Photo Credit: peakSTOCK
The following is a abstract of “Practical Clinical Role of Machine Learning Models with Different Algorithms in Predicting Prostate Cancer Local Recurrence after Radical Prostatectomy,” printed within the February 2024 challenge of Oncology by Hu et al.
Detection of native recurrence in prostate most cancers (PCa) sufferers post-radical prostatectomy (RP) poses a big medical problem, impacting remedy selections.
Their goal was to develop and validate machine studying fashions using three distinct algorithms primarily based on post-operative mpMRI to foretell PCa native recurrence after RP, evaluating their medical utility with the Prostate Imaging for Recurrence Reporting (PI-RR) rating supplied by skilled radiologists.
A retrospective evaluation included 176 sufferers randomly divided into coaching (n = 123) and testing (n = 53) cohorts. PI-RR assessments have been performed by two skilled radiologists with entry to operative histopathological and pre-surgical medical knowledge. Radiomics fashions predicting native recurrence have been constructed utilizing help vector machine (SVM), linear discriminant evaluation (LDA), and logistic regression-least absolute shrinkage and choice operator (LR-LASSO) algorithms.
A mixed mannequin integrating radiomics options and PI-RR rating was developed utilizing the best classifier. Model efficiency was evaluated utilizing receiver working attribute (ROC) curve evaluation. There have been no important variations in affected person traits between coaching and testing units. The LR-LASSO-based radiomics mannequin demonstrated superior efficiency with an AUC of 0.858 within the testing set, whereas PI-RR yielded an AUC of 0.833, with no important distinction between the perfect radiomics mannequin and the PI-RR rating. The mixed mannequin achieved the best predictive efficiency, with an AUC of 0.924, considerably outperforming the PI-RR rating alone.
Their findings indicated that the radiomics mannequin successfully predicts PCa native recurrence post-RP. Integration of radiomics options with the PI-RR rating enhances predictive accuracy, surpassing skilled radiologists’ assessments.
Source: cancerimagingjournal.biomedcentral.com/articles/10.1186/s40644-024-00667-x
https://www.physiciansweekly.com/predictive-utility-of-diverse-machine-learning-models-for-pca-local-recurrence-following-rp/