Optimizing Prostate Cancer Management: A Review on the Diagnostic Accuracy and Health Disparities using Machine Learning

Abstract

Prostate cancer requires diagnostic tools that transcend conventional clinical paradigms. This systematic literature review aggregates primary research articles from 2021 to 2025 on the application of machine learning in prostate cancer management. We specifically selected studies on diagnostic accuracy and health disparities while excluding secondary sources. Overall, the results indicate a promising trend in multimodal fusion and explainable machine learning. All models had an area under the curve ranging from 0.84 to 0.91, which was either comparable to or even surpassing the performance of human experts such as radiologists and pathologists. Notably, the in-tegrated models had a positive impact on biopsy specificity, while AI-based pathology had a kappa statistic of above 0.90 in Gleason grading. One of the issues, however, remains how to ensure the generalizability of the models to different racial and geographic populations. Overall, machine learning significantly enhances the accuracy of diagnosis and the efficiency of management. Therefore, it can be said that while machine learning significantly enhances the accuracy of diagnosis and management of prostate cancer, its effective application remains a matter of prospective validation of models and fairness in machine learning.