ECP 2023 Abstracts

S139 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 Results: We identified eight transcriptionally different stromal popula- tions: c0-c7. Three were common to all models: c0 expressed contrac- tile marker genes; c1 displayed Sfrp1, Gpx3, and complement system- related genes expression, and c2 JNK pathway-related genes. Five were specifically linked to specific epithelial mutations: c3 and c4, enriched in T-ERG, Hi-MYC, and NP models, expressed WNT path- way-related genes; c5, c6, and c7, enriched in PRN model, revealed high levels of proliferative markers, collagen and metalloproteinase genes, adaptive immune responses, TGF-β and WNT pathways- related genes. These clusters extensively map in human primary and metastatic samples: c3 was the most conserved one, whereas stromal transcriptional profiles of PRN model and human metastases were strikingly similar. Conclusion: We provided clear molecular evidence of stromal changes during PCa carcinogenesis in mouse models, driven in a genotype-spe- cific manner, which can be extended to humans. We demonstrated how TMPRSS2-ERG fusion reprograms the mesenchyme in the early phases of PCa; while, in advanced PCa models, transcriptional mesenchymal programs are associated with metastatic potential, and similar to the human bone microenvironment ones. Overall, these findings support mesenchymal changes as major contributors to PCa carcinogenesis and phenotypic heterogeneity, to a previously unrecognized level. This work was partially supported by the National Cancer Institute (Prostate Cancer SPORE Grant P50CA211024 [M.L., H.P., R.C., M.O., S.R., C.R.], P01 CA265768-01 [M.L., H.P., R.C., M.O., S.R., C.R.], R01CA200859 [L.M.]), by an AIRC Fellowship for Abroad “Ezio, Maria e Bianca Panciera” [F.P.] and by a Junior Assistant Professor/ Temporary Researcher (RTDA) contract - Italian Ministry of University and Research - PON "Research and Innovation" 2014-2020 (PON R&I) Actions IV. 4 "Doctorates and research contracts on innovation topics" [G.N.F.]. PS-26-009 Using AI-powered solution in clinical practice for primary diagnosis of prostate biopsies R. Grobholz*, P. Nahm, L. McDuffus, D. Mevorach, M. Vecsler *Medical Faculty, University of Zurich, Institute of Pathology, Cantonal Hospital Aarau, Switzerland Background & objectives: We aimed to test the accuracy of an artifi- cial intelligence (AI)-based solution for primary diagnosis of prostate carcinoma using real-world data. Emphasis was given on the accuracy of detection of Gleason patterns and assigning to the Gleason grade groups (GG). Methods: The project included validation of the AI solution in clini- cal practice prior to deployment. One pathologist used the solution for primary diagnosis of 91 prostate biopsies, reporting on 678 parts (1,563 H&E slides), 19 cases with GG1, 15 with GG2, 16 with GG3, 22 with GG4 and 19 with GG5. Accuracy for adenocarcinoma detection and Gleason grading was assessed. Results: The AI solution agreed with the ground truth (GT) in 100% of cases for carcinoma, and generated identical or +1/-1 GG to the GT for 98.7% of cases (n=81). The highest levels of agreement were for GG 1 and 5 (100% and 90%, respectively). For GG 2-4, the AI find- ings were identical or +1/-1 GG to the GT for 100% of cases but had higher disagreement rates with respect to the amount of pattern 4. For GG2 and GG3 concordance rate was 57.1% and 26.6%, respectively. Part-level analysis demonstrated high accuracy, with an AUC of 0.992 (95% CI: 0.9857-0.999), NPV of 98.5% and PPV of 98.7% for carci- noma detection. Conclusion: This study demonstrates the successful validation of a multi feature AI solution that aids to accurately detect all cancer areas and assign an accurate Gleason score. However, detection and quantify- ing of Gleason pattern 4 remains challenging, and AI can aid to detect and quantify the amount of this pattern in order to accurately assign the respective GG. PS-26-010 Assessment of tumour infiltrating lymphocytes in urothelial carci- nomas of the bladder and their prognostic value E. Haddad*, A. Blel, H. Ayed, S. Rammeh *CHU ROUEN, France Background & objectives: Analysis of tumour infiltrating lympho- cytes (TILs) represents a new immunological marker considered as a prognostic factor in several solid tumours. The aim of our work was to evaluate the level of TILs in muscle-invasive bladder UC and their prognostic value. Methods: Retrospective and descriptive study including 74 Urothelial Carcinoma collected between 2010 and 2017. TILs were evaluated based on the recommendations of the International TILs Working Group. A rate of TILs <10% was considered low and a rate ≥10% was considered high. The prognostic value of TILs was assessed. The correlation between clinical-pathological parameters and the rate of TILs was evaluated. Results: A rate of TILs≥10% was observed in 58 cases (78%). The mean follow-up period was 19 months. Mean overall survival was 42 months and progression-free survival was 57 months. Aggressive histologic variants, high tumour stage, perineural neoplastic invasion (p=0.04), tumour surgical limits, lymph node invasion, and a TILs rate of less than 10% were prognostic factors for overall survival. In multi- variate analysis, rate of TILs, tumour stage, metastasis were independ- ent factors of overall survival. The rate of TILs was not a prognostic factor in progression-free survival. A low rate of TILs was correlated with the presence of aggressive histological variants (p=0.02), high tumour stage (p=0.02) and tumour surgical limitations (p=0.02). Conclusion: Our results confirm that TILs in bladder UC are a prog- nostic factor in overall survival. We recommend the inclusion of this factor among prognostic markers in anatomo-pathological reports. PS-26-011 Assessment of positive margins on prostatectomy specimens using fluorescence confocal microscopy: the LaserSAFE technique A. Haider*, R. Almeida-Magana, A. Mathew, T. Al-Hammouri, K. Dinneen, G. Shaw *Department of Histopathology, University College London Hospital, United Kingdom Background & objectives: Frozen section of the posterolateral pro- static margins (NeuroSAFE) has been described as a safe method to evaluate positive surgical margins (PSM) during radical prostatectomy to guide nerve-sparing decisions during surgery. However, its associ- ated time and cost requirements prevent widespread adoption. Methods: We developed a standardised fluorescence confocal micros- copy (FCM) method to analyse margins of prostatectomy specimens. The sample cohort included RP specimens from April 2022 to Febru- ary 2023 from consented patients. Intact specimens were dipped in a photoreactive solution and placed on the microscope for en-face imag- ing. Specimens were then processed using standard paraffin analysis (PA). FCM images were analysed retrospectively. Results: Thirty-one patients were included. It took 1 operator less than 10 minutes to produce images of both posterolateral surfaces of the prostate. 60 images were acquired. 59 images had good quality and contrast, and 56 had >90% of the specimen surface analysable. Eight positive surgical margins were identified on both FCM and PA with a median length of 4 mm. We found one false positive (benign glands on PA) and one false negative on FCM. Sensitivity was 87.5% (CI 86.4%- 88.6%), specificity was 98.1% (CI 97.6%-98.5%), and Cohen’s kappa agreement was 0.86. No artefacts were found on PA attributed to the LaserSAFE technique. Conclusion: The LaserSAFE technique is easy to perform and dem- onstrates high accuracy and very good agreement with PA. These