Adopting machine learning to predict nomogram for small incision lenticule extraction (SMILE)

Area of Science:

• Ophthalmology
• Medical Technology
• Data Science

Background:

• Small Incision Lenticule Extraction (SMILE) is a refractive surgery procedure.
• Accurate nomograms are crucial for optimizing refractive outcomes in SMILE surgery.
• Predicting nomograms traditionally relies on surgeon experience and clinical data.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting nomograms in SMILE surgery.
• To compare the accuracy of machine learning predictions against traditional methods and junior residents.
• To assess the potential of machine learning in assisting refractive surgeons and improving nomogram adjustment.

Main Methods:

• Utilized preoperative clinical data from 1025 eyes undergoing SMILE.
• Applied six machine learning algorithms: XGBoost, Gradient Boosting Regression (GBR), Random Forest (RF), LightGBM, Linear Regression (LR), and Support Vector Regression (SVR).
• Assessed model performance using Root Mean Absolute Error (RMSE) and Mean Absolute Error (MAE); compared accuracy with junior residents at specific diopter thresholds.

Main Results:

• Machine learning models demonstrated accurate nomogram prediction, with no significant difference from actual nomograms (P > 0.05).
• RMSE values ranged from 0.075 to 0.110, and MAE values ranged from 0.055 to 0.085 across the models.
• XGBoost showed significantly higher accuracy than SVR and junior residents (P < 0.001) within ±0.05D to ±0.25D thresholds; other top ML models performed comparably.

Conclusions:

• Machine learning effectively predicts nomograms for SMILE surgery using preoperative clinical data.
• ML methods show promise in assisting refractive surgeons and shortening the learning curve for junior residents in nomogram adjustment.
• The study highlights the potential of AI in enhancing refractive surgery precision and efficiency.