Transforming lung transplantation with artificial intelligence: a narrative review from organ allocation to post-transplant management.

Lung transplantation (LTx) serves as the only definitive therapy for end-stage lung disease, yet, its clinical success is chronically constrained by the severe shortage of donor organs and the high incidence of postoperative complications such as primary graft dysfunction (PGD) and chronic rejection. As a transformative technology, artificial intelligence (AI) demonstrates substantial potential to address these systemic challenges and reshape the entire transplantation clinical pathway. The primary purpose of this article is to consolidate existing research findings to comprehensively assess the current status, technical mechanisms, and future ecosystem of AI applications in various stages of LTx, with a focus on its profound implications for organ allocation optimization, surgical assistance, complication prediction, and personalized medication management. A systematic search was performed in PubMed from inception to September 30, 2025, using a combination of Medical Subject Headings (MeSH) terms and keywords related to "artificial intelligence" (e.g., machine learning, deep learning) and "lung transplantation". Inclusion criteria focused on English-language original research, reviews, and landmark case reports. In the pre-transplant phase, AI optimizes organ allocation by shifting focus from "urgency" to "utility" and enhances donor assessment and matching via computer vision. Intraoperatively, AI integrates with robotic platforms to enable augmented reality navigation and real-time risk warnings. Post-transplant applications, currently the most mature area, utilize machine learning to accurately predict complications like PGD and chronic lung allograft dysfunction (CLAD), enabling non-invasive monitoring (e.g., electronic nose) and personalized immunosuppressant dosing through deep learning analysis of time-series data. AI has demonstrated distinct advantages in improving decision-making precision, optimizing resource allocation, and improving patient prognosis. However, the heterogeneity of data quality, model interpretability, and the complexity of clinical integration remain major barriers to its widespread adoption. Future efforts need to construct a data ecosystem based on FAIR (Findable, Accessible, Interoperable, and Reusable) principles and strengthen human-machine collaboration mechanisms to ensure that algorithmic precision translates into substantive improvements in patient survival and quality of life.