical success before procedure (4-point scale), and safety root accessibility on conventional USG (step 1), USG-Fusion (step 2), and CEUS-Fusion (step 3). Occurrence of change in biopsy target also assessed.
Results: Among 36 target lesions, 8 (22.2%) lesions were invisible on Step 1 and 2. After applying the CEUS fusion, 7 of 8 (87.5%) focal hepatic lesions were visualized. Con dence level of technical success before procedure is signi cantly increased on CEUS-Fusion compared Step 1 (p=0.029) or 2 (p=0.002). Presumed target lesion were changed in 16 of 36 (44.4%) patients after CEUS-Fusion, and as the lesion is more necrotic, the presumed target lesions were more frequently changed (61.2%, 13/21 necrotic mass; 20%, 3/15 non-necrotic mass). Con rmative diagnostic results were reported in 35 patients (97.2%, 33 malignant and 2 benign). Accessibility of safety root to target lesion did not reach statistical differences. Conclusion: Applying real-time CEUS fusion with CT/MRI improved tumor visibility and viable portion assessment, and leading to higher operator condence and diagnostic yield, comparing conventional USG and real-time CT/ MRI fusion with USG.

SS 5.2
The impact of a liver specific multidisciplinary assessment in patients with colorectal cancer liver metastases: a population-based study J. Engstrand 1 , N. Kartalis 1 , C. Strömberg 1 , M. Broberg 1 , A. Stillström 1 , T. Lekberg 1 , E. Jonas 2 , H. Nilsson 1 , J. Freedman 1 ; 1 Stockholm/ SE, 2 Cape Town/ ZA Purpose: To evaluate the potentially improved resection rate in a de ned cohort if all patients with colorectal cancer (CRC) liver metastases (LM) were evaluated at a dedicated liver multidisciplinary tumour-board (MDT). Material and methods: A retrospective analysis of 272 patients diagnosed with LM within a 5-year follow-up period after CRC-diagnosis in the greater Stockholm region was conducted. All patients with LM were re-evaluated at a ctive liver-MDT where previous imaging studies, tumour characteristics, medical history and patients' treatment preferences were presented. Treatment decisions for each patient were compared to the decisions at the original MDT. Odds ratios (ORs) and 95% con dence intervals were estimated (logistic regression) for factors associated with referral to the liver-MDT. Results: Out of 272, 102 patients were discussed at an original MDT and 69 patients were eventually resected. At the ctive liver-MDT, a further 22 patients were considered as resectable/potentially resectable, none previously assessed at liver-MDT. At the original MDT, 47 (18%) patients had inadequate imaging; in additional 27 (10%) patients the available imaging was inadequate but other patient-related factors permitted decision-making on treatment strategy. Factors in uencing referral to liver-MDT were age (OR 3.12, 1.72 to 5.65), American Society of Anesthesiologists (ASA) score (OR 0.12, 0.01 to 1.17, ASA 2 versus ASA 4) and number of LM (OR 0.10, 0.04 to 0.22, 1-5 LM versus >10 LM), while gender (p=0.194) and treatment at a teaching hospital (p=0.838) were not. Conclusion: A substantial number of patients with CRC liver metastases are not managed according to best available evidence and the potential for higher resection rates is high.

SS 5.3
Can CT texture features predict imminent development of hepatic colorectal metastatic disease? S.J. Lee, M. Lubner, D. Kim, P. Pickhardt; Madison, WI/ US Purpose: Preclinical studies have demonstrated CT textural changes in liver enhancement related to micrometastases, and early human clinical studies have suggested that these changes may be associated with the eventual development of hepatic metastases. The purpose of this study was to determine if hepatic textural changes are present at CT in patients immediately prior to the development of grossly detectable colorectal liver metastases, compared to those without liver metastatic spread. Material and methods: This case-control study was comprised of 89 colorectal cancer patients (29 cases who developed liver metastases and 60 controls who did not). Liver texture features were assessed on portal-venous phase CT using a software tool (TexRAD) at both pre-treatment staging and interval surveillance scans immediately prior to metastatic detection. Matched interval scans were also assessed in controls. Multivariate Cox proportional hazards models were tted with time to occurrence of liver metastases as the primary outcome. Results: After adjusting for multiple hypothesis testing, none of the features were different between cases and controls. Both models of staging and interval CT texture features failed to reach statistical signi cance (p = 0.06, p = 0.13 by likelihood ratio testing). Staging and interval entropy values were also not associated with hepatic metastatic disease (p = 0.61, p = 0.62). Conclusion: We found no evidence that liver parenchymal texture features from contrast-enhanced staging or interval pre-detection CT scans could predict the subsequent development of hepatic metastasis in colorectal cancer patients.

SS 5.4
Intravoxel incoherent motion (IVIM) of colon cancer liver metastases for the response assessment of antiangiogenic treatment: preliminary results A. Oz, E. Namal, S. Server, B. Koyuncu Sokmen, S.B. Barlas, N. Inan, C. Balci; Istanbul/ TR Purpose: To evaluate the time related intra-voxel incoherent motion (IVIM) parameter changes of colon cancer liver metastases during anti-angiogenic therapy. Material and methods: Eight-patients with 16 liver metastases were included in this prospective study. All patients were treated with anti-angiogenic agent Bevacizumab in combination with FOLFOX-FOLFIRI. All patients were examined pretreatment (PT) as well as at the 3 rd , 6 th , and 9 th months of therapy. MRI was performed with a 1.5T scanner. In addition to routine abdominal MRI, an IVIM-DWI sequence was obtained using a free breath single-shot echo planar spin-echo (EPI) sequence with 17 different b factors (0-1400s/mm 2 ). The mean D (true diffusion coef cent), D* (pseudo-diffusion coef cient associated with blood ow), and f (perfusion fraction) values of each metastasis were calculated for each time point and time related changes were recorded. Results: Compared to the value of PT; both f and D* values of the metastases were signi cantly decreased at the time point 6 months after the initiation of anti-angiogenic therapy (p<0.001). The time-related changes of the mean f value were as follows: 0.2035, 0.2552, 0.3024, and 0.2492 at PT, 3 rd , 6 th , and 9 th months; of the mean D* values were: 84.486x10 -3 , 197.911x10 -3 , 188.241x10 -3 , 121.853x10 -3 at PT, 3 rd , 6 th , and 9 th months, respectively.
Conclusion: The f value of IVIM may quantitatively re ect the response of antiangiogenic therapy. The effect of anti-angiogenic response becomes evident after 9 months of therapy. Purpose: The purpose of our study was to correlate in benign hepatocellular tumors the quantitative analysis of HBP contrast agent uptake to the quantitative level of OATP expression. Material and methods: This single-center retrospective study included between September 2009 to March 2015, 20 consecutive patients with a proven pathological and immunohistochemical (IHC) diagnosis of FNH or liver adenoma and underwent Gd-BOPTA-enhancement MRI, including the realisation of HBP. The analysis of HBP utpake was performed using the Liver to Lesion Contrast Enhancement Ratio (LLCER). The mean LLCER and OATP expression were calculated and compared between FNH and HCA (Mann-Whitney) and expression of OATP were correlate with LLCER value (Spearman). Results: Of the 23 hepatocellular benign tumors 9 (39%) were FNH and 14 (61%) were HCA. On HBP 100% of the FNH appeared either hyperintense or isointense and 79% of the adenomas appeared hypointense. The mean OATP expression of FNH (46,67+/-26,58%) was signi cantly higher than that of HCA (22,14+/-30,74%) (p=0,0273) and the mean LLCER of FNH (10,66 +/-7,403%) was signi cantly higher than that of HCA (-13,5 +/-12,25%) (p<0,0001). A signi cant correlation was found between OATP expression and LLCER values in all patients (r=0,661; p =0,0006). Conclusion: The quantitative analysis of HBP lesion uptake is correlated to quantitative OATP expression in benign hepatocellular tumors, and can help for the differential diagnosis of FNH, -catenin-activated HCA and other HCA. To determine the performance of diffusion-weighted imaging magnetic resonance enterography (DWI-MRE) compared to contrast material-enhanced MR enterography (CE-MRE) for evaluating bowel in ammation activity in Crohn's disease (CD). Material and methods: We retrospectively review 55 consecutive adults patients with CD that underwent clinical assessment, MR enterography, and ileocolonoscopy within 4 weeks. Inclusion criteria were: a) DWI-MRE with the following b values (50, 400 and 800 s/mm 2 ) and CE-MRE and b) histologically examination after colonoscopy. The nal study population included 39 patients (19 M, 20 F, mean age 42 years). The results were compared with the histopathological examination after colonoscopy as the reference standard. The sensitivity (se), speci city (sp), positive predicting value (PPV), negative predicting value (NPV) and accuracy (acc) of DWI and CE-MR enterography in distinguishing active in ammatory lesions from inactive lesions were determined.

SS 6.2
Comparison of contrast-enhanced and diffusionweighted MRI in assessment of the terminal ileum in Crohn's disease patients  To determine the diagnostic value of dynamic contrast-enhanced perfusion-MRI in detection and characterization between active small bowel in ammation and mural brosis in patients with Crohn's disease (CD). Material and methods: We analyzed a total of 37 patients (11 women; 23-69 years) with known biopsy proven CD, who underwent MR-enterography study, performed on a 1.5T MRI system (Achieva, Philips), using a phased array sense body multi-coil, after oral administration of 1,5-2 of PEG solution. MRE protocol included T1 weighted, sSShT2, sBTFE and gadolinium-enhanced THRIVE sequences acquired on coronal and axial planes. A dedicated workstation was used to generate colour permeability maps, and after placing the Region of interest (ROI) on the bowel site involved by CD localization, the following parameters were calculated and statistically analyzed: relative arterial, venous and late enhancement (RAE, RVE, RLE), maximum enhancement (ME) and time to peak (TTP Purpose: To investigate the ef cacy of magnetic resonance enterography (MRE) in patients with endoscopically diagnosed terminal ileitis (TI) and suspected to be Crohn's disease (CD). Material and methods: Retrospective review of a prospective database of patients referred to our institute with suspected Crohn's disease. Patients were included if they underwent MRE and colonoscopy for suspected CD. All MRE were performed on a 1.5T scanner (Magnetom Avanto, Siemens Healthcare, Erlangen, Germany) with the same protocol. Images were independently reviewed by two board-certi ed radiologists blinded of clinical information and the results were compared to endoscopic and pathology ndings. Results: Seventy-two patients (median age 37, 53M:19F) were included with all but one one scan deemed adequate for analysis. Inter-observer agreement for abnormality on MRE was high (keppa=0.91) and pathology was identi ed on 43 scans of which 39 (90.7%) involved the terminal ileum and 26 (60.4%) had multi-segment disease. Twenty-eight patients had histologically conrmed Crohn's disease, 6 TB enteritis, 2 intestinal Behcet's disease and 1 oesinophilic gastroenteritis. MRE has sensitivity of 85.7% and the presence of skip lesions has 100% speci city for CD. However, isolated MRE TI was also in seen all TB enteritis and 1 intestinal Bechet's. Conclusion: Our results show MRE compares favorably to colonoscopy in the detection of CD in the Chinese population where the incidence is much lower than the West. However, for isolated TI, other differential diagnoses should be considered.

SS 6.6
MR-Enterography: small-bowel wall-thickening diseases not associated with Crohn's disease G. Skouroumouni, D. Panagiotidou, G. Papaderakis, I. Petmezaris, M. Lelegianni, A. Morichovitou, I. Tsitouridis; Thessaloniki/ GR Purpose: In this study, we describe our experience with MR-Enterography (MRE) in accessing small-bowel wall-thickening diseases, other than its main use, that of Crohn's disease. Material and methods: We have retrospectively reviewed 227 consecutive patients who underwent MRE in our institution during a two-year period (January 2015-December 2016). We document and evaluate all other small-bowel wall-thickening diseases, including benign and malignant neoplasms, other in ammatory bowel diseases, infectious processes and celiac disease. All of the patients had a nal diagnosis on the basis of the clinical presentation, follow-up, and in some cases histological proof. Results: 25 patients (11.01%) with small-bowel wall-thickening diseases other than Crohn's disease were found. 11 out of 227 (4.85%) patients had benign and malignant neoplasms arising in isolation or in polyposis syndromes, 9 patients (3.08%) had non-speci c terminal ileitis and were nally diagnosed with intestinal infections, such as Yersinia enterocolitica, 1 patient (0.44%) had ulcerative colitis with backwash ileitis, 2 patients (0.88%) were diagnosed with eosinophilic gastroenteritis and 6 patients (2,64%) had celiac disease. Conclusion: MRE can be a useful diagnostic modality in small bowel diseases, which are not associated with Crohn's disease, and are dif cult to access with other diagnostic modalities. Therefore, radiologists should be familiar with the imaging appearances of these conditions. Delft/ NL, 4 London/ GB, 5 Zurich/ CH Purpose: MRI scores show promise for evaluation of Crohn's disease (CD) activity, although reported reproducibility is variable. Potentially, reproducibility could be improved by use of computer-assisted semi-automated measurements to reduce interobserver variation. The aim of this study was to develop and validate a predictive MRI activity score for ileocolonic CD activity based on computer-assisted semi-automatic measurements of MRI features. Material and methods: The "VIGOR" MRI activity score was developed using subjective radiologist observations and semi-automatic measurements of mural thickness, excess bowel wall volume and DCE (initial slope of increase; ISI) using a retrospective cohort of 27 CD patients against the CDEIS. A subjective score was developed using only radiologist observations. For validation, scores were applied by two observer groups to a prospective dataset of 106 CD patients (59 female, median age 33), along with three existing MRI activity scores (MaRIA, London score and CDMI). Results: The VIGOR score (17.1*ISI+0.2*excess volume+2.3*mural T2) showed moderate correlation to CDEIS (r=0.58-0.59), which was comparable to other activity scores (r=0.34-0.51,p>0.05). The VIGOR score had a signicantly higher interobserver agreement than other activity scores (ICC=0.81 vs. 0.44-0.59, p<0.001). Comparable diagnostic accuracy was seen for the VIG-OR score (80%-81%) to other activity scores (70%-86%). Conclusion: The new VIGOR score achieves comparable accuracy to conventional MRI activity scores, but with signi cantly improved reproducibility, favouring its use for therapy evaluation and monitoring of disease activity. Material and methods: Two experienced GI radiologists reviewed all CT colonography (CTC) studies carried out between 20/10/2011-1/4/2016. Both observers categorised extracolonic ndings according to C-RADS criteria and discrepancies were resolved by consensus. A resident, blinded to study purpose, performed a retrospective review of relevant databases and case-notes for all patients classi ed with an indeterminate stricture (C-RADS C3c) to establish histopathological and clinicoradiological outcomes. Strictures were assessed for shouldering, mural thickness, ulceration, perienteric stranding and lymphadenopathy. Descriptive statistics were produced. Results: In total 1646 studies were performed (94% symptomatic; 6% screening) with a median follow-up 20months (range 1-55). Of a total 73 patients classi ed C-RADS C3c, 14 were deemed too frail or refused further investigation, 4 died of unrelated conditions, 45 underwent endoscopy and 7, interval CTC. Diverticular strictures were con rmed in all cases. The positive predictive value was 0 for cancer and 100% for diverticular stricture. Consequently, no meaningful analysis of morphological features suggestive of underlying malignancy could be performed. Conclusion: In our series, all patients classi ed as CRADS C3c were conrmed as having benign disease by either endoscopy or clinicoradiologic follow-up. Further research to compare morphological characteristics of benign and malignant strictures is required to avoid unnecessary endoscopy.

SS 6.9
Imaging features of missed colon cancers on non targeted abdominal CT: assessment of the missed diagnosis rate M.M. Amitai, M. Eifer, U. Kopilov, V. Belsky, E. Klang; Ramat Gan/ IL Purpose: To assess the missed diagnosis rate of colon cancer on non-targeted abdominal CT and to evaluate the imaging features characteristic of missed cancers. Material and methods: Consecutive patients diagnosed with colorectal cancer by colonoscopy that underwent an abdominal CT scan for reasons other than tumor evaluation within a year before the colonoscopy were included. Missed diagnosis rate by the original radiologists that interpreted the CTs was evaluated. A repeat interpretation oriented for colon cancer detection was separately obtained by two radiologists (study readers) who were blinded to tumors' location. The study readers analyzed the imaging features of detected tumors (shape, length, l wall thickness, free uid, fat stranding, vessel engorgement, stenosis and lymphadenopathy). Associations of imaging features and cancer misses were evaluated. Results: 127 patients were included. Missed diagnosis rate by the original readers was 25/127 (19.7%). Each study reader could not identify the cancer in 8/127 (6.3%) patients. Imaging features associated with misses were absence of fat stranding (p=0.007, p=0.003), absence of vessel engorgement (p<0.0001, p<0.0001) and absence of lymphadenopathy (p=0.005, p=0.004). Missed tumors were shorter than non-missed (1 st

SS 6.10
Ileal neuroendocrine tumours with mesenteric involvement: suggestions for an improved pre-surgical risk stratification with multi-phase CT L. Funicelli, F. Zugni, E. Bertani, F. Ferrari, N. Fazio, M. Bellomi; Milan/ IT Purpose: To suggest additional CT features for the pre-surgical risk strati cation of patients with small intestinal neuroendocrine tumours (siNETs), and to provide preliminary analysis of the performance of such criteria on our patient database. Material and methods: We employed the recently proposed classi cation of mesenteric nodules associated with siNETs, based on superior mesenteric artery (SMA) in ltration [Lardière-Deguelte et al], including as additional criteria the degree of superior mesenteric vein (SMV) in ltration and the presence of extensive mesenteric retraction. A radiologist with 5 years of experience in the staging of siNETs retrospectively and blindly reviewed the images of patients scheduled for surgery in our institution (2006-2016) providing a strati cation into either high-risk or low-risk for unresectability. Results were matched to the surgical reports. Results: 20 multi-phase CT scans of patients with siNETs were selected for evaluation. All 5 patients in the high-risk group had received non-curative surgery (impossible or incomplete resection). Of these, 1/5 was classi ed into the high-risk group for the presence of severe SMV in ltration alone, and 1/5 only for the presence of extensive mesenteric retraction. Fourteen patients in the low-risk group had received curative surgery (14/15), while one (1/15) was unresectable due to the primary tumour extension. Conclusion: The CT evaluation of SMV in ltration and mesenteric retraction, in addition to the standard evaluation of SMA in ltration allowed a correct presurgical risk strati cation in the majority of patients. Material and methods: Institutional review board approval and informed consent were waived. CSI, T2WI, DWI and DCE-MRI were performed in 7 pathologically proved PM-RCCs patients with 14 tumors and in 24 pathologically proved pNETs patients with 24 tumors. Signal intensity was measured in the pancreatic tumor and the spleen on in-phase and opposed-phase images. The signal intensity index (SII) and tumor-to-spleen ratio (TSR) in PM-RCC and pNET were calculated and statistically compared. Receiver operating characteristic (ROC) curve was performed to evaluate the diagnostic ef ciency of SII and TSR in the differentiation of PM-RCC and pNET.
Results: The SII between PM-RCC and pNET (21.0% ± 19.2 vs -3.2% ± 11.4) was statistically different (P < 0.001), so was the TSR (-18.9% ± 19.4 vs 6.0% ± 13.8 ) (P < 0.001). The area under the ROC curve was 0.869 for the signal intensity index (SII) and 0.872 for the tumor-to-spleen ratio (TSR) and permitted the differentiation of PM-RCC from pNET with sensitivity of 71.4%, specicity of 100%, and accuracy of 89.5%. A TSR cutoff value of -13.3% allowed the differentiation of the two groups with sensitivity of 71.4%, speci city of 95.8%, and accuracy of 86.8%. Conclusion: Double-echo GRE chemical shift MR imaging can accurately differentiate PM-RCC with pNET.

SS 7.2
Differences in low-dose CT perfusion the pancreatic neuroendocrine tumors by different commercial software: the one-compartment (or slope) method and deconvolution Y. Nerestyuk, G.G. Karmazanovsky, N. Rubtsova; Moscow/ RU Purpose: To evaluate CT perfusion values for pancreatic neuroendocrine tumors for one-compartment (or slope) method and deconvolution. Material and methods: 6 patients with known pancreatic neuroendocrine tumor (3 patients: well-differentiated tumors (G1) and 3 patients: moderately differentiated tumors (G2)) underwent whole-pancreas perfusion by a 256-slice CT (Brilliance iCT; Philips). 80-kVp/100mAs (low-dose) image data were reconstructed with iDose 5 iterative reconstruction. Perfusion parameters were calculated with method 1: one-compartment (or slope) model (Philips), and method 2: deconvolution (General Electric). The parameters generated included: the one-compartment (or slope) method: the blood ow (BF, ml/min/100 ml tissue), blood volume (BV, ml/100 ml tissue), arterial blood ow (AF, ml/min/100 ml tissue), portal blood ow (PF, ml/min/100 ml tissue), perfusion index (%) [PI = AF/(AF + PF) x 100]; deconvolution: BF, BV, time to peak (TTP), mean transit time (MTT), permeability surface (PS) and other. Results: CT-perfusion characteristics of the healthy pancreatic parenchyma were: method 1: BF-72±32.2 ml/min/100ml, BV-26.2±10.7 ml/100ml, AF-58.9±17.5 ml/min/100ml, PF-30.3±28.8 ml/min/100ml; method 2: BF-162±20.5 ml/min/100ml, BV-19.2±7.7 ml/100ml. There was a signi cant difference neuroendocrine pancreatic tumors between G1 and G2 for PI (p<0,05). TTP for G1 tumor decreases, for G2 maybe not changed in comparison with the parenchyma. TTP for the healthy pancreatic parenchyma was 16±6.4 sec. The radiation dose was 7.4±0.4mZv. Conclusion: The different commercial software can be used as an additional parameter for identifying pancreatic neuroendocrine tumors grading. Material and methods: Morphological analysis was performed on 38 nonH-NETs and on 38 PDAs matched by size and included tumour size, margins, homogeneity, enhancement pattern, MPD and vascular invasion. Texture analysis was performed using MaZda in 19 nonH-NETs for whom non-contrast CT was available and in their matched PDAs. Data were analysed with t-test and Fisher's test. Results: Morphological analysis revealed a mean tumour diameter of 46 mm for nonH-NETs and 45.5 mm for adenocarcinomas (p=ns). 41/57 nonH-NETs had well-de ned margins, in contrast to 15/57 adenocarcinomas (p<0,0001). Calci cations were seen in 5/57 nonH-NETs and none of the adenocarcinomas (speci city 100%). No difference in tumour homogeneity was observed (p=ns). In the arterial phase 33/57 nonH-NETs were hypodense and 22/57 were isodense, 54/57 adenocarcinomas were hypodense and 3/57 isodense (p<0.0001). 15/57 nonH-NETs and 11/57 adenocarcinomas showed increasing enhancement from the arterial to the venous phase (p=ns). The upstream MPD had signi cantly larger calibre in adenocarcinomas (5.8 vs 2.2 mm; p<0.0001). No difference was observed for vascular invasion (p=ns). Texture analysis did not reveal differences between nonH-NETs and PDAs regarding variance, skewness and kurtosis. Conclusion: NonH-NETs have well-de ned margins, cause less MPD dilation, and are more commonly isodense than adenocarcinoma. Tumoural calci cations are speci c for nonH-NETs. Texture analysis is not helpful for the differential diagnosis. The 18 FDG uptake of the tumor was compared to uptake of the liver and dened as: low/no uptake, homogeneous high or heterogeneous high uptake. Kaplan-Meier curves and log-rank test were used to correlate ow-metabolic phenotypes to overall survival (OS). Results: Three hyperattenuating PDAC were excluded from analysis. The overall survival for PVP isoattenuating PDAC (n=62) is signi cantly better compared to hypoattenuating PDAC (n=35): 63±18.1 weeks and 35±4.6 weeks respectively (p=0.003). Low/no 18 FDG uptake is correlated with better OS compared to high homo-or heterogeneous uptake (n=81): 63±18.8 weeks, 46±5.1 weeks and 33±5.3 weeks, respectively (p=0,049). Hypoattenuating PDAC with high 18 FDG uptake has the poorest prognosis whereas the isoattenuating PDAC with low 18 FDG uptake has the best prognosis (32±2,9 weeks and 63±24,7 weeks, respectively) (p=0.009). Conclusion: The ow-metabolic phenotype on CECT and 18 FDG-PET scan of pancreatic ductal adenocarcinoma can be used as a prognostic factor for overall survival. Purpose: This analysis aimed to establish the most appropriate post neo-adjuvant (PNA) imaging modality for those with borderline resectable pancreatic cancer, how the imaging ndings should be interpreted, its ability to predict resectability and likelihood of an R0 resection rate. Material and methods: Evidence-based medicine methods were used. Ranking and appraisal was performed using the Oxford Centre for Evidence Based Medicine tools. Research studies and trials were identi ed by searching MED-LINE, EMBASE, the Cochrane Central Register of Controlled Trials and Pubmed from January 2006 to December 2016. Review articles and those inadequately powered were excluded. Results: No studies were retrieved from the secondary literature. From the primary literature, four relevant articles were identi ed, analysed and ranked in a "hierarchy of evidence". All studies were retrospective, published since 2012 and provided level three evidence. Three studies demonstrated that the majority of patients still had borderline disease post neoadjuvant treatment. All four studies showed that CT staging was not predictive of resectability and pathological response, that radiographic down-staging post neoadjuvant therapy is rare and RECIST response is not an appropriate treatment goal.
Conclusion: Best available evidence shows that CT staging PNA chemotherapy does not predict resectability or R0 resection rate. PNA imaging should consist of CT thorax, abdomen and pelvis rather than CT pancreas protocol and the interpretation should focus on evidence of distant metastases rather than local disease staging.

SS 7.6
CT features of pancreatic adenocarcinoma: are there any differences between tumors < or ≥2 cm? A. Rossi, G.A. Zamboni, M.C. Ambrosetti, G. Marchegiani, R. Pozzi Mucelli; Verona/ IT Purpose: According to the surgical literature, the prognosis of patients affected by pancreatic adenocarcinoma depends on the size of the primary tumor. Purpose of this study is to assess if the size measured at CT corresponds to the size at pathology and if there is any difference between tumors < or ≥2 cm. Material and methods: We reviewed 107 patients with resected pancreatic carcinoma and a preoperative multiphasic MDCT. One reader measured tumor diameter and conspicuity both in the arterial and venous phase, and assessed arterial and venous involvement. Tumor diameters measured at CT were compared with diameters at pathology. Data were analyzed by means of Bland-Altman's test, Cohen's Kappa, t-test and Fisher's test. Results: Patient population included 39 pts with tumors <2cm and 68 ≥2cm at pathology. Tumor conspicuity was higher in the arterial phase than in the venous phase for both groups (differences in conspicuity between the two phases were respectively 19.41 HU and 18.16 HU, p < 0.0001). Between the two groups there were no differences in conspicuity, in upstream MPD dilatation, in arterial or venous invasion, neither regarding complete (R0) or incomplete resections (R1-2) (all p=ns). The agreement between arterial phase CT and pathology measurements was 0.423 (IC 95% 0.243-0.602). Conclusion: Preoperative CT does not show different features for tumors smaller or larger than 2 cm. CT measurements in arterial phase are as reliable as pathology measurements.

SS 7.7
Dedicated diffusion weighted MR imaging for staging peritoneal metastases: a preoperative selection tool for cytoreduction surgery (CRS) candidates I. Van Purpose: There is an urgent need for non-invasive imaging methods that accurately stage peritoneal metastatic burden to select candidate patients for cytoreductive surgery (CRS)/HIPEC. We compared the peritoneal cancer index (PCI) estimated preoperatively by diffusion weighted MR imaging (DW-MRI) with PCI found at surgery to assess whether DW-MRI can be used to select CRS/HIPEC candidate patients. Material and methods: Twenty-three consecutive patients (M/F= 8/15) with histologically proven peritoneal carcinomatosis of either colorectal (n=16) or ovarian (n=7) origin were included. Patients underwent preoperative DW-MRI (scan time = 30 min) before exploratory laparoscopy or CRS/HIPEC. Two independent radiologists prospectively determined PCI following DW-MRI, categorizing patients as low-risk (PCI 0-19, operable) or high-risk (PCI 20-39, inoperable). PCI following DW-MRI was compared to PCI found at surgery. Results: Mean PCI found at surgery was 10.5 (range 0-27). Mean radiological PCI for reader 1 was 15.0 (range 2-30) and 13.3 (range 0-29) for reader 2. Interobserver agreement was excellent (k = 1.0). Both readers staged 20 out of 23 patients correctly (accuracy 87%) compared to surgical ndings; a single patient was overstaged following DW-MRI. Conclusion: Data from this ongoing study suggest that DW-MRI is a reproducible and accurate selection tool to noninvasively select patients who could potentially bene t from CRS/HIPEC. Purpose: Preoperative assessment of the extent of peritoneal carcinomatosis (PC) is invaluable in the work up of potential cytoreductive surgery (CRS)/hyperthermic intraperitoneal chemotherapy (HIPEC) candidates. The aim of this study was to assess whether there is a role for ADC value measurement (as a biomarker for tumor aggressiveness) in this process. The correlation between mean ADC values of peritoneal lesions and the operability of these patients was analyzed. Material and methods: In this ongoing study 17 patients (8 female/9 male) with peritoneal carcinomatosis originating from colorectal carcinoma were included. All patients underwent diffusion weighted MRI prior to CRS/HIPEC. An experienced radiologist measured mean ADC value of the most assessable peritoneal lesion in each patient. Accordingly it was veri ed if the patient had undergone CRS/HIPEC (n=11) or did not (n=6). Eligibility for CRS/HIPEC was evaluated by diagnostic laparoscopy. Unpaired t-test was used to measure the relationship between these two variables. Results: Mean ADC value was 1036 (range 813-1256) for patients who had undergone CRS/HIPEC and 1161 (range 632-2157) who were deemed inoperable at laparoscopy. Mean ADC values were signi cantly higher (p< 0.0001) in patients who underwent cytoreductive surgery. Conclusion: This preliminary data demonstrates a correlation (p<0.0001) between low ADC-values and non-operability in patients with PC originating from colorectal carcinoma. This entails that a simple ADC measurement might be used as an additional biomarker in the operability assessment of these patients.

SS 7.10
CT imaging of primary pancreatic lymphoma E. Boninsegna, D. Facchinelli, G.A. Zamboni, R. Negrelli, A. Ambrosetti, R. Pozzi Mucelli; Verona/ IT Purpose: Primary pancreatic lymphoma (PPL) is a rare neoplasm; the correct diagnosis is essential because it bene ts from chemotherapy, indeed surgery may increase morbidity. Our aim is to evaluate the CT characteristics of PPL basing upon a relatively large series of cases examined in a single institution. Material and methods: Five patients were enrolled. CT examinations included: unenhanced scan, contrast-enhanced pancreatic phase and portal phase. We evaluated: tumor location; peripancreatic vessel encasement; necrosis; enlarged lymphnodes; fat stranding; enlarged hepatic duct or main pancreatic duct; neoplasm longest dimension, volume and density. Results: Qualitative analysis 3/5 PPL were located in the pancreatic head, 1 in the tail, 1 involved the whole gland. In 4/5 cases superior mesenteric artery and vein were encased; splenic vein and artery encasement was depicted in one PPL. Enlarged lymphnodes were present in 3/5; fat stranding was depicted in 5/5. Hepatic duct was dilated in 3/5; main pancreatic duct was enlarged in 2/5. Quantitative analysis Mean neoplasm longest dimension and volume were 12.8 cm and 605.4 cm 3 . Mean tumor density was 39.4HU in basal condition, 56.2HU in the pancreatic phase and 67HU in the venous phase. Conclusion: PPL presents several CT features in common with adenocarcinomas and autoimmune pancreatitis; useful signs for the differential diagnosis are: presence of a large lesion with homogeneous delayed enhancement; vessel encasement without in ltration; mild or absent dilatation of the main pancreatic duct; positive peripancreatic fat stranding. ) who underwent liver transplantation after loco-regional therapy with transarterial chemoembolization plus radiofrequency ablation (n=56) or 90 Yttrium radioembolization (n=5) were included in this retrospective study. All patients underwent MRI using routine liver protocol using gadolinium-based contrast agents with image subtraction within 90 days of transplant. The following criteria were assessed by 3 independent readers: RECIST, mRECIST, EASL, percentage of necrosis on subtraction images (on arterial and portal venous phases), and qualitative/quantitative diffusion-weighted imaging (signal intensity and ADC). Degree of tumor necrosis/viability was retrospectively assessed in all index tumors at histopathology. Logistic regression and ROC analyses were used to determine predictors of complete pathologic necrosis (CPN=100% To determine the prognostic value of CT-perfusion (CT-p) imaging in evaluation of blood ow changes related to therapeutic effects of sorafenib by quantitative analysis of tumor vascularization. Material and methods: Eighty-one CT-p studies were performed in 22 patients, with biopsy-proven diagnosis of HCC lesion, who underwent target antiangiogenetic therapy. Perfusion studies were performed at baseline and during treatment follow-up (every 3 months) on 256 multidetector CT (iCT, Philips), with following parameters:100 Kv,100 mAs;16 dynamic slices/scan; 40 dynamic scans; 50 ml of contrast medium. Target lesions and surrounding parenchyma were evaluated using a dedicated perfusion software which generated a quantitative colour map of vascularization. Following perfusion parameters were considered: hepatic perfusion (HP); arterial perfusion (AP); blood volume (BV); hepatic perfusion index (HPI) and time to peak (TTP) and statistically compared between responders (complete response, stable disease or partial response) and non-responders. Purpose: Esophageal functional magnetic resonance (fMR) represents a dynamic test which is rarely applied in clinical practice. The aim of this study was to assess diagnostic value of fMR in patients with achalasia in comparison with manometric ndings regarding 3 types of achalasia. Material and methods: Examination was performed on a 1.5T scanner using T2W SSFSE sequences speci ed protocol. Twenty-ve achalasia patients underwent functional testing while swallowing water. Esophageal dilation was measured on axial images and was de ned as a diameter >3.5 cm. Loss of peristalsis and motility dysfunction were determined as good, poor and absent. These results were compared to manometric ndings, as a gold standard, and were divided according to 3 types of achalasia. Results: Manometric examination showed 6 patients with achalasia type 1, 18 with type 2 and one with type 3, while fMR indicated 7, 16 and 2 patients according to types, respectively. Overall accuracy of fMR was statistically similar to that of manometry (p=0.58). In six (24%) patients we found mismatch between types 1 and 2. Diagnostic performance of fMR was satisfactory, with sensitivity and speci city of 80% and 100%, respectively. Conclusion: Esophageal fMR was found to be a feasible and highly speci c dynamic technique, useful in determining types of achalasia and possibly clinically signi cant, regarding the outcomes of therapy.

SS 9.2
The potential usefulness of MDCT in diagnosis of acute appendicitis in patients with atypical clinical presentation and to compare findings S. Merola; Narbonne/ FR In these cases, histopathology con rmed the diagnosis of AA in 58/78 (74.4%) patients. The three remaining MDCT-positive patients (4.3%) were false-positive with a nal diagnosis of perforated diverticulitis of the sigmoid colon. In those patients, the appendix could not be identi ed retrospectively, because of abscess formation in the RLQ. In the remaining 14/78 patients (17.9%) who underwent surgery, ischemic colitis was diagnosed on MDCT in 10 patients (10.2%), and acute cholecystitis in 8 patients (7.7%) with histopathologically con rmation, respectively. Among the remaining patients, who did not undergo surgery (60/138), colitis was diagnosed on MDCT (n=10), urolithiasis (n=6), pyelonephritis (n=4), pancreatitis (n=4), and MDCT was unremarkable (n=36). Conclusion: MDCT usefulness has a fundamental impact in clinically equivocal cases of acute appendicitis.

SS 9.3
Significance of timed barium esophagography in evaluation of laparoscopic myotomy for achalasia S. Jovanovic, A. Djuric-Stefanovic, O. Skrobic, N. Ivanovic, D. Masulovic, A. Simic; Belgrade/ RS Purpose: Laparoscopic Heller myotomy (LHM) represents therapy of choice for patients with achalasia, providing excellent long-term outcomes. Aim of this study is to access signi cance of timed barium esophagography (TBE) in evaluation of esophageal emptying before and after LHM. Material and methods: Fifty-one patients with achalasia, con rmed by manometric studies, underwent TBE which was performed by swallowing 250 ml of low-density barium sulfate. Radiographic plane upright frontal lms were performed 1, 2 and 5 min after ingestion. In all patients a LHM was done, and on the second postopertive day control TBE was obtained in the manner described above. Premyotomy and postmyotomy height and width of the barium column were compared. The quantitative assessment based on measurements of the barium column is an accurate and bene cial method, which estimates esophageal emptying success of LHD at short-term outcomes.

SS 9.4
Evaluating response of locally advanced gastric adenocarcinoma to neoadjuvant chemotherapy using intravoxel incoherent motion MRI: preliminary results Y. Zhu fusion fraction (f) and pseudo-diffusion coef cient (D * ) maps were calculated from the bi-exponential model. IVIM parameters (D, D * , and f) of LAGC were measured by region-of-interest (ROI) methods using the FuncTool on GE AW4.6 workstation. All patients received radical resection within 2 weeks after the second examination. According to the Mandard pathologic tumor regression grade (TRG), subjects were divided into responders (TRG 1-3) and nonresponders (TRG 4,5). The IVIM parameters before (pre-parameters) and after (post-parameters) NCT and their corresponding changes (∆parameters) between the two groups were compared using the Student's t test or nonparametric test. The diagnostic performance of different parameters was judged by the receiver-operating characteristic curve (ROC) analysis.
Results: Based on the Mandard TRG criteria after 4-6 cycles of NCT, 25 patients were categorized into the responder group whereas the other 17 patients were considered nonresponders. The D value was signi cantly higher after treatment and the f value was signi cantly lower (all P < 0.05). In contrast, D * value was only slightly lower after treatment. Compared with nonresponders, a notably higher post-D value, ∆D and ∆f were observed in responders (all P < 0.05), but no signi cant change other parameters among the 2 groups (P > 0.05). The ROC curve analysis indicated that the cutoff of ∆D value in best predicting TRG was 0.42×10 -3 mm 2 /s, and the corresponding AUC, sensitivity, and speci city were 0.841, 66.7%, and 100.0%, respectively. Conclusion: IVIM-derived parameters, especially the D and f value, showed potential value in the prediction and response monitoring to neoadjuvant chemotherapy in LAGA. .014), respectively. There was weak positive correlation for His angle in both groups, weak positive correlation in patient and high positive correlation in control group for hiatus area between readers. Although there was not signi cant difference in crural thicknesses between two groups, there was high positive correlation in crural thicknesses at celiac level between readers. Conclusion: Hiatus area and His angle are the main affected parameters in hiatal hernia. Hiatus area is the most powerful criteria in interobserver agreement that could easily take place in routine preoperative CT reports of hiatal hernia patients for the purposes of surgery management and outcome prediction.

SS 9.6
Diffusion-weighted MRI in the assessment of response of gastric cancer to neoadjuvant therapy: correlation of the ADC values with tumor regression grade Y. Zhu, L. Jiang, Y. Li; Beijing/ CN Purpose: To investigate the value of diffusion-weighted (DW) MRI in response assessment after neoadjuvant chemotherapy (NCT) in patients with locally advanced gastric cancer (LAGC). Material and methods: 40 patients with LAGC underwent respiratory gated DWI with b=0 and 800 s/mm 2 on a 3.0-Tesla MRI scanner before starting therapy and after NCT. All patients underwent radical resection in 2 weeks after the second examination; surgical pathologic tumor regression grade (TRG) represented the reference standard. Subjects were divided into responders and nonresponders according to the TRG. ADC value before (pre-ADC) and after (post-ADC) NCT and their corresponding change (∆ADC) between the two groups were compared using the Student t test. The diagnostic performance of different parameters was judged by the receiver-operating characteristic (ROC) curve analysis.
Results: Based on Mandard TRG criteria after 4-6 cycles of NCT, 25 patients were categorized into the responder group whereas the other 16 patients were considered nonresponders.The ADC value was signi cantly higher after treatment. ∆ADC value was notably higher in responders compared with nonresponders, but no signi cant change in other parameters among the 2 groups. The ROC curve analysis indicated that the cutoff of ∆ADC value in best predicting tumor NCT response was 0.25×10 -3 mm 2 /s, and the corresponding AUC, sensitivity, and speci city were 0.770, 70.8%, and 81.2%, respectively. Conclusion: The ADC value showed potential value in the prediction and response monitoring to NCT in LAGC.

SS 9.7
Non-enhanced Fast-MRI for radiological evaluation of acute appendicitis H. Goessmann, L.M. Dendl, B. Pregler, M. Scherer, S. Opitz, A.G. Schreyer; Regensburg/ DE Purpose: Evaluation of a fast, non-enhanced and sequence-reduced MRI (Fast-MRI) without the use of oral, rectal or intravenous contrast-medium regarding diagnostic accuracy in patients with suspected acute appendicitis. Material and methods: 47 consecutive patients with suspected acute appendicitis in a tertiary care hospital were enrolled in this study. In addition to the routine workup (clinical examination and laboratory ndings) a Fast-MRI (dura-tion=12 min) with an orienting T2 Haste and DWI (b=0; 600; 1000) was performed. A diffusion restriction of the appendix was interpreted as in ammation. The results of the MRI were compared to the clinical course or the intraoperative ndings.
Results: 25 of the 47 patients underwent surgery. Among those 16 diffusion restrictions and 7 normal DWIs were found. Two examinations were inconclusive. 14 of the 16 diffusion-restricted appendices showed an acute in ammation intraoperatively. The other 2 had malignant ndings (appendix and coecum carcinoma). 5 of the 7 not restricted appendices showed no signs of in ammation intraoperatively. All of the 22 patients, who were treated conservatively, had a normal nding in the MRI. Sensitivity as well as the PPV was 87.5%, while speci city and NPV were 93.1%, respectively. Conclusion: In this prospective study, Fast-MRI proved to be a quick and radiation-free option to rule out or con rm an acute appendicitis with a high degree of certainty. Material and methods: All patients sent to ultrasound over a year period with concern of appendicitis were included. Outcomes were gathered. Patients were scanned by 1 of 15 sonographers with experience of 2-33 years or by a resident. Chi-square tests and one-way ANOVA were used to test the associations of ultrasound results with sonographer's years of experience, length of exam, patient age and the impact on patient care. Results: Of 455 patients, the appendix was not visualized in 52% of cases. Non-visualization of the appendix per sonographer ranged from 5% up to 69%. Subsequent CT was performed in 36% of these patients with 83% having normal results. The sonographer's success in visualizing the normal appendix ranged from 8% to 76% with an average of 24%. There was no correlation between sonographer's years of experience, time spent scanning or resident year of training with success of visualizing the normal appendix. Conclusion: An individual sonographer's perception skill plays signi cant role in the ability to visualize the appendix, which affects CT utilization and hospital costs and impacts patient care.

Scienti c Session SS 10
Best rated papers on colorectal cancer evaluation

SS 10.1
Response to neoadjuvant therapy in locally advanced rectal cancer: assessment combining standardized index of shape of dynamic contrast-enhanced MRI and intravoxel incoherent motion method of diffusionweighted MRI R. Fusco, M. Petrillo, V. Granata, A. Petrillo; Naples/ IT Purpose: To evaluate MRI for neo-adjuvant therapy response in locally advanced rectal cancer (LARC) using dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and diffusion-weigthed imaging (DWI). Material and methods: 40 patients with LARC underwent MR examination before and after pCRT. SIS, apparent diffusion coef cient (ADC), tissue diffusion (D t ), pseudodiffusion (D p ) and perfusion fraction (f) were calculated for each patient before and after pCRT. Percentage changes were evaluated. Non-parametric sample tests, receiver operating characteristic curve (ROC) and diagnostic performance were performed. Results: 23 Patients were classi ed as responders (TRG≤2) and 17 as nonresponders (TRG>3). Statistically signi cant differences in SIS and ADC percentage change median values between responders and non-responders patients were found. The best parameters to discriminate responder by non-responders patients (ROC analysis) were ∆SIS and ∆ADC with an accuracy of 85% and 83% (cut-off value of 5.2% and 21.4%, respectively) while to discriminate pathological complete response were ∆f and ∆D p with an accuracy of 80% and 85% (cut-off value of 64.0% and 70.6%, respectively). Combining ∆SIS and ∆D t an accuracy of 95% and 93% was obtained to discriminate responder from non-responder patients and to assess pathological complete response. Conclusion: SIS is a promising DCE-MRI angiogenic biomarker useful to assess preoperative treatment response directing surgery for more or less conservative treatment. If combining ∆SIS and pure tissue diffusion coef cient change (∆D t ), a further increase of accuracy could be obtained. Material and methods: We prospectively enrolled 40 consecutive CRC patients, who underwent pretreatment, midtreatment and posttreatment 3T MRI. A region of interest was drawn manually around the tumor on unenhanced T2-weighted images and analyzed using TA (TexRAD), evaluating texture parameters (skewness, kurtosis). After CRT, all patients underwent complete surgical resection and the surgical specimen served as the gold standard. Receiver operating characteristic (ROC) curve analysis was performed to assess the discriminatory power of each quantitative parameter to predict complete response.
Results: Thirteen patients showed pathological complete response (pCR). Twenty-two patients showed partial response (PR) and ve patients were nonresponders (NRs). pCR patients showed a signi cantly lower kurtosis at pre-CRT analysis compared to PR+NR subgroup (P=0.01). During-treatment kurtosis was signi cantly higher in pCR compared to PR+NR (P=0.043). PR+NR showed a little change in kurtosis between pre-CRT and during-CRT compared to pCR (P = 0.039). The optimal cutoff value for pretreatment kurtosis was 0.17, resulting a sensitivity and speci city for pCR prediction of 100% and 79.4%, respectively. Conclusion: TA from T2w images can potentially have an important role as imaging biomarkers of tumoral response to neoadjuvant CRT in rectal cancer.

SS 10.4
Prevalence of post-investigation colorectal cancer ("interval cancer") after computed tomographic colonography: a systematic review A. Obaro 1 , A. Plumb 2 , T. Peer-reviewed studies published after 1994, and with over 12 months followup, were selected and data extracted by two independent authors. The PICRC rate was pooled across studies using random-effects meta-analysis. Following World Endoscopy Organization recommendations, PICRC rates were calculated using (a) number of cancers identi ed and (b) number of CTCs completed as denominators. Material and methods: One hundred patients with biopsy-proven primary rectal adenocarcinoma referred for thoracic and abdominal CT staging were enrolled in this study. The contrast-enhanced MDCT scans were performed on a 256 (ICT, Philips) row scanner with the following acquisition parameters: tube voltage 120 KV and tube current 150-300 mAs. Multiplanar CT reconstructions were performed and imaging data were reviewed as axial and as MPR images, along with rectal tumour axis. MR study, performed on 1.5 T magnet included standard multiplanar T2-weighted and axial T1-weighted sequences. The MRF involvement was assessed on axial and MPR images independently and compared with MRI imaging nding. Diagnostic accuracy of both modalities was compared and statistically analyzed.
Results: Multidetector-row CT images agreed with those of MRI, obtaining CT axial images sensitivity and speci city of 80.4% and 75%, PPV of 80.4%, NPV 75% and accuracy of 78%. Using MPR the sensitivity increased up to 88% and speci city to 87%, PPV was 90%, NPV 85.36% and accuracy of 88%. Conclusion: New-generation CT scanners, using high-resolution MPR images along tumor axis, can be considered a complementary technique for loco-regional and whole body staging in rectal cancer patients, especially in those with MRI contraindications.

Scienti c Session SS 11
New insights into pancreatic abnormalities and techniques SS 11.1 Definition of age-dependent reference values for diameter of the common bile duct and pancreatic duct on MRCP from a population-based cohort study Results: 1385 subjects were initially scanned, 865 measured PDs and 938 CBDs were included for analysis. Subjects were excluded for missing data or evidence of pancreaticobiliary disease. Median age was 53y (21-89y) and 48.5% were female. The diameters increased with age (PD median (1st-3rd quartile), 20-29 years: 1.33mm (1.20-1.57), >70 years: 2.49mm (1.85-3.01); CBD median (1st-3rd quartile), 20-29 years: 4.53mm (3.87-5.17), >70 years: 6.50mm (5.10-8.23)) and the historic upper limit of normal of 3mm for PD and 7mm for CBD were exceeded by 11% and 18.2%, respectively. Subjects that underwent cholecystectomy presented with signi cantly increased diameter of CBD, but not PD (CBDw  Material and methods: This retrospective study was approved by institutional review board and the requirement of informed consent was waived. A total of 42 patients (M:F =25:17, mean age 59.4±14.9 years) who underwent conventional respiratory-triggered MRCP and breath-hold MRCP using a 3T scanner were included. Three board-certi ed radiologists independently reviewed both exams for bile duct and pancreatic duct visualization and overall image quality on a ve-point scale, and image sharpness and background suppression on a four-point scale, with a higher score indicating better image quality. The results and acquisition time were compared using either the Wilcoxon signed-rank test or paired t test between the two MRCP exams as appropriate.
Results: The acquisition time of breath-hold MRCP was 21 seconds, and that of respiratory-triggered MRCP was 5:30 seconds on average. Overall image quality was signi cantly higher in breath-hold MRCP than respiratory-triggered MRCP ( Purpose: To assess the interobserver agreement of categorizing pancreatic and peripancreatic collections using the revised Atlanta classi cation (RAC) in the rst month of acute pancreatitis (AP), and to correlate the type of collection with outcome. Material and methods: Among 131 consecutive patients hospitalized for 139 episodes of AP, 195 CTs performed within the rst month, and presenting peripancreatic anomalies were analysed. The episodes of AP were classi ed into three degrees of severity according to the RAC: mild, moderately severe and severe. Two radiologists blinded to the clinical data restrospectively categorized the collections as acute peripancreatic uid collections (APFC) or acute necrotic collections (ANC), using the RAC criteria. Interobserver agreement was assessed using statistics, and compared according to the timing of the CTs. Results: Interobserver agreement to categorize peripancreatic collections was moderate ( =0.45). This agreement did not increase with the timing of the CTs: was 0.51 before day 3, 0.36 between day 3 and day 7 and 0.53 after day 7. For the detection of parenchymal necrosis, the interobserver agreement was also moderate ( =0.42). There was a tendency for less severity among patients with APFC compared with patients with ANC (p=0.06). Conclusion: The RAC shows a moderate interobserver agreement to categorize both peri-and intra-pancreatic collections. The type of collection occuring during the rst month of an AP is associated with the severity of the pancreatitis.

SS 11.4
Evaluation of extracellular volume fraction and T1 mapping for the diagnosis of chronic pancreatitis T. Tirkes, E. Cui, P.R. Territo, C. Lin, B.L. Bick, E.L. Fogel, F. Akisik; Indianapolis, IN/ US Purpose: To determine if extracellular volume (ECV) fraction and T1 relaxation times can be useful diagnostic criteria for chronic pancreatitis (CP). Material and methods: Total of 122 patients were grouped into the normal (n=53), suspected (n=26) and de nite (n=43) CP cohorts based on enrollment criteria, which included history, clinical ndings, endoscopic US, ERCP and MRCP results. T1 mapping was performed using dual ip angle (2° and 13°) gradient echo imaging technique on the same 3T scanner (TR: 3.87, TE: 1.32). T1 maps were acquired in the pre-and 5-minute delayed post-contrast phases and reconstructed with Siemens MapIt Software. ECV fraction was calculated using the formula ECV= [1-hematocrit] x ∆R1 pancreas / ∆R1 blood . Results: There was statistical signi cance between the average ECV fraction of the 3 groups; normal (0.27 ±0.08) and suspected (0.38 ±0.10, p=0.0009), normal and de nite (0.46 ±0.10, p<0.00001) and suspected and de nite (p=0.015) CP groups. Average T1 relaxation times also showed statistical signi cance between the normal (751±221 msec) and suspected CP group (1018 ±295 msec, p=0.01) and normal and de nite CP groups (1105 ±356 msec, p< 0.00001). ECV fraction >0.32 was 80% sensitive and 84% speci c for the diagnosis of suspected and de nitive CP (AUC= 0.88). Combined criteria of ECV fraction >0.30 and T1 >1000 msec showed 92% sensitivity and 71% speci city (AUC= 0.89). Conclusion: ECV fraction and T1 relaxation times can be useful imaging techniques to diagnose CP.

SS 11.6
Incidence, prevalence and pancreatic-related mortality of pancreatic cysts in a population-based study

Munich/ DE
Purpose: To analyze the prevalence, incidence and clinical relevance of pancreatic cysts detected as incidental nding in a population-based longitudinal study. Material and methods: 1,077 participants (521 men, mean age 55.8±12.8 years) underwent T2-weighted MRCP at baseline in a population-based study. 676 subjects received a 5-year follow-up, mortality follow-up was performed after 6 years. MRCP was analyzed for pancreatic cysts with a diameter >2 mm. The prevalence and incidence of pancreatic cysts were assessed in association to age, gender and suspected risk factors (BMI,smoking,lipase,triglycerides,total cholesterol,HbA1c,alcohol,systolic blood pressure). Results: At baseline pancreatic cysts were detected with a prevalence of 49.1% (total 1,681 cysts in 494 subjects), with an average number of 3.9 (CI 95% 3.2; 4.5) cysts per subject. Prevalence (p<0.001), number (p=0.001) and maximum size (p<0.001) of pancreatic cysts increased signi cantly with age. Concerning risk factors, a signi cant association was seen between cyst development and BMI, triglyceride level and total cholesterol. The 5-year followup revealed an incidence of 12.9% newly detected pancreatic cysts. 35.4% of the subjects initially harboring pancreatic cysts, showed an increase in number and/or maximum cyst size. No participant died of pancreatic diseases within mortality follow-up. Conclusion: The prevalence of pancreatic cysts in general population is unexpectedly high and increases with age. Prospective follow-up data in a population-based setting suggests that most pancreatic cysts are harmless incidental ndings.
Purpose: To evaluate the prevalence of pancreaticoduodenal artery (PDA) aneurysms, and to de ne possible etiological associations. Material and methods: We conducted a record-based cross-sectional study at our institution, collecting radiological reports of abdominal contrast-enhanced CT scans of 12,736 patients, performed from January 2012 until December 2016. Reports were reviewed searching for visceral aneurysms. Then the corresponding CT images were evaluated, selecting only studies showing PDA aneurysms.
Results: PDA aneurysms were identi ed in 7 patients (1 male and 6 females). The mean age was 60.7 years (range 30-81 years). The mean aneurysm size was 2.0 cm (range 1.1-3.5 cm). One patient had a gastroduodenal artery (GDA) aneurysm associated. Six patients (86%) had severe stenosis or occlusion of the celiac artery, most likely due to compression by the median arcuate ligament. Three patients (43%) had portal hypertension signs. Six patients (86%) had enlargement of at least one PDA; 6 patients (86%) had enlargement of GDA; 1 patient (14%) had enlargement of dorsal/transverse pancreatic artery. We calculated a prevalence of PDA aneurysms of 0.055% (55 /100,000 patients).
Conclusion: The prevalence of PDA aneurysms found in our study (55/100,000) is higher than that reported in the current literature (1-2/100,000) derived from autopsy studies. We found an association with celiac artery compression in 86% of patients, in disagreement with a literature frequency of 50-60%, most likely due to the low number of cases studied.

Stockholm/ SE
Purpose: To develop and validate a supervised learning model for automatic segmentation of pancreatic adenocarcinoma based on texture features extracted from diffusion-weighted MR images. Material and methods: 14 patients with pancreatic adenocarcinoma underwent MRI examination on a 1.5T system. B1000 DWI images were used to extract texture features. An expert radiologist manually traced regions of interest separately for pancreatic cancer and non-cancerous tissue. 11 Haralick texture features were extracted on a pixel by pixel basis. 4 different models based on Naïve Bays, J48, k-NN and Random Forests were trained using 80% of the data with a 10-fold cross-validation method. The number of features were reduced to 100 from 310 by means of 4 feature selection methods (correlation between features and with classes, entropy and relief). Testing was based on 20% of the data that were not included in the training phase. Area under the receiver operating characteristic curve (AUROC) was calculated and compared between all 4 models. Results: Random forest algorithm provided with the highest AUROC (1.00, F value: 0.99) while the k-NN resulted in the second highest AUROC (0.992, F value: 0.995). Differences between RF and k-NN in AUROC did not reach statistical signi cance (p=0.76).
Conclusion: Automatic segmentation of pancreatic cancer is feasible using texture features from b1000 diffusion images in combination with a random forest model.

SS 11.9
Role of the pre-operative CT scan in predicting postoperative pancreatic fistula after pancreaticoduodenectomy A. Ng 1 , E. Neo 2 , K. Tew 2 , H. Kanhere 2 ; 1 Southport, QLD/ AU, 2 Woodville South, SA/ AU Purpose: Pancreaticoduodenectomy (PD) is the mainstay operative treatment of pancreatic head and periampullary tumours. Despite advances in operative technique and centralisation of PD to high-volume centres, peri-operative morbidity remains high. Postoperative pancreatic stula (POPF) is a major contributor to both mortality and morbidity. The aim of this study was to assess the role of the pre-operative CT scan in assessing for pancreatic duct location and size, and fatty pancreatic in ltration and their association with POPF. Material and methods: 65 patients who underwent PD were retrospectively identi ed from a database at an Adelaide tertiary hospital. Age, sex and POPF were obtained from the patients' records. Pancreatic diameter and position, and pancreatic tissue attenuation was measured from pre-operative CT scans. Purpose: To clarify the correlation of the co-activation of -catenin and hepatocyte nuclear factor (HNF) 4 with the ndings of gadoxetic acid-enhanced MRI, organic anion transporting polypeptide (OATP) 1B3 expression and histological differentiation grade in HCC. Material and methods: Surgically resected 196 HCC in 174 patients were enrolled in this study. HCC were classi ed into four groups by immunohistochemical expression of -catenin, glutamine synthetase (GS) and HNF4 : -catenin/GS(positive: +) HNF4 (+), -catenin/GS(+) HNF4 (negative: -), -catenin/GS(-) HNF4 (+) and -catenin/GS(-) HNF4 (-). We compared the four groups regarding enhancement ratio on hepatobiliary phase of gadoxetic acid-enhanced MRI, immunohistochemical OATP1B3 expression and histological differentiation grade. Kruskal-Wallis test and Steel-Dwass multiple comparison test were used for the statistical analyses. Results: Enhancement ratio in HCC with -catenin/GS(+) HNF4 (+) was signi cantly higher than those of the other 3 groups (median 2.49, 0.53, 0.66, 0.64, in the above-mentioned order, P<0.001). HCC with -catenin/GS(+) HNF4 (+) showed the highest OATP1B3 expression grade (P<0.001) and the highest differentiation grade among the 4 groups (P=0.002). Conclusion: Co-activation of -catenin and HNF4 promotes OATP1B3 expression, and consequently increase enhancement ratio on gadoxetic acidenhanced MRI and differentiation grade in HCC. Using imaging to identify the molecular background of liver lesions (radiogenomics/proteomics) will help future personalized medicine in HCC.

SS 12.8
The added value of diffusion-weighted imaging at 3T MRI in differentiating malignant from benign thrombus of the portal vein E. Guler 1 , E. Ozturk 1 , M. Yuksel 2 , T. Kose 1 , M. Harman 1 , N.Z. Elmas 1 ; 1 Izmir/ TR, 2 Kahramanmaras/ TR Purpose: To determine the utility of diffusion-weighted imaging (DWI) and conventional 3T MRI in differentiating malignant portal vein thrombus (PVT) from benign PVT. Material and methods: A retrospective database search for examinations obtained with 3T MRI including DWI from January 2011 through December 2016 for "PVT" was performed. A thrombus was considered malignant if it enhanced on MRI (≥15% during arterial phase when compared with precontrast images). Twenty-three patients with malignant PVT and 14 patients with benign PVT were identi ed. Two independent reviewers recorded the maximum diameter of portal vein, signal intensity of PVT on T2-weighted images and DWI. The mean apparent diffusion coef cient (ADC) of the malignant and benign PVTs were calculated and compared using Mann-Whitney U test. Comparisons of the signal intensity of the PVTs on T2-weighted images and DWI were evaluated using Pearson Chi-square and Fisher's exact tests. The interobserver aggreement was assessed using the kappa statistic and intraclass correlation. Results: The mean ADC values of malignant and benign PVTs were 0.95±0.19x10 -3 mm 2 /s and 1.87±0.26x10 -3 mm 2 /s, respectively, with signi cant difference (p=0.00). There were signi cant differences between two groups for the comparison of signal intensity on T2-weighted images and DWI (p<0.05). The interobserver correlation for ADC values was 0.95, indicating excellent correlation. Conclusion: DWI enables differentiation between malignant and benign PVT. An interesting nding is that conventional sequences and DWI can predict malignant PVT in the absence of apparent tumor on MRI.
SCIENTIFIC SESSIONS / FRIDAY, JUNE 23, 2017 S619 SS 12.9 Diagnostic performance of intravoxel incoherent motion and conventional diffusion-weighted imaging in the differential diagnosis of benign and malignant portal thrombosis E. Kaya, B. Koyuncu Sokmen, S. Server, A. Oz, N. Inan, C. Balci, Y. Tokat; Istanbul/ TR Purpose: To evaluate the diagnostic accuracy of intravoxel incoherent motion (IVIM) and conventional diffusion-weighted imaging (DWI) parameters in the differential diagnosis of benign and malignant portal vein thromboses (PVT). Material and methods: Twenty-eight patients (18 men,10 women) with PVT (13 benign, 15 malignant) were included in this retrospective study. All patients were examined by 1.5 T MRI with the use of four-channel phased array body coil. In addition to routine pre-and postcontrast sequences, IVIM (16 different b factors of 0-1300s/mm2) and conventional DWI (3 different b factors of 50,400,800s/mm2) were obtained. Two different ADC maps reconstructed from conventional DWI (ADCcon) and IVIM (ADCivim). The mean D (true diffusion coef cent), D* (pseudo-diffusion coef cient associated with blood ow) and f (perfusion fraction) values were also calculated from IVIM. Quantitatively, both ADCcon and ADCivim, D, D* and f values were compared between the groups by Student's t test. To evaluate the diagnostic performance of the parameters, recevier operating characteristic (ROC) analysis was performed. Results: The ADCcon, ADCivim, and D values of the malignant PVT were signi cantly lower than those of benign ones (p=0.011, p=0.008, and p=0.046, respectively). The differences of f and D values were statistically not signicant. The best discriminative parameter was the ADCivim (mean ADCivim was 1,15 ±0,32x10-3 for benign PVT; 0,86 ±0,26x10-3 for malignant PVT We aimed to determine if a new photoacoustic imaging (PAI) system can successfully depict microvascular circulatory changes in response to thermal stimuli. Material and methods: Following ethical permission, 13 consenting subjects underwent PAI of the index ngertip as proof-of-concept. The images were obtained after immersion in either warm (30-35°C) or cold (10-15°C) water to promote vasodilation or vasoconstriction, respectively. The PAI instrument used a Fabry-Perot interferometer as the ultrasound sensor, a 30Hz 750nm excitation laser and a 1550nm interrogation laser. Images were acquired through a 14x14x14mm volume over 90seconds and reconstructed of ine for analysis. Volumetric imaging datasets were evaluated subjectively by two independent radiologists, and quantitatively by voxel-counting, to determine if PAI could depict cold-induced vasoconstriction. We also measured the fullwidth at half-maximum (FWHM) of resolvable vessels. Material and methods: All CT-guided drainages of uid collections between January 2014 and September 2016 in a maximal care hospital were included in this study. The patients were divided into 2 subgroups; patients that underwent surgery before developing the uid collection and patients that did not. Drainages that were additionally ushed with CM were evaluated to assess whether the additional scan was helpful for either detecting the source of the uid collection (e.g. anastomosis insuf ciency) or additional cofactors (e.g. intestinal stulas in pancreatitis) and whether or not this information led to an immediate change of therapy. Results: 499 drainages in 352 patients were detected, 197 thereof were postinterventionally ushed with CM. 51 (26%) of those showed a clinically signi cant additional nding. An immediate change of therapy was found in 19 cases (9%). The group that underwent surgery showed no statistically signicant difference to the group that did not. Conclusion: An additional scan with CM injection via the newly placed drain revealed clinically signi cant additional information in almost 26% in this study. In 9% of the cases this information led to an immediate change of therapy.

SS 13.6
The effect of adaptive statistical iterative reconstruction (ASIR) on CT image heterogeneity analysis in primary colorectal cancer K. Owczarczyk 1 , D. Prezzi 1 , M. Siddique 1 , P. Bassett 2 , C. Grierson 3 , D.J. Breen 3 , G.J. Cook 1 , V. Goh 1 ; 1 London/ GB, 2 Amersham/ GB, 3 Southampton/ GB Purpose: To determine the effect of adaptive statistical iterative reconstruction (ASIR) on the calculation of image heterogeneity features using contrast-enhanced CTs from patients with colorectal cancer. Material and methods: CTs from 32 prospective patients were reconstructed using 6 different ASIR percentages (0-100%) yielding 6 datasets/case. A single-slice region of interest was drawn around the tumor on the ASIR0% scan and propagated onto remaining datasets. First-order (mean, skewness, kurtosis and entropy), second-order (GLCM: contrast, entropy and homogeneity) and fractal (dimension and lacunarity) features were compared using Bland-Altman statistics, repeat-measure ANOVA (rANOVA) and univariate regression (R 2 ). Results: Mean feature variation (expressed as percentage difference) between ASIR levels ranged from 0.13 to 10.14%; in the majority of cases was <5%. ASIR had no signi cant effect on rst-order features (rANOVA p values: 0.29, 0.1, 0,2 and 0.6; R 2 : 0.08, 0.16, 0.11 and 0.03, respectively). Second-order/ fractal features varied signi cantly across ASIR levels (rANOVA p value<0.0001 in all cases) and ASIR had an incremental effect on feature calculation (R 2 : 0.86, 0.75, 0.57, 0.95 and 0.86). Conclusion: Heterogeneity feature variation was overall modest. While ASIR had little/no effect on rst-order (non-texture) features, there was a strong correlation between increasing ASIR percentages and higher order features, suggesting that different reconstruction algorithms should not be used interchangeably in image heterogeneity research.

SS 13.7
Role of perfusion computed tomography in evaluation and comparison of gastrointestinal stromal tumors (GISTs) and GI lymphomas: a pilot study S. Gupta, J. Kumar, V. Chowdhury; New Delhi/ IN Purpose: To assess perfusion characteristics of GIST and GI lymphomas on CT and differentiate and compare CT perfusion parameters with histopathology/immunohistochemistry. Material and methods: 23 adult patients with suspected GIST/GI lymphoma were examined with an initial NCCT of abdomen followed by dynamic CT acquisition of the region of interest using a dedicated perfusion protocol. Postprocessing perfusion maps were created and perfusion parameters: blood ow (BF); blood volume (BV); permeability (PMB) and mean transit time (MTT) of both involved and uninvolved parenchyma were recorded for comparison. Only histopathologically/immunohistochemistry con rmed tumors were used in nal analysis (18/23). Statistical signi cance was calculated using Mann-Whitney U test and p value<0.05 was considered as signi cant. Results: Both GISTs and GI lymphomas; irrespective of the size and location showed signi cantly higher perfusion parameters compared to the baseline (p=0.001). High-grade GISTs (graded histopathologically) showed signi cantly higher BF, PMB & lower MTT (p<0.05) than low-grade GISTs; however, with similar BV. MTT was the only parameter which showed statistical signi cance (p=0.021) in differentiating GISTs & GI lymphomas (reduced more in GI lymphomas; 6.07+/-2.67 vs 11.57+/-4.93s). Mean BF was signi cantly lower in GI lymphomas showing mural thickening >5cm than those showing mural thickening<5cm (75.87+/-31.25 vs 110.24+/-61.54 ml/100ml/min) indicating outgrowth of blood supply by increased mural thickening. Conclusion: Perfusion CT is useful in complementing MDCT in characterizing and differentiating between different grades of these tumors which are close imaging mimics.

SS 13.8
Whole-liver perfusion CT with a 160-mm/256-row scanner CT in cirrhotic patients: preliminary experience R. Faletti, M. Fronda, C. Dianzani, M. Gatti, A. Ferraris, F. Marchisio, P. Fonio, G. Gandini; Turin/ IT Purpose: To analyse the feasibility and the dosimetric aspects of whole-liver perfusion CT (WLpCT) in cirrhotic patients. Material and methods: Eight cirrhotic patients underwent WLp with a 160-mm/256-row CT scanner (GE Healthcare, Revolution CT). Blood ow (BF), time to peak (TTP), blood volume (BV), mean transit time (MTT), hepatic arterial fraction (HAF) and permeability surface area product (PS) were measured by two double-blinded radiologists both in the HCC and in the surrounding parenchyma. The WLp CT dose index (CTDI) and dose-length product (DLP) were compared to those of a standard 4-phase CT performed in a control group of 127 patients. Data were analysed using Wilcoxon signed-rank test. Results: A total of 10 HCC were detected and histologically proven. Five of them appeared hypovascular on a previous examination (3 contrast-enhanced ultrasound and 2 gadoxetic acid-enhanced MRI). HCC had increased HAF and BV when compared with the surrounding parenchyma (respectively, p=0.01 and p=0.02), without signi cant differences in HAF between typical and atypical nodules (p>0.05). The mean DLP of the WLpCT were similar to that of 4-phase CT performed with the same scanner CT (967.5 vs. 1093 mGy-cm; p>0.05). Conclusion: WLpCT is a feasible technique in the evaluation of cirrhotic liver with promising results in the assessment of typical and atypical HCC, without signi cant increase in radiation dose.

SS 13.9
Noise-optimized virtual monoenergetic dual-energy CT angiography can improve diagnostic accuracy for the detection of active arterial abdominal bleeding J.L. Wichmann, S.S. Martin, D. Leithner, T.J. Vogl,

M.H. Albrecht; Frankfurt am Main/ DE
Purpose: To evaluate the diagnostic accuracy of noise-optimized virtual monoenergetic imaging (VMI+) regarding the detection of active arterial abdominal bleeding at dual-energy CT angiography (DE-CTA) in comparison with standard image reconstruction. Material and methods: DE-CTA datasets of 71 patients (46 men; 63.6±13.3 years) with suspected arterial bleeding of the abdomen or pelvis were reconstructed with standard linearly blended (F_0.5), VMI+, and traditional monoenergetic (VMI) algorithms in 10-keV increments from 40 to 100 keV. Attenuation measurements were performed in the descending aorta, area of hemorrhage, and the feeding artery to calculate contrast-to-noise ratios (CNR) in patients with active arterial bleeding. Based on quantitative image quality results, the best series for each reconstruction technique were chosen to separately analyze the diagnostic performance of three blinded radiologists for the detection of active arterial bleeding. Results: Thirty-six patients showed ndings of acute arterial bleeding. Mean CNR was superior in 40-keV VMI+ compared to all VMI series (all P<0.001), which showed highest CNR at 70 keV, as well as standard F_0.5 images (21.6±7.9, 12.9±4.7, and 10.4±3.6, respectively). Area-under-the-curve analysis for detection of arterial bleeding showed signi cantly superior (P<0.001) results for 40-keV VMI+ (0.963) compared to 70-keV VMI (0.775) and F 0.5 image series (0.817). Conclusion: Diagnostic accuracy for the detection of active arterial bleeding in the abdomen can be signi cantly improved using noise-optimized VMI+ reconstructions at 40 keV compared with standard linearly blended and traditional VMI series in DE-CTA.

SS 13.10
withdrawn by the authors SCIENTIFIC SESSIONS / FRIDAY, JUNE 23, 2017 Conclusion: Utilizing advanced analytics, multi-textural data obtained from CT images have the capability to detect genetic aberrations in intrahepatic cholangiocarcinoma, including FGFR2 gene fusion. This is particularly compelling in light of improved treatment options for patients with FGFR2 gene fusions through the use of FDA-approved kinase inhibitors with pan-FGFR activity and/or FGFR2 inhibitory activity. If imaging could be used to identify patients with an increased likelihood of FGFR2 gene fusion, targeted chemotherapy treatment could be more rapidly deployed with the potential to eliminate the need for invasive core biopsies as well as the expense and turnaround time of genomic analysis.

SS 14.2
Biliary leak after percutaneous biliary stenting: prevention with biopsy sealing device T. Biondi, D. Bellini, D. Caruso, M. Rengo, M. Zerunian, A. Saltarelli, A. Laghi; Latina/ IT Purpose: To prospectively evaluate the effectiveness of biliary sealing with a compressed collagen foam plug in preventing the development of complication related to biliary leak following biliary stenting. Material and methods: This single-center, HIPAA-compliant prospective study was approved by our Institutional Review Board. Forty patients with malignant biliary obstruction (16 men, 24 women), with clinical indication for biliary stenting, were randomly assigned in a 1:1 ratio, to receive the biliary sealing device or not. All patients were masked to treatment allocation. The primary endpoint was the prevalence of complications related to perihepatic biliary leak after procedure; hospital stay was also evaluated after procedure and abdominal pain was perceived using a 10-point scale.
Results: The prevalence of postprocedural biliary leak was 83% (15 of 18 patients) in the control group and 5% (1 of 18 patients) in the plug group. None of patients in both groups required abdominal drainage. Hospital stay was longer for the control group compared to plug group (30% increase in hospital stay) and the scores of abdominal pain perceived after procedure were signi cantly higher for the control group (5 ± 2 vs 3± 2; P value 0.032). Conclusion: Although further experience is necessary, transhepatic collagen foam plug placement might be a simple and effective supplement for avoiding biliary leak after percutaneous biliary stenting. Gd-EOB-DTPA) at 1.5T and 3T were retrospectively included from two centers. Signal intensity of uids were classi ed as hypo/iso/hyperintense compared to the splenic parenchyma, before and after injection. The relative signal enhancement (RE) and the signal to noise ratio (SNR) on the pre-and contrastenhanced sequences were calculated. Results: 139 patients with peritoneal/pleural effusions without biliary/vascular leakage (mean 60±10-yo, 96 males, 69%) were included. MR was performed for chronic liver disease (n=105), cancer staging (n=21), and others (n=15). On hepatobiliary phase (HBP) after Gd-BOPTA, peritoneal uid appeared hyper/ isointense in 88-100%, and pleural effusion in 100% of the patients. On HBP after Gd-EOB-DTPA, all effusions remained hypointense. Signal intensity increased with both HBCA over time but RE was signi cantly higher after Gd-BOPTA (p=0.002 and <0.001 for peritoneal and pleural uid). After Gd-BOPTA, it was signi cantly higher in patients with chronic liver disease (p=0.009).
Conclusion: Signal intensity of pleural and peritoneal uid progressively increases after HBCA in the absence of vascular of biliary leakage, regardless of the eld strengh. As most patients explored with Gd-BOPTA had hyperintense uid effusions during HBP, we do not recommend this contrast agent for diagnosing biliary leak.

SS 14.4
Imaging findings in gangrenous cholecystitisretrospective analysis of histopathologically proven cases: largest Indian tertiary care experience B. Sureka, A. Rastogi, A. Mukund, S.T. Laroia, T.K. Chattopadhyay; New Delhi/ IN Purpose: To identify the radiological signs/markers in histopathologically proven cases of gangrenous cholecystitis. Material and methods: A total of 31 patients were evaluated. A retrospective study was done wherein imaging (USG, CT or MRI) data of patients and cases operated at our institute with histopathological evidence of gangrenous cholecystitis were analyzed. For this, the histopathology records of operated cases of acute cholecystitis between January 2012 and August 2016 were searched and cases in which pathology reports mentioned necrosis, transmural in ammation with transmural ulceration were included in the present study. Results: 31 patients (male 13; female 18; mean age 49.2 years, age range 24-74 years) were included in the study. The mean wall thickening of the gallbladder wall was 6 ± 1.93 mm. Gallstones were present in 30 cases. Mural edema, mural striation, pericholecystic uid and intraluminal membranes were seen in 27, 18, 20 and 14 cases, respectively. Pericholecystic stranding was seen in 24, gas in 3. Intraluminal membranes were present in 14 cases with mean short-axis distension of 4.6 cm and absent in 17 cases with mean shortaxis distension of 3.94 cm (p=.041). 11/14 (78.6%) had mural striation (p=.036). All cases with gallbladder wall hemorrhage had intraluminal membranes. Focal decreased wall enhancement was seen in 9/10. Conclusion: Imaging signs and markers like gallbladder distension (≥ 4cm), membranes, mural striation and abnormal wall enhancement suggest high probability of gangrenous change in acute cholecystitis. Purpose: The aim of the study was to analyze the diagnostic role of contrast agent administration in CT in emergency patients with suspected complication after ERCP. Material and methods: From January 2013 till December 2016, ninety-six patients (46 males and 39 females, mean age 66.7 years), who recently underwent ERCP, arrived at emergency department of radiology with clinically suspected complication. All patients were studied with CT. The results were statistically analyzed. Results: In sixty-one patients the CT exam was performed with contrast agent administration, thirty-ve had only the baseline CT. The diagnostic accuracy of contrast-enhanced CT was signi cantly higher than baseline CT (93% vs 73%, P <0.05). Fifty-four patients had normal CT scans except for the presence of ordinary post-procedural ndings such as air and contrast agent in the biliary tree and gallbladder. The presence of complications was found in 42 cases (43.8%). The most common nding was perforation (16.6% of complications), followed by acute pancreatitis (14 cases, 14.5%), infection (8.3%) and gastroduodenal acute bleeding (3.1%). Finally, there was one case of stent misplacement. Conclusion: Although ERCP has a low incidence of adverse events (5% of cases described in the literature), contrast-enhanced CT is the technique of choice for emergency imaging when they are suspected. It allows evaluation of the type and severity of complication, and it is a necessary exam for surgical management and treatment. S625