Enhancing Prognostic Prediction for Patients with Brain Metastasis from Lung Cancer after Radiosurgery Based on Tumor and Vasculature Radiomics

Motivation: Recent studies showed the relationship between the tumor-associated vasculature (TAV) and the treatment response to radiosurgery in brain metastasis (BM). However, the enhancement of outcome prediction achieved by combining tumor and vasculature features of BM was unexplored. Goal(s): This study aimed to integrate tumor and vasculature radiomic features to investigate the improvement of overall survival prediction in BM. Approach: The DeepSurv models were built with intra-tumoral, peri-tumoral, and TAV features and assessed with the time-dependent AUC and concordance index. Results: The model showed the best performance in predicting survival over 24 months for BM patients after radiosurgery by combining tumor and vasculature features. Impact: We identified the benefit of combining tumor and vasculature features in prognostic prediction for lung cancer patients with BM after radiosurgery. This study also reflected the association between TAV and prognosis for the patient with overall survival over 24 months.