Machine Learning and Deep Learning for Multi-omics and Signal-Based Biomarker Discovery in Cardio–Brain Oncology: A Comprehensive Survey

The convergence of oncology, cardiology, and neuroscience has led to the emergence of cardio–brain oncology, a considered area of inter-disciplinary research addressing the complex factors involved in the cross-interactions of oncogenesis, oncologic therapeutics, dysfunction of the cardiovascular system, and cognitive/neurological dysfunction. Recent improvements in machine learning and deep learning have enabled data-driven biomarker discoveries through the integration of diverse data sources from patients, including multi-omics profiles, physiological signals, and medical imaging. This chapter presents an extensive overview of machine and deep learning-based discovery methodologies of biomarkers from multi-omics and signal data in the setting of cardio–brain oncology with special focus on representation learning, multimodal fusion frameworks, explainability and privacy-preserving learning frameworks. By offering a critical review of studies published in the past few years, the chapter identifies major methodological trends, comparative understandings, dominating problems, and emerging research directions, in order to hopefully provide a coherent and clinically relevant view of the future of next generation of precision diagnostics and risk stratification of cardio–brain oncology.