Machine Learning for Predicting Pulmonary Graft Dysfunction After Double-Lung Transplantation: A Single-Center Study...

Area of Science:

• Thoracic Surgery
• Transplant Immunology
• Medical Informatics

Background:

• Primary graft dysfunction at 72 hours (PGD3-T72) is a critical complication after lung transplantation.
• Early identification of PGD3-T72 is crucial for timely intervention and improved patient prognosis.

Purpose of the Study:

• To develop and validate an intraoperative machine-learning tool for predicting PGD3-T72.
• To identify key perioperative predictors of PGD3-T72.

Main Methods:

• Retrospective analysis of perioperative data from 477 double-lung transplant recipients.
• Development and comparison of supervised machine-learning models (XGBoost, logistic regression) for PGD3-T72 prediction.
• Hyperparameter optimization using grid search and cross-validation.

Main Results:

• PGD3-T72 occurred in 17.3% of patients.
• The XGBoost model achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.84 at the second graft implantation stage.
• Key predictors included extracorporeal membrane oxygenation (ECMO) use, lactate levels, PaO2/FiO2 ratio, and lung capacity mismatch.

Conclusions:

• Intraoperative prediction of PGD3-T72 is feasible and reliable using machine learning.
• The developed tool demonstrates potential for facilitating early interventions in lung transplant recipients at risk for PGD3-T72.