Predicting subretinal fluid absorption with machine learning in patients with central serous chorioretinopathy

Area of Science:

• Ophthalmology
• Medical Imaging
• Machine Learning

Background:

• Central serous chorioretinopathy (CSC) is a condition affecting vision.
• Subretinal fluid absorption (SFA) is a key indicator of treatment success in CSC.
• Predicting SFA aids in managing patient prognosis and treatment planning.

Purpose of the Study:

• To develop and validate machine learning models for predicting SFA in CSC patients.
• To identify crucial clinical and imaging features influencing SFA.
• To create a simplified model for practical clinical application.

Main Methods:

• Utilized clinical and imaging data from 480 eyes of 461 CSC patients.
• Employed machine learning algorithms including Random Forest and XGBoost for SFA prediction.
• Validated models using internal (ZOC) and external (XEC) datasets.

Main Results:

• Random Forest achieved high internal validation accuracies for SFA prediction at 1, 3, and 6 months.
• XGBoost demonstrated superior performance in external validation across all time points.
• A simplified model using fewer features showed comparable predictive power.

Conclusions:

• Machine learning models can accurately predict long-term SFA in CSC patients.
• The study identifies key features relevant to patient prognosis.
• Findings support individualized treatment and follow-up strategies for CSC management.