Predictive Analysis Of EGFR Gene Mutation Types In Brain Metastasis Of Lung Cancer Based On CE-T1WI Radiomic And Machine Learning

Objective: Based on magnetic resonance imaging(MRI)radiomic analysis of brain metastases(BM)from NSCLC to identify radiomic features that were important for predicting EGFR resistance mutation;Methods: This retrospective study enrolled96 patients with proven of BM in NSCLC.Radiomic features extracted from contrast-enhanced T1-weighted imaging(T1CE)sequence.The most predictable features were selected based on the least absolute shrinkage selection operator(LASSO).Then seven radiomics models were built to characterize EGFR resistance mutation by the best classifier;Results: The AUC of EGFR mutation and non-mutation,exon 20 and exon 21,exon 20 and exon 19 predicted by the best-performing model in the training and test set were 0.956 and 0.928,0.987 and0.750,0.962 and 0.643,respectively;Conclusions: MRI-based radiomic signatures have potential to noninvasively predict the emergence of EGFR resistance mutations,facilitating targeted clinical treatment options.