From data to decisions: Machine learning applications in kidney transplantation management

Abstract

Kidney transplantation (KT) is a complex clinical process requiring accurate risk stratification, timely decision-making, and individualized management across all stages of care. Traditional clinical scores and center-specific algorithms often fail to capture the multidimensional interactions among donor, recipient, and procedural factors. In this context, artificial intelligence (AI) and machine learning (ML) have emerged as promising tools to support precision medicine in KT. We conducted a narrative review of the literature published between 2010 and 2024, focusing on AI and ML applications in KT. Studies addressing pretransplant evaluation, surgical assistance, posttransplant outcomes, and immunosuppressive therapy optimization were included and qualitatively synthesized. Artificial intelligence and ML models have demonstrated improved performance compared with conventional approaches across multiple stages of KT. In the pretransplant setting, ML-based models outperform traditional risk scores in donor–recipient matching and graft survival prediction, with the potential to reduce organ discard. Deep learning techniques applied to preimplantation biopsies enhance histopathologic assessment and reproducibility. Intraoperatively, AI-driven imaging and surgical data science tools support anatomic visualization and decision-making. Posttransplant, ML models enable earlier detection of acute rejection, prediction of delayed graft function, and noninvasive assessment using laboratory, imaging, and molecular data. Furthermore, AI-based pharmacokinetic models facilitate personalized immunosuppressive dosing, improving time within therapeutic ranges and reducing toxicity. Artificial intelligence represents a powerful adjunct to clinical expertise in KT, offering data-driven decision support across the transplant continuum. However, most available models remain retrospective and lack prospective multicenter validation, limiting immediate clinical implementation. Prospective multicenter studies and responsible implementation strategies are essential to translate AI-driven tools into routine clinical practice and improve transplant outcomes.