RADIOMICS TO PREDICT THE RESPONSE OF BRAIN METASTASES TO STEREOTACTIC RADIOTHERAPY

Abstract AIMS Radiomics converts routinely obtained medical imaging into mineable, high-dimensionality data. We assessed its utility for incorporation a multivariate model to predict the likelihood of response brain metastases (BMs) stereotactic radiotherapy (SRT). METHOD Patients treated with SRT over ten-year period BMs at three tertiary centres were retrospectively analyzed. planning MRI scans pre-processed, and radiomic features extracted from each BM using PyRadiomics software. Follow up analyzed quantified by modification Response Assessment in Neuro-Oncology Brain Metastases criteria. A signature predictive was created stepwise process involving correlation coeffcients binary least absolute shrinkage se- lection operator regression analysis. clinical predicting com- pared against combination clinical/radiomic McFadden’s pseudo-R-squared test. RESULTS 520 patients included contributing total 1891 BMs. 1131 BM. 12 used construct signature. Mean dose volume independent predictors both radiomics models. The an prognostic predictor (HR=1.095-1.231). Goodness fit greater (McFadden’s score 0.243 vs 0.211), statistically significant ratio test (p=0.021). CONCLUSIONS This multi-centre retrospective analysis unique is first study incorporate factors. prospective validation planned.