Development and validation of a machine learning model to predict recurrence in polypoidal choroidal vasculopathy: a...

This study aims to develop and validate machine learning models for predicting recurrence in polypoidal choroidal vasculopathy (PCV) patients using optical coherence tomography (OCT) and OCT angiography (OCTA) biomarkers.

This multicentre prospective study was conducted at 14 hospitals between June 2019 and December 2023. Patients who achieved remission after anti-vascular endothelial growth factor treatment were followed up for at least 1 year with serial OCT/OCTA imaging. Predictive features were selected using the least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation. Five classifiers (logistic regression, support vector machine, random forest, k-nearest neighbours and extreme gradient boosting (XGBoost)) were developed and evaluated using area under the curve (AUC), accuracy, sensitivity and specificity. Shapley additive explanations (SHAP) were applied for model interpretation and feature ranking.

A total of 204 eyes were included, with 125 from Peking Union Medical College Hospital (training set) and 79 from 13 other centres across China (external validation set). Ten features were selected for model development. In the external validation set, AUCs ranged from 0.801 to 0.861, with the XGBoost model achieving the highest AUC (0.861). SHAP analysis revealed that the percent change in polyp height, the change in branching neovascular network area and the change in subfoveal choroidal thickness were the top three significant predictors.

The XGBoost model demonstrated the best predictive performance, providing a reliable tool for recurrence prediction, aiding personalised treatment decisions and optimising clinical resources.