A Novel Machine Learning-Based Predictive Model of Clinically Significant Prostate Cancer and Online Risk Calculator

This recent publication presents a novel machine learning-based predictive model for clinically significant prostate cancer, along with an online risk calculator. The model was developed using a diverse cohort from three countries (New Zealand, Australia, and Switzerland), enhancing its generalizability across populations and clinical settings. Rigorous internal and external validation was conducted, including the use of a Spanish cohort of patients who underwent transrectal ultrasound-guided (TRUS) biopsies. This external validation confirmed the model's applicability for both TRUS and transperineal prostate biopsies (TPPB), underscoring its versatility.

From the three machine learning models tested, Light GBM was selected for developing the predictive model due to its superior performance. The authors focused on creating a balanced, calibrated, and accurate model. A calibration plot was presented, showcasing excellent alignment between predicted and observed probabilities.

Decision curve analysis demonstrated significant clinical utility, with a net benefit in decision-making, including reduced unnecessary biopsies without compromising the detection of clinically significant cancer. Compared to existing models based on linear regression, the Light GBM model demonstrated better calibration and a more balanced performance.

To translate this innovation into practice, the authors created a risk calculator based on this AI/Machine Learning model, aiming to assist clinicians in daily clinical decision-making and improve patient care.

Written by: Flávio Vasconcelos Ordones, MD, MSc, FACS, Urologist-Tauranga Hospital, New Zealand

Read the Abstract

Login