Artificial Intelligence and Computational Medicine

Abstract

This chapter provides a hands-on, code-driven introduction to artificial intelligence in computational medicine, bridging the gap between theoretical understanding and practical implementation. Covering three interconnected domains (medical imaging, multimodal data integration, and AI-assisted computing), we demonstrate current best practices using reproducible Jupyter notebooks with publicly available and synthetic datasets. First, we address brain tumor segmentation from multiparametric MRI using a 3D U-Net trained with Dice loss, alongside promptable segmentation with the MedSAM2 foundation model. Second, we present multimodal data integration through adaptive fusion networks that learn modality-specific importance weights, combined with graph neural networks operating on patient similarity networks for semi-supervised learning. Third, we explore the role of large language models in medical research workflows, including text summarization, zero-shot classification, named entity recognition, and retrieval-augmented generation for grounding responses in verified evidence. Throughout, we discuss key challenges: class imbalance, clinical interpretability, hallucination mitigation, missing modalities, privacy-preserving deployment, algorithmic fairness, and regulatory considerations. All code, pretrained model weights, and materials are openly available, designed to run locally or on Google Colab, enabling researchers and clinicians to engage directly with working implementations.