Abstract
ABSTRACT
Background and Aims
Prostate cancer affects many people around the world and is fatal. Today, artificial intelligence‐based techniques are widely used to predict and diagnose diseases. This study aimed to use machine learning‐based techniques to predict and identify the most significant risk factors of prostate cancer.
Methods
This retrospective study was conducted at the Shahid Beheshti Hospital, Hamadan Province, Iran, in five steps, including the identification of risk factors, data collection, preprocessing, modeling, and evaluation. The Web of Science, Scopus, Medline, and PubMed Central databases were searched from the beginning of 2000 to 2024. Different prediction algorithms, including Logistic Regression, Gradient Boosting, Random Forest, XGBoost, Support Vector Machine, and Neural Networks algorithms developed based on 49 confirmed factors. 597 medical records were reviewed. 342 patients (57.29%) had prostate cancer, and 255 patients (42.71%) had benign prostate hyperplasia. The algorithm's performance was evaluated using accuracy, precision, recall, F1‐score, and support metrics.
Results
The results of this study showed the XG Boosting model performed best in predicting the effective factors, with an accuracy of 77.5% a sensitivity of 74.5%, and a specificity of 79.7%. Key predictors included total and free prostate‐specific antigen (PSA) levels, hemoglobin, BMI, and fish consumption.
Conclusion
Machine learning can enhance and personalize the identification of potential risks of PCa, but more research is necessary to refine these algorithms and address data biases.