Mp74-10 AI-Based Risk-Stratification Tool Pathomiq_prad Analyzes Histopathology Images to Accurately Predict Metastasis in Black Men and Helps Discover Cancer Drivers

You have accessJournal of UrologyProstate Cancer: Detection & Screening VI (MP74), Moderated Poster 741 May 2024MP74-10 AI-BASED RISK-STRATIFICATION TOOL PATHOMIQ_PRAD ANALYZES HISTOPATHOLOGY IMAGES TO ACCURATELY PREDICT METASTASIS IN BLACK MEN AND HELPS DISCOVER CANCER DRIVERS Dimple Chakravarty, Hassan Muhammad, Chensu Xie, Parag Jain, Rachel Brody, Lawrence Leung, Monali Fatterpekar, Wei Huang, Swati Bhardwaj, Hirak Basu, George Wilding, Rajat Roy, Sujit S. Nair, and Ashutosh K. Tewari Dimple ChakravartyDimple Chakravarty , Hassan MuhammadHassan Muhammad , Chensu XieChensu Xie , Parag JainParag Jain , Rachel BrodyRachel Brody , Lawrence LeungLawrence Leung , Monali FatterpekarMonali Fatterpekar , Wei HuangWei Huang , Swati BhardwajSwati Bhardwaj , Hirak BasuHirak Basu , George WildingGeorge Wilding , Rajat RoyRajat Roy , Sujit S. NairSujit S. Nair , and Ashutosh K. TewariAshutosh K. Tewari View All Author Informationhttps://doi.org/10.1097/01.JU.0001008632.59099.b9.10AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Black men have a greater incidence and mortality rate of prostate cancer (PCa) than white men. There is an unmet need for risk-stratification tools for early detection of lethal phenotypes for all patients and black men in particular. In recent years, artificial intelligence (AI) has shown great promise in signature extraction for the prediction of cancer progression; however, AI models underperform on patients from underprivileged groups. PATHOMIQ_PRAD is a previously validated AI-based tool that identifies clinically significant histological patterns from whole slide images (WSIs) to predict patient outcome. Here, we show that PATHOMIQ_PRAD performs equally well with black men as it does with white men. METHODS: 215 radical prostatectomy samples were collected from the Icahn School of Medicine at Mount Sinai to form a held-out test set. Each sample included a digitized H&E WSI, CAPRA-S score, and retrospective time-to-metastasis (MET). A subset of 147 patients also had an associated genomic score. Racial stratification of the test set was 40.7% (n=87) Black, 51.4% (n=110) White, and 7.9% (n=17) other. PATHOMIQ_PRAD was used to produce a risk score for each sample by analyzing the associated WSI. Concordance index (CI) was used to measure the performance of predicting MET. Hazard ratios (HR) for patients in each racial group were measured by stratifying between high and low risk, as determined by PATHOMIQ_PRAD. ROIs were analyzed by spatial transcriptomics using NanoString GeoMx DSP WTA platform. RESULTS: PATHOMIQ_PRAD predicted MET with a CI of 0.782 on the whole test set, 0.879 on the Black patient subset, and 0.769 on the White patient subset. HR for Black patients was 7.977 (p<0.01); White patients was 7.921 (p<0.01). On the subset of samples with an associated genomic score and CAPRA-Score, PATHOMIQ_PRAD outperformed both methods in predicting MET (CI: PATHOMIQ_PRAD, 0.850; CAPRA-S, 0.669; Genomic score, 0.708). ROI guided GeoMx RNA and Protein Assay (NanoString Technologies) identified known PCa biomarkers including androgen receptor (AR) target genes in the high-risk ROIs. CONCLUSIONS: PATHOMIQ_PRAD successfully predicted MET in patients with prostate cancer, with better performance in black patients. PATHOMIQ_PRAD significantly outperformed the genomic score and CAPRA-S in predicting MET in PCa patients. Further, the identification of molecular drivers of lethal phenotypes may lead to novel therapeutic targets against aggressive PCa in black men. Source of Funding: The Arthur M. Blank Family Foundation © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e1196 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Dimple Chakravarty More articles by this author Hassan Muhammad More articles by this author Chensu Xie More articles by this author Parag Jain More articles by this author Rachel Brody More articles by this author Lawrence Leung More articles by this author Monali Fatterpekar More articles by this author Wei Huang More articles by this author Swati Bhardwaj More articles by this author Hirak Basu More articles by this author George Wilding More articles by this author Rajat Roy More articles by this author Sujit S. Nair More articles by this author Ashutosh K. Tewari More articles by this author Expand All Advertisement PDF downloadLoading ...