Constructing a Prediction Model for Clinically Significant Prostate Cancer Combined with Radiomics Features of MRI and PRKY Promoter Methylation Level in Urine Samples

Abstract Purpose To enhance the diagnostic value of MRI for clinically significant prostate cancer (csPCa) and optimize the diagnostic process of prostate cancer (PCa), we developed a machine learning-based prediction model for csPCa combined with MRI features, clinical data, and PRKY promoter methylation level in urine samples. Methods Thirty-nine patients who underwent prostate biopsy or transurethral laser enucleation of the prostate from 2022 to 2023 were selected for this study, and their clinical data and MRI images were obtained before the operation. The urine samples of these patients were collected after prostate massage. Methylation level of two PRKY promoter sites, cg05618150 and cg05163709, were tested through Methylation-Specific PCR. The PI-RADS score of each patient was estimated and the region of interest (ROI) was delineated. After being extracted by a plug-in of 3D-slicer, radiomics features were selected through LASSO regression and t-test. Selected radiomic features, methylation levels and clinical data were used for model construction through the random forest (RF) algorithm in Python. The model based on the PI-RADS score was also constructed for comparison with the radiomics model. The predictive efficiency of each model was analyzed by the area under the receiver operation characteristic (ROC) curve (AUC), and all the models have gone through 3-fold cross-validation. Results Methylation level of cg05163709 in csPCa patients was higher than that in clinically insignificant PCa and benign prostatic hyperplasia patients. The AUC of cg05163709 in csPCA prediction was 0.75. The AUC of the model combined with T2WI and ADC features was 0.91. And the model combined with radiomics features, Methylation level of cg05163709 and clinical data reached an AUC of 0.97, which was greater than that of the model based on the PI-RADS score (AUC = 0.86). Conclusion An effective prediction model for csPCa was successfully established by integrating T2WI features, clinical data, and methylation level of cg05163709 in urine specimens.