Prediction of the Cause of Fundus-Obscuring Vitreous Hemorrhage Using Machine Learning
Jinsoo Kim1, Bo Sook Han1, Joo Eun Ha1
1Department of Ophthalmology, Hallym University Sacred Heart Hospital, Hallym University College of Medicine, Anyang 14068, Republic of Korea.
Diagnostics (Basel, Switzerland)
|February 13, 2025
Summary
Machine learning accurately predicts the cause of fundus-obscuring vitreous hemorrhage (FOVH) using preoperative data. Key factors include diabetic retinopathy, retinal vein occlusion, and fellow eye conditions, aiding treatment planning.
Area of Science:
- Ophthalmology
- Medical Artificial Intelligence
- Retinal Diseases
Background:
- Fundus-obscuring vitreous hemorrhage (FOVH) presents diagnostic challenges.
- Identifying the etiology preoperatively is crucial for effective treatment.
Purpose of the Study:
- To predict the unknown etiology of FOVH using machine learning (ML) based on preoperative conditions.
- To identify key preoperative factors influencing FOVH etiology.
Main Methods:
- Retrospective review of 223 eyes with FOVH of unknown etiology.
- Collected preoperative data including demographics, systemic diseases, and ophthalmic history.
- Trained and evaluated four ML algorithms (ANN, XG-Boost, LASSO, Decision Tree) using seven-fold cross-validation.
Main Results:
- ML algorithms achieved predictive accuracies ranging from 68.5% to 76.2%.
- Key predictive factors identified: proliferative diabetic retinopathy (PDR) in the fellow eye, diabetes mellitus, subarachnoid hemorrhage, and history of retinal tear, retinal vein occlusion (RVO), or neovascular age-related macular degeneration (nAMD).
Conclusions:
- ML models can predict the etiology of FOVH preoperatively with significant accuracy.
- Preoperative ophthalmic conditions and fellow eye status are critical predictors.
- This ML approach can assist in formulating appropriate treatment strategies for FOVH.
