Project Summary: Brain metastases are an emerging challenge in modern Neuro-Oncology due to increasing incidence and limited treatments. Recent translational studies have demonstrated the potential efficacy of checkpoint inhibitors in brain metastases. However, despite extensive molecular and functional imaging analyses, a definitive biomarker for intracranial efficacy of checkpoint inhibitors remains elusive.

In an effort to accelerate the clinical translation of this promising therapeutic strategy, our objective is to harness three complementary data modalities in MRI, clinical metadata, and tumor genomics into robust biomarkers that capture clinically significant facets of brain metastases.

To achieve this, the thesis explores the application of deep learning models that can seamlessly integrate these diverse data sources into a unified model, outperforming existing clinical single-modal biomarkers. Given the limitations of single-modal networks in capturing the complexity of diseases like brain cancer, effective multimodal fusion becomes increasingly pivotal in enhancing personalized care.