Aussavavirojekul P, Hoonlor A, Srinualnad S, et al. - PIRADS 3 patient selection for MRI/US fusion biopsies is facilitated by machine learning (ML) models, and optimization of previously known clinical risk factors to their full potential could be possible using ML.

• There is a low cancer-detection rate in patients with PIRADS category 3 lesions.

• This retrospective, single-center study was conducted to develop ML models to facilitate decision-making regarding whether to perform prostate biopsies or monitor clinical data without biopsy results.

• Participants included 101 eligible patients with at least one PIRADS category 3 lesion but no higher PIRADS lesions who had MRI/US fusion biopsies; 30 additional patients were enrolled as the validation cohort.

• Age, prostate-specific antigen, prostate volume, prostate-specific antigen density, and the number of previous biopsies were the clinical inputs, and the number of lesions, maximum lesion diameter, lesion location, and lesion zone were the radiology-report inputs.

• Logistic Regression, Support Vector Machine, Naive Bayes, Decision Tree, Random Forest, and eXtreme Gradient Boosting Tree (XGBoost) were used.

• XGBoost afforded the best Area Under the ROC Curve (AUC) of 0.76, in the validation cohort, which considered 80% sensitivity and 72% specificity at a probability cutoff of 57%.

• Worse performance of the rest of the possible ML models was evident with lesser AUC.

• Naïve Bayes was the worst, with AUC of 0.50.