Models in kidney transplantation: from AI models to organ models

Abstract

Kidney transplantation stands as the most effective treatment for end-stage renal disease. However, its success has led to an increased demand for donor organs, resulting in a growing reliance on kidneys from older donors. These kidneys often exhibit lower quality, reflected in high discard rates. Consequently, enhancing the pre-transplant assessment of donor kidneys is pivotal. This can be achieved through the use of various predictive models to evaluate organ quality and predict transplantation outcomes.

Advanced machine learning algorithms, such as random forest, or extreme gradient boosting, hold significant promise in predicting both short-term and long-term graft survival. Despite their potential, these models necessitate rigorous external validation to ensure their accuracy and generalizability.

Additionally, ex vivo normothermic machine perfusion emerges as a valuable technique for assessing organ quality. When combined with advanced imaging modalities and multi-omics analyses, these methods represent innovative strategies for improving pre-transplant viability assessment.

Ultimately, adopting a comprehensive approach that integrates pre-transplant donor and recipient data, ex vivo perfusion-derived data, and post-transplant information can provide a holistic framework for predicting transplant outcomes. This integrative strategy has the potential to significantly enhance long-term graft survival rates, advancing the field of kidney transplantation.