Photo Credit: peakSTOCK
The following is a summary of “Practical Clinical Role of Machine Learning Models with Different Algorithms in Predicting Prostate Cancer Local Recurrence after Radical Prostatectomy,” published in the February 2024 issue of Oncology by Hu et al.
Detection of local recurrence in prostate cancer (PCa) patients post-radical prostatectomy (RP) poses a significant clinical challenge, impacting treatment decisions.
Their objective was to develop and validate machine learning models employing three distinct algorithms based on post-operative mpMRI to predict PCa local recurrence after RP, comparing their clinical utility with the Prostate Imaging for Recurrence Reporting (PI-RR) score provided by expert radiologists.
A retrospective analysis included 176 patients randomly divided into training (n = 123) and testing (n = 53) cohorts. PI-RR assessments were conducted by two expert radiologists with access to operative histopathological and pre-surgical clinical data. Radiomics models predicting local recurrence were constructed using support vector machine (SVM), linear discriminant analysis (LDA), and logistic regression-least absolute shrinkage and selection operator (LR-LASSO) algorithms.
A combined model integrating radiomics features and PI-RR score was developed using the most effective classifier. Model performance was evaluated using receiver operating characteristic (ROC) curve analysis. There were no significant differences in patient characteristics between training and testing sets. The LR-LASSO-based radiomics model demonstrated superior performance with an AUC of 0.858 in the testing set, while PI-RR yielded an AUC of 0.833, with no significant difference between the best radiomics model and the PI-RR score. The combined model achieved the highest predictive performance, with an AUC of 0.924, significantly outperforming the PI-RR score alone.
Their findings indicated that the radiomics model effectively predicts PCa local recurrence post-RP. Integration of radiomics features with the PI-RR score enhances predictive accuracy, surpassing expert radiologists’ assessments.
Source: cancerimagingjournal.biomedcentral.com/articles/10.1186/s40644-024-00667-x
