Intelligent organ-on-chip platforms: Machine learning in predictive and personalized drug delivery

The pursuit of physiologically relevant preclinical models has driven the emergence of three-dimensional tumor spheroids and organ-on-chip (OOC) platforms, which recapitulate the structural, biochemical, and functional complexity of human tissue microenvironments. Unlike conventional two-dimensional cultures and animal models, these systems provide human-specific insights into disease biology, drug delivery, and therapeutic response, addressing critical translational gaps in biomedical research. They reproduce cell–cell and cell–matrix interactions, vascularization, nutrient and drug gradients, and organ-level functions, offering new opportunities to evaluate drug penetration, efficacy, and safety. Their complexity generates high-dimensional imaging, omics, and functional datasets that are increasingly difficult to analyze using traditional approaches, setting the stage for the integration of artificial intelligence (AI) and machine learning (ML) as transformative tools. Recent advances demonstrate how AI-driven methods can automate spheroid image segmentation, quantify spatial heterogeneity, and extract predictive biomarkers of treatment response with minimal human bias. Deep learning algorithms are also being applied to predict drug transport dynamics, optimize dosing strategies, and analyze complex interactions within OOC systems, enhancing experimental precision and translational relevance. Moreover, AI integration with Multi-OOC platforms is beginning to capture systemic pharmacokinetics and pharmacodynamics, linking in vitro performance with in vivo outcomes and reshaping drug delivery research by coupling biological fidelity with computational intelligence. This review critically analyzes AI-enhanced spheroid and OOC platforms for drug delivery, highlighting progress and key challenges, such as data standardization, interpretability, and reproducibility, while outlining future directions, including the convergence of digital twins, ML, and multi-scale modeling toward intelligent, personalized, and clinically translatable drug delivery systems.