ECP 2023 Abstracts

S42 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 OFP-10-015 An international multi-institutional validation study of deep learn- ing-based classifier for prostate cancer detection and Gleason grad- ing in biopsy samples Y. Tolkach*, V. Ovtcharov, A. Pryalukhin, M. Eich, N. Gaisa, M. Braun, W. Hulla, G. Netto, P. Caie, R. Büttner *University Hospital Cologne, Germany Background & objectives: Pathologic examination of prostate biop- sies is time consuming due to the large number of slides per case. In this study, we develop and validate a deep learning-based classifier for prostate cancer (PCA) detection and Gleason grading in biopsy samples. Methods: Five external cohorts of patients with prostate biopsy were analysed. A total of 5922 H&E sections (7473 biopsy cores, 423 patient cases) were assessed concerning tumour detection. Whole Slide Images (WSI) were digitized using three scanners. Two datasets (core n=227 and 159) were graded by international group of pathologists including expert urologic pathologists (n=11) to validate the Gleason Grading classifier. Results: The sensitivity, specificity, and NPV for the detection of tumour-bearing biopsy slides was in a range of 0.971-1.000, 0.875- 0.976, and 0.988-1.000, respectively, across the different test cohorts. In several biopsy slides tumour tissue was correctly detected by the AI tool that was missed by pathologists during initial evaluation. Most false positive misclassifications represented lesions suspicious for carcinoma or cancer mimickers. The quadratically weighted kappa levels for Gleason grading agreement for single pathologists was 0.62-0.80 (0.77 for AI tool) and 0.64-0.76 (0.72 for AI tool) for the two grading datasets, respectively. In cases where consensus for grading was reached among pathologists, kappa levels for AI tool were 0.903 and 0.855. Conclusion: The PCA detection classifier showed high accuracy for PCA detection in biopsy cases during external validation, independent of the institute, scanner or magnification used. High levels of agree- ment for Gleason grading were indistinguishable between experienced genitourinary pathologists and the AI tool. OFP-11 | Joint Oral Free Paper Session Neuropathology / Ophthalmic Pathology OFP-11-001 Use of immunohistochemical surrogate biomarkers as an alterna- tive to molecular methods in classifying adult-type diffuse gliomas F. Gundogdu*, B. Babaoglu, F. Soylemezoglu *Hacettepe University Department of Pathology, Turkey Background & objectives: Nowadays, the use of difficult-to-access and expensive molecular methods has become mandatory in the diag- nosis and grading of IDH-mutant diffuse gliomas. In this study, we aimed to find immunohistochemical (IHC) surrogate biomarkers that could be an alternative to molecular methods. Methods: Tissue microarrays were prepared using 93 astrocytoma and 73 oligodendroglioma cases diagnosed between 2014-2021. 1p/19q and CDKN2A FISH, and H3K27me3 and MTAP IHC studies were per- formed. The stainings were evaluated independently by two observers. The relationship between CDKN2A HD and loss of MTAP expression and the relationship between 1p/19q codeletion and loss of H3K27me3 expression were investigated for two observers. Results: CDKN2A HD was detected in 25 (15.1%) of 161 diffuse glio- mas, while the first observer detected MTAP loss in 30 tumours (18.1%), and the second observer in 29 (17.5%). As a surrogate marker for CDKN2A FISH, the sensitivity of MTAP was 88% for both observers, and the specificity was 95.52%-96.27% respectively. Loss of H3K27me3 expression was observed in 67.1-68.5% of oligodendrogliomas, and 21.1-22.6% of astrocytomas. As an alternative to 1p/19q FISH, the sensitivity of H3K27me3 was 77.42-76.34%, and the specificity was 65.75%-67.12%. In both biomarkers, an almost perfect agreement with a 0.938 Kappa coefficient was recorded among observers. Conclusion: MTAP IHC appears to be a reliable surrogate biomarker with high sensitivity and specificity as an alternative to CDKN2A FISH. However, the loss of H3K27me3 expression is not sensitive and specific enough as an alternative to 1p/19q FISH although it is more frequently seen in oligodendroglioma. OFP-11-002 Prognostic value of the Ki-67 proliferative index in paediatric medulloblastoma, in relation to histology and molecular subgroups M. Al-Hussaini*, A. Abu Shanab, J. Amarin, A. Al-Ani, N. Amayiri *KHCC, Jordan Background & objectives: A high proliferative index, as measured by Ki-67, has been shown to predict poor prognosis in adult medulloblastoma. This study explores the prognostic significance of Ki-67 proliferative index in paediatric medulloblastoma and its relation to molecular subgroups. Methods: Medulloblastoma specimens from paediatric patients (3 – 18 years) were stained with Ki-67/ MIB-1 immunostain. At a mean cut-off of 30.0%, we correlated the Ki-67 proliferative index with histology and molecular subgroups. Kaplan Meier analysis and Cox proportional hazard analysis were conducted to examine the relationship between Ki-67 prolif- erative index and clinicopathological characteristics including molecular subgroups. Results: We included 85 patients (mean age of 7.7 years). There were 17 (20.0%) WNT-activated, 21 (24.7%) SHH-activated, 20 (23.5%) group-3, and 27 (31.8%) group-4 tumours. There were 45 (52.9%) classic, 17 (20%) desmoplastic/ nodular, and 23 (27.1%) large cell/ anaplastic medulloblastoma. Mean Ki-67 score correlated with molecular subgroups (p-value = 0.016). Group-4 has the low- est mean Ki-67 score. It also correlated with histology (p-value = 0.006). Large cell/anaplastic has the highest mean Ki-67 score. The 5-year OS and PFS rates were 70.2% (SE, 5.5%) and 67.9% (SE, 5.8%), respectively. There was no significant survival difference between patients stratified by Ki67 mean score (PFS: p-value = 0.529; OS: p-value = 0.493). Conclusion: Unlike its adult counterpart, the Ki-67 proliferative index mean score failed to be associated with survival outcomes in paediatric medulloblastoma. However, Ki-67 proliferative index correlated with histology and molecular subgroups. Funding: Intra-mural grant from King Hussein Cancer Center OFP-11-003 Colorectal cancer brain metastases, the use of immunohistochemistry markers in the assessment of prognosis J. Gama*, R. Caetano Oliveira, P. Teixeira, F. Silva, Â. Jesus, O. Rebelo, M.A. Cipriano *Centro Hospitalar e Universitário, Portugal Background & objectives: Brain metastases in colorectal cancer patients are common and are associated with a dismal prognosis. Immunohistochemistry assessment of biomarkers may predict prog- nosis and guide therapy. Our goal was to investigate the role of different immunohistochemistry makers in prognosis. Methods: A retrospective transversal study of 77 patients (diagnosed between 2001-2016) with colorectal cancer brain metastases was car- ried out using archival biological material. Expression of CD44, HER2, MLH1, MSH2, MSH6, PMS2, BerEp4, EGFR, NTRK, HIF1-alfa and PD-L1 was assessed by immunohistochemistry. Clinical and pathological data was retrieved from the hospital database. The local ethical committee approved this study. Results: Of the 77 patients (44 male and 33 female) enrolled in the study, the median age at diagnosis was 65 years (33-84). After a median follow-up