Diagnostic Model Research Of Prostate Cancer Based On Machine Learning Algorithm

Objective Establish diagnostic prediction models based on three machine learning algorithms, and compare the diagnostic value of the three models for prostate cancer (PC). Methods Retrospectively analyze the clinical data of 956 patients (463 cases of prostate cancer and 493 cases of benign prostatic hyperplasia) with prostate biopsy in the General Hospital of PLA during 2008-2014, and screen predictors by Logistic regression which include age, free prostate-specific antigen (free PSA, fPSA), the percentage of free prostate-specific antigen (free/total PSA, f/tPSA), prostate volume and PSA density. Further compare the diagnostic performance of three models for prostate cancer, using BP neural network, Logistic regression(LR) and random forest algorithm, based on machine learning. Results The diagnostic capability of Logistic regression, BP neural networks and random forest model for prostate cancer is higher than any a single indicator. The sensitivity, specificity, and accuracy of LR model were 77.5%,74.8%, and 76% versus 77.4%,76.8%, and 77% for BP neural networks model versus 76.2%,76,9%,77% random forest model. The area under the ROC curve (AUC) is 0.831 for LR model,0.832 for BP neural networks model and 0.833 for the random forest model, respectively, which shows that three models have no statistically significant difference. Conclusion The above results verified the high diagnostic validity of these models. Three models can be incorporated into urologic decision making to assist the clinician making diagnosis and treatment, to reduce the unnecessary biopsies.