Machine Learning Support for Decision-Making in Kidney Transplantation: Step-by-step Development of a Technological...

Area of Science:

• Nephrology
• Transplant Surgery
• Biomedical Informatics

Background:

• Kidney transplantation is the leading treatment for end-stage renal disease.
• Optimizing donor-recipient matching is crucial for maximizing patient and graft survival.
• Accurate prediction of transplant outcomes is essential for informed decision-making.

Purpose of the Study:

• To develop an innovative technological solution for predicting kidney transplant survival.
• To enhance the selection process for donor-recipient pairs.
• To improve outcomes for kidney transplant patients.

Main Methods:

• Utilized deidentified data from the Scientific Registry of Transplant Recipients.
• Compared survival analysis models: Cox regression, random survival forests, and artificial neural networks (DeepSurv, DeepHit, RNN).
• Evaluated model performance using C-index, integrated Brier score, and integrated calibration index.

Main Results:

• Neural network models, particularly RNN, demonstrated superior discriminative ability compared to traditional models.
• The RNN model achieved a C-index of 0.659, showing strong predictive performance.
• The developed RNN-based solution reached technology readiness level 4.

Conclusions:

• The RNN-based technological solution effectively predicts kidney transplant survival.
• This tool supports medical professionals and patients in identifying optimal donor-recipient matches.
• The innovation has the potential to improve kidney transplant success rates.