A novel method to predict the haemoglobin concentration after kidney transplantation based on machine learning:...

Area of Science:

• Nephrology
• Medical Artificial Intelligence
• Clinical Prediction Modeling

Background:

• Anemia is a frequent complication post-kidney transplant, with hemoglobin levels being a key diagnostic indicator.
• Artificial intelligence (AI) shows significant promise and has achieved notable success in various medical applications.

Purpose of the Study:

• To refine the construction process of machine learning-based clinical prediction models.
• To develop an advanced classification prediction model for post-kidney transplant hemoglobin concentrations.

Main Methods:

• Retrospective analysis of real-world data from 854 kidney transplant patients.
• Utilized a combined K-nearest neighbor and multilayer perceptron imputation method for missing data.
• Employed recursive feature elimination and extreme gradient boosting for feature selection and dimensionality reduction.
• Developed and optimized classification models using random forest, extreme gradient boosting, light gradient boosting machine, and support vector machine variants with error-correcting output codes.

Main Results:

• The K-nearest neighbor and multilayer perceptron imputation method outperformed standard techniques for handling missing values.
• Error-correcting output code optimization enhanced the predictive performance of tree-based machine learning models.
• The optimized extreme gradient boosting model achieved the highest accuracy, reaching 87.22% post-optimization from 85.98%.

Conclusions:

• The developed machine learning classification model, optimized through advanced methods, demonstrated a high accuracy of 87.22%.
• This model can serve as a valuable tool for clinicians in predicting preoperative risks associated with abnormal hemoglobin levels in kidney transplant recipients.