Predicting Prostate Cancer Risk Stratification Based on PSA and Functional Subsets of Peripheral Lymphocyte

Prostate cancer (PCa) is the second most common cancer among men worldwide, but its diagnosis has long grappled with issues of over-and under-treatment. To address this challenge, novel screening tools and liquid biopsy techniques have emerged in recent years. These methods not only enhance risk stratification for prostate cancer but also reduce patient discomfort and risk. Notably, Prostate-specifc antigen (PSA) screening is no longer the sole choice. In this study, we selected functional subsets of peripheral lymphocyte along with PSA levels to predict the three risk stratifications of PCa. We collected clinical data from 197 prostate cancer patients, subjected it to data cleaning, and used lasso feature selection to choose the optimal feature subset. A decision tree model was employed with ten-fold cross-validation to predict low, intermediate, and high-risk stratifications for PCa. The study yielded an AUC of 0.8688, F1 score of 0.7950, sensitivity of 0.7944, and specificity of 0.9093.