Machine Learning-Based Clinical Decision Support Systems for Personalized Stem Cell Treatments in Regenerative Medicine

Abstract

The incorporation of machine learning (ML) into clinical decision support systems (CDSS) has brought new possibilities for the further development of personalized stem cell therapies in the field of regenerative medicine. Furthermore, ML-based CDSS ae) provide a characterization of disease and patient-specific information (e.g., genomic profiles, biomarker levels, clinical histories), b) assist clinicians to predict therapy efficacy and c) optimize cell type selection and tailor therapeutic procedures at the individual patient level. In this paper, current methods for the implementation of ML in stem cell therapy were reviewed with focus on supervised, unsupervised, and hybrid algorithms for patient stratification, treatment recommendation, and risk prediction. Electronic health records, high-throughput sequencing datasets and clinical trial repositories are explored for their contribution to the accuracy and reliability of the model. We describe issues of data heterogeneity, model interpretability, and clinical integration, and address ethical and regulatory issues that affect patient safety and treatment efficacy. Conclusions: Looking beyond this paper to the implementation path to market of ML-based CDSS for regenerative medicine, we highlight the importance of careful validation processes, cross-disciplinary collaboration, and iterative learning in order to maximize the potential of CLT while minimizing the risks.