Development, External Validation, and Deployment of RFAN-ML: A Machine Learning Model to Estimate Renal Function...

Area of Science:

• Nephrology
• Oncology
• Data Science

Background:

• Partial nephrectomy is preferred for small kidney tumors but carries perioperative risks.
• Accurate estimation of post-nephrectomy renal function is crucial for patient counseling and surgical decisions.
• Existing prediction models often lack external validation and user-friendly interfaces.

Purpose of the Study:

• To develop and externally validate a machine learning (ML) model, RFAN-ML, for estimating long-term renal function after nephrectomy.
• To provide a user-friendly tool for predicting renal function post-nephrectomy.

Main Methods:

• Utilized data from two academic medical institutions.
• Employed Boruta feature selection to identify key predictors: age, BMI, preoperative renal function, and nephrectomy type.
• Trained and evaluated six ML regression models, selecting the best performing one as RFAN-ML.

Main Results:

• RFAN-ML demonstrated superior or competitive performance compared to existing benchmarks.
• Achieved a root mean squared error (RMSE) of 16.6 (95% CI, 15.6 to 17.5).
• Performance was evaluated using R-squared and mean absolute error metrics.

Conclusions:

• RFAN-ML, a validated ML model, accurately predicts renal function post-nephrectomy.
• The model is available online, facilitating personalized patient counseling and surgical planning.
• RFAN-ML has the potential to enhance care and outcomes for kidney tumor patients.