Machine learning methods for predicting early recurrence in Ta stage bladder cancer and comparison with conventional statistical methods

BACKGROUND: Predicting whether recurrence will occur during follow-up in bladder cancer cases confined to the mucosa (Ta-stage) is one of the crucial aspects of management. In this study, we aimed to compare the models obtained by conventional statistical methods and machine learning (ML) methods in order to predict the development of recurrence in the 2-year postoperative period in Ta stage Non-Muscle Invasive Bladder Cancers. METHODS: The data of patients who underwent complete transurethral resection of the bladder and were found to have Ta pathologies due to primary bladder cancer between 2018-2021 was retrospectively screened. Patients with no recurrence during the two-year follow-up were classified as Group 1 (N.=107, 58.2%), and those with recurrence were classified as Group 2 (N.=77, 41.8%). The demographic, clinical, imaging and pathological data were recorded. These parameters were analyzed using a conventional statistical method and ML methods to construct prediction models. RESULTS: Body Mass Index, American Society of Anesthesiologists (ASA) score, and the presence of macroscopic hematuria were found to be significant variables to predict early recurrence (P