Pd36-03 Artificial Intelligence in Prostate Cancer: The Potential of Machine Learning Models and Neural Networks to Predict Biochemical Recurrence After Robot-Assisted Radical Prostatectomy

You have accessJournal of UrologySurgical Technology & Simulation: Artificial Intelligence III (PD36)1 May 2024PD36-03 ARTIFICIAL INTELLIGENCE IN PROSTATE CANCER: THE POTENTIAL OF MACHINE LEARNING MODELS AND NEURAL NETWORKS TO PREDICT BIOCHEMICAL RECURRENCE AFTER ROBOT-ASSISTED RADICAL PROSTATECTOMY Gurpremjit Singh, Mayank Agrawal, Gopal Sharma, Puneet Ahluwalia, and Gagan Gautam Gurpremjit SinghGurpremjit Singh , Mayank AgrawalMayank Agrawal , Gopal SharmaGopal Sharma , Puneet AhluwaliaPuneet Ahluwalia , and Gagan GautamGagan Gautam View All Author Informationhttps://doi.org/10.1097/01.JU.0001008916.72488.6a.03AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Machine learning (ML) and Neural networks (NN) are a subdivision of artificial Intelligence (AI). It enables the development of novel and user-friendly algorithms by uncovering complex relationships and patterns that traditional statistical methods cannot reveal, mainly when dealing with extensive databases. To compare the classical statistical methods with Machine learning (ML) and Neural networks (NN) models for estimating BCR in patients after Robot-Assisted Radical Prostatectomy. METHODS: A retrospective evaluation of the prospectively maintained database was conducted on patients who received robot-assisted radical prostatectomy from November 2011 to July 2022. Patients who did not exhibit biochemical recurrence (BCR) were allocated to Group 1, whereas those identified with BCR were assigned to Group 2. The demographic information of the patient, as well as their preoperative and postoperative parameters, were all documented in the database. This study utilized two distinct machine learning techniques, K nearest neighbor and XGboost machine learning models and one neural network - Radial basis function neural network (RBFNN) to predict BCR. RESULTS: Data from 916 patients were assessed, and after exclusion criteria, 516 patients were found suitable for the study. Among these, 282 (54.7%) did not have BCR, and 234 (45.3%) developed BCR. The median duration of follow-up and the diagnosis of BCR were calculated as 24 (15-42) months and 12.23 +/- 15.58 months, respectively. The Cox proportional hazard analysis had an area under the curve (AUC) of 0.77. The AUCs for receiver-operating characteristic curves for K nearest neighbor and XGBoost models were 0.69 and 0.82, respectively. The AUC of the Radial basis function neural network was 0.82. CONCLUSIONS: We showed that XGBoost and RBFNN models better predict BCR than classical statistical models. It is suggested that developing more robust and effective models will offer physicians and patients several benefits, including improved accuracy in risk assessment, estimation of prognosis, early intervention, avoidance of unneeded treatments, and reduced morbidity and mortality rates. Download PPT Source of Funding: None © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e793 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Gurpremjit Singh More articles by this author Mayank Agrawal More articles by this author Gopal Sharma More articles by this author Puneet Ahluwalia More articles by this author Gagan Gautam More articles by this author Expand All Advertisement PDF downloadLoading ...