Comprehensive Summary
This study outlines the use of machine learning(ML) derived nomograms and the use of certain biomarkers to predict the occurrence of side-specific extraprostatic extension(EPE), which is extrusion of prostate cancerous tissue outside the prostatic capsule. This was a retrospective study of the datasets of 108 patients whose biomarkers, primarily from tools such as Prostate-Specific Membrane Antigen Positron Emission Tomography(PSMA-PET), Multiparametric Magnetic Resonance Imaging(mpMRI), and Decipher Genomic Classifier(DGC), were used by two MLs Logistic Regression (LR) and Extreme Gradient Boosting (XGBoost) in order to make judgments about the existence of EPEs in the patients. The PSMA-PET essentially used a radiotracer to highlight cancer cells, the mpMRI combined many MRI images to form a more complete scan, and the DGC was basically a genomic analysis of needle-biopsied cancerous tissue. The study found out that the XGBoost ML outperformed the LR ML in EPE predictive capability when using just the PSMA-PET and DGC biomarkers. In summary, this study addressed the critical need for accurate preoperative EPE predictions using the novel approach of feeding data from 3 important biomarkers to machine learning algorithms and having them predict the possibility of an EPE.
Outcomes and Implications
This was the first study to successfully use machine learning models to predict the existence of EPE using the mpMRI, PSMA-PET and DGC biomarkers. If ML models are used to predict the possibility of EPE before radical prostatectomy procedures, surgeons would have much more insight about what to cut out before arriving at the operating table which may improve postoperative outcomes for patients. The possibility of the existence of extraprostatic extensions presented a big challenge to clinicians especially since it surgeons doing the tumor biopsy often resected more tissue than they may have needed in order to err on the side of caution because they often were not very sure if there was an EPE or not due to the limitations of previous EPE diagnostic tools which could often only be used post-surgery. These unnecessary extensive resections could cause adverse outcomes such as urinary incontinence and erectile dysfunction, which highlights why the introduction of MLs is so important for improving pre-surgical predictions and post-surgical outcomes. Additionally, this approach of using MLs to analyze critical biomarkers could also be used in other types of cancer screenings so their respective resections are more accurate.