Machine-learning Algorithm-based Risk Prediction and Screening-detected Prostate Cancer in A Benign Prostate...

Area of Science:

• Urology
• Oncology
• Data Science

Background:

• Prostate cancer (PCa) poses a significant lethal threat.
• Benign prostate hyperplasia (BPH) affects a large patient population.
• Accurate PCa risk prediction in BPH patients is crucial for timely intervention.

Purpose of the Study:

• To predict PCa risk in BPH patients using machine learning.
• To identify key risk factors for PCa development in this cohort.
• To optimize predictive model performance for enhanced clinical utility.

Main Methods:

• Retrospective cohort study utilizing a clinical database (2000-2020).
• Inclusion of BPH patients prescribed specific medications, excluding those with prior cancer diagnoses.
• Application of machine learning algorithms: Extreme Gradient Boosting (XGB), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN).

Main Results:

• Support Vector Machine (SVM) and Extreme Gradient Boosting (XGB) models demonstrated superior accuracy and area under the curve compared to K-Nearest Neighbors (KNN).
• Key predictors identified include Body Mass Index (BMI), late Prostate-Specific Antigen (PSA), and PSA velocity.
• Use of 5-alpha-reductase inhibitors was associated with increased PCa incidence, though survival outcomes were similar.

Conclusions:

• Machine learning offers a promising avenue for personalized PCa risk assessment in BPH patients.
• Further research is needed to refine models and mitigate data biases.
• Clinicians should consider these ML tools as adjuncts to conventional screening methods.