ECP 2023 Abstracts

S3 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 were associated with the peritumoral budding category in 70% of the tumours. The complete data will be presented at the congress. Conclusion: ITB can be reliably scored on biopsies, based on the pres- ence of TB in the superficial portion of the tumour, provided that suf- ficient tumour is captured in those biopsies. OFP-01-007 A real-time histologic evaluation of gastric cancer tissue by using artificial intelligence-assisted confocal laser endomicroscopic system S. Kim*, H. Bae, H. Cho, J. Chu, S. Choi, S.M. Heo, K. Hwang, Y. Jo, K. Kim, D. Lee *Ajou University School of Medicine, Republic of Korea Background & objectives: The processing time and preparation arti- facts of frozen sections demand a novel technique for the instant his- tologic evaluation. A confocal laser endomicroscopic system provides real-time imaging of the tissue without glass slide preparation. We evaluated its applicability in gastric cancer. Methods: A confocal laser endomicroscopic system using light laser scanning was developed. Forty-five advanced gastric cancer tissues and the normal tissues were obtained from fourteen patients. Five patholo- gists interpreted 200 images and detected histologic locations and the presence of cancer. An artificial intelligence (AI)-based tumour detec- tion algorithm was developed with 1,305 tumour and 1,137 normal tissue images, by using EfficientNet V2. Results: Pathologists initially detect the histologic location of the images with 65.7% accuracy and differentiate cancer tissue from normal with 74.7% accuracy. The sensitivity and specificity of cancer detection were 71.9% and 76.1%. The interpretation time per image was 12.0 seconds. Following the review of matched H&E images, the accuracy of identify- ing the histologic location was increased to 92.8% (P<0.0001), and that of detecting cancer tissue was also increased to 90.9% (P<0.001). The sensitiv- ity and specificity of cancer detection were enhanced to 89.1% and 93.2% (P<0.0001). The interpretation time per image was decreased to 5.3 seconds (P<0.001). The AI algorithm revealed 94% accuracy, 96% sensitivity, and 93% specificity for detecting cancer. Conclusion: High-quality histologic images were immediately acquired by the confocal laser endomicroscopic system. The trained pathologists could accurately detect cancers and histologic location and the developed AI algorithm showed high performance in detecting cancer, raising its potential applicability as a real-time tissue imaging modality. Funding: Korea Medical Device Development Fund (RS-2022-00140721) OFP-01-008 Characteristics of a modern, screened UICC stage II colon cancer cohort M. Pihlmann Kristensen*, U. Korsgaard, I. Zlobec, T. Frøstrup Hansen, H. Hager, S. Kjær-Frifeldt *Department of Pathology, Lillebaelt Hospital, University Hospital of Southern Denmark, Denmark Background & objectives: Current high-risk factors used for select- ing stage II colon cancer patients to adjuvant chemotherapy are based on outdated cohorts. The aim of this study was to establish a modern cohort of stage II colon cancer patients, post-screening. Methods: A cohort consisting of resected colon cancers from 2014-2016 in the Region of Southern Denmark was collected retrospectively using the Danish Colorectal Cancer Group database and the Danish Pathology Registry. All diagnostic slides were reviewed, and the tumours re-staged according to the UICC TNM Classification, 8th edition. A comprehen- sive histopathological characterization was performed, and clinical data extracted from medical journals. Results: In total, 497 patients met the inclusion criteria of which 20% were diagnosed through the Danish colorectal cancer screening. Adjuvant chemotherapy was administered to 14 % of the patients. Sufficient lymph node sampling (≥ 12) was obtained in 98% of the tumours and lymphatic, venous and perineural invasion was con- firmed in 5, 23 and 12% of the tumours respectively. A test of prog- nostic biomarkers revealed a high stroma content (>50%) in 52% of tumours and a distribution of tumour budding as low, intermediate and high in 53, 22 and 25% of the tumours respectively. Local or distant relapse was determined in barely 9% of patients. The survival analyses are ongoing. Conclusion: This screened cohort of stage II colon cancers is thoroughly characterized and brought up to current pathological evaluation. The relapse rate is notable lower than in other cohorts reflecting the contemporary approach in this cohort. Consequently, we propose that it is highly relevant to examine high risk factors in modern cohorts in order to select the right patients for adjuvant chemotherapy. Funding: Lillebaelt Hospital Region of Southern Denmark OFP-01-009 A new algorithm for coeliac disease based on gamma delta intraep- ithelial lymphocytes detected with an antibody working on FFPE sections E.N. Kozan*, C. Tuna Kirsaclioglu, A. Kansu, A.I. Soykan, D. Gok- men, B. Savaş, A. Ensari *Ankara University Medical School, Pathology Department, Turkey Background & objectives: Diagnosis of coeliac disease (CD) with mild mucosal changes is difficult for all parties involved. We aimed to determine the power of γδ+ IELs in discriminating CD from other causes of intraepithelial lymphocytosis using a new monoclonal antibody. Methods: A total of 167 cases comprising coeliacs with 117 active CD (29 type 1, 29 type 2, 39 type 3) and 20 treated CD, and non- coeliacs including 24 controls and 26 non-coeliac IELosis were studied. IEL counts were evaluated on H&E, CD3, TCRδ-stained (with moAb H-41) sections. Morphometric features were evalu- ated on digitalized images. Discriminant analysis was applied for statistics. Results: γδ+ IELs were significantly higher in CD (24.83±16.13) compared to non-CD (6.72±6.32) and were positively correlated with the degree of mucosal damage (type 1=20.41±13.57, type 2=24.24±12.65, type 3=33.69±19.15). Both γδ+ IEL count and γδ+/CD3+IEL ratio showed higher performance in differentiating coeliacs from non-coeliacs with a sensitivity of 89,69; 84,62 and specificity of 87,5; 76,00, respectively. γδ+ IEL counts also differ- entiated groups showing normal villus/crypt architecture including type 1 CD (20.41±13.57) and non-coeliac IELosis (9.42±7.28) with great significance (p=0.001). Discriminant analysis revealed that villus/crypt ratio, IEL distribution pattern, γδ+/CD3+IEL ratio were the most potent discriminants. Discriminant function correctly clas- sified 82.3% of study cases; 97.5% of new cases and was therefore validated. Conclusion: The new antibody detecting γδ+IELs in FFPE sections allowed us to determine thresholds to discriminate coeliacs from non-coeliacs. Using 10.5 for γδ+ IELs and 14% for γδ+/CD3+IEL ratio proved highly sensitive and specific in correctly classifying cases with normal villus/crypt axis as Type1 CD, non-CD IELosis and controls. γδ+ IEL counts also correlated with Marsh classifica- tion. We propose a ‘histopathological algorithm’ based on γδ+IELs and hope that it will be used as a simple diagnostic test by our fellow pathologists.