Value of machine learning-based transrectal multimodal ultrasound combined with PSA-related indicators in the diagnosis of clinically significant prostate cancer.

Objective: To investigate the effect of transrectal multimodal ultrasound combined with serum prostate-specific antigen (PSA)-related indicators and machine learning for the diagnosis of clinically significant prostate cancer. Methods: Based on Gleason score of postoperative pathological results, the subjects were divided into clinically significant prostate cancer groups(GS>6)and non-clinically significant prostate cancer groups(GS ≤ 6). The independent risk factors were obtained by univariate logistic analysis. Artificial neural network (ANN), logistic regression (LR), support vector machine (SVM), decision tree (DT), random forest (RF), and K-nearest neighbor (KNN) machine learning models were combined with clinically significant prostate cancer risk factors to establish the machine learning model, calculate the model evaluation indicators, construct the receiver operating characteristic curve (ROC), and calculate the area under the curve (AUC). Results: Independent risk factor items (P< 0.05) were entered into the machine learning model. A comparison of the evaluation indicators of the model and the area under the ROC curve showed the ANN model to be best at predicting clinically significant prostate cancer, with a sensitivity of 80%, specificity of 88.6%, F1 score of 0.897, and the AUC was 0.855. Conclusion: Establishing a machine learning model by rectal multimodal ultrasound and combining it with PSA-related indicators has definite application value in predicting clinically significant prostate cancer.