Towards Multimodal Deep Learning for Predicting Survival of Brain Metastatic Cancer

Inês N. Carvalho

Keywords: Brain metastases, Survival prediction, Deep learning, Multimodal data, Magnetic Resonance Imaging, Convolutional Neural Networks, Foundation models

Abstract: Brain metastases (BM) are associated with poor prognosis and high morbidity, yet survival prediction remains limited. Existing methods often rely on radiomic features, overlooking raw imaging and heterogeneous clinical data. We propose a deep learning (DL) framework that integrates multisequence magnetic resonance (MRI) and clinical variables for survival prediction. Our study compares a foundation model (FM) paired with a classifier against conventional convolutional neural networks (CNNs) to develop a robust model.

Serve As Reviewer: ~Inês_N._Carvalho1

Submission Number: 44