Predicting Central Serous Chorioretinopathy Recurrence Using Machine Learning.

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Central serous chorioretinopathy (CSC) is a condition affecting vision.
• Predicting CSC recurrence after treatment is crucial for patient management.
• Laser treatment is a common intervention for CSC.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting CSC recurrence at 3 and 6 months post-laser treatment.
• To compare the performance of different machine learning algorithms in predicting CSC recurrence.
• To assess the utility of a simplified model using clinical and OCT data.

Main Methods:

• Data from 461 patients (480 eyes) with CSC were collected from two centers.
• Machine learning algorithms, including an ensemble model, were trained on clinical and imaging data.
• Internal and external validation datasets were used to test model performance.

Main Results:

• The ensemble machine learning model achieved high prediction accuracies: 0.941 (3 months, internal), 0.970 (3 months, external), 0.903 (6 months, internal), and 1.000 (6 months, external).
• A simplified model using only clinical and OCT features demonstrated comparable predictive power.
• The models successfully predicted recurrence at both 3 and 6 months post-treatment.

Conclusions:

• Machine learning models demonstrate high accuracy in predicting central serous chorioretinopathy recurrence.
• An intelligent prediction model can assist in identifying factors contributing to CSC recurrence.
• This approach supports precise, individualized treatment strategies for CSC patients.