首页|IEO European Institute of Oncology IRCCS Reports Findings in Prostate Cancer (Ca n we predict pathology without surgery? Weighing the added value of multiparamet ric MRI and whole prostate radiomics in integrative machine learning models)
IEO European Institute of Oncology IRCCS Reports Findings in Prostate Cancer (Ca n we predict pathology without surgery? Weighing the added value of multiparamet ric MRI and whole prostate radiomics in integrative machine learning models)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-New research on Oncology - Prostate Ca ncer is the subject of a report. According to news reporting out of Milan, Italy , by NewsRx editors, research stated, "To test the ability of highperformance m achine learning (ML) models employing clinical, radiological, and radiomic varia bles to improve non-invasive prediction of the pathological status of prostate c ancer (PCa) in a large, singleinstitution cohort. Patients who underwent multip arametric MRI and prostatectomy in our institution in 2015-2018 were considered; a total of 949 patients were included." Our news journalists obtained a quote from the research from the IEO European In stitute of Oncology IRCCS, "Gradient-boosted decision tree models were separatel y trained using clinical features alone and in combination with radiological rep orting and/or prostate radiomic features to predict pathological T, pathological N, ISUP score, and their change from preclinical assessment. Model behavior was analyzed in terms of performance, feature importance, Shapley additive explanat ion (SHAP) values, and mean absolute error (MAE). The best model was compared ag ainst a naive model mimicking clinical workflow. The model including all variabl es was the best performing (AUC values ranging from 0.73 to 0.96 for the six end points). Radiomic features brought a small yet measurable boost in performance, with the SHAP values indicating that their contribution can be critical to succe ssful prediction of endpoints for individual patients. MAEs were lower for low-r isk patients, suggesting that the models find them easier to classify. The best model outperformed (p 0.0001) clinical baseline, resulting in significantly fewe r false negative predictions and overall was less prone to under-staging. Our re sults highlight the potential benefit of integrative ML models for pathological status prediction in PCa. Additional studies regarding clinical integration of s uch models can provide valuable information for personalizing therapy offering a tool to improve non-invasive prediction of pathological status. The best machin e learning model was less prone to under-staging of the disease. The improved ac curacy of our pathological prediction models could constitute an asset to the cl inical workflow by providing clinicians with accurate pathological predictions p rior to treatment. • Currently, the most common strategies for pre-surgical stra tification of prostate cancer (PCa) patients have shown to have suboptimal perfo rmances. • The addition of radiological features to the clinical features gave a considerable boost in model performance."
MilanItalyEuropeCancerCyborgsE merging TechnologiesHealth and MedicineMachine LearningOncologyPathologyProstate CancerProstatic NeoplasmsSurgery
NETL
NSTL
