Machine Learning-Based Models Enhance the Prediction of Prostate Cancer

Area of Science:

• Urology
• Oncology
• Medical Informatics

Background:

• Prostate-specific antigen (PSA) is a common but imperfect biomarker for prostate cancer screening due to limited specificity.
• Accurate and early diagnosis of prostate cancer is crucial for effective patient management and treatment outcomes.

Purpose of the Study:

• To develop and evaluate machine learning (ML)-based models for enhanced prostate cancer prediction.
• To compare the performance of various ML algorithms against traditional methods for improved diagnostic accuracy.

Main Methods:

• Retrospective analysis of data from 551 patients undergoing prostate biopsy.
• Development of five prediction models using logistic regression (univariate and multivariate), decision tree, random forest, and support vector machine algorithms.
• Evaluation of models using metrics like AUC, accuracy, sensitivity, specificity, calibration curves, and clinical decision curve analysis (DCA).

Main Results:

• All five ML models demonstrated good calibration in the training dataset.
• Random Forest, Decision Tree, and multivariate LR models showed superior discrimination in the training set (AUCs 1.0, 0.922, 0.91).
• The multivariate LR model achieved the best discrimination in the test set (AUC=0.918), with strong generalizability. Four models showed better net clinical benefits than univariate LR.

Conclusions:

• Machine learning techniques significantly enhance prostate cancer prediction capabilities.
• ML models offer improved accuracy, discrimination (AUC), and clinical utility compared to traditional PSA screening.
• The study highlights the potential of ML in improving prostate cancer diagnosis and patient management.