Artificial intelligence and machine learning in heart and lung transplantation

Artificial intelligence (AI) and machine learning (ML) are increasingly applied across the heart and lung transplant continuum, from donor organ evaluation and waitlist prioritisation through peri-operative decision support to post-transplant rejection surveillance and personalised immunosuppression. Deep learning for endomyocardial biopsy interpretation has achieved an area under the curve (AUC) of 0.962, with performance comparable to expert pathologists, while ML-augmented ex-vivo lung perfusion platforms have demonstrated the potential to expand organ utilisation in real-world implementation studies. Noninvasive donor-derived cell-free DNA assays and electrocardiogram-based deep learning models are beginning to reduce dependence on invasive surveillance biopsies. Across published studies, models predicting 1-year post-transplant mortality have shown only moderate discrimination. Most published models remain retrospectively validated on single-centre datasets with limited external validation, inconsistent calibration reporting, and unresolved concerns about algorithmic bias and global health equity. This scoping review synthesises current evidence, critically appraises model performance, and identifies the regulatory, ethical, and methodological steps required for responsible clinical integration. Future research should prioritise prospective, multicentre external validation, transparent calibration reporting, and formal evaluation of clinical utility before these tools are adopted in routine transplant care.