Construction of Predictive Model for Type 2 Diabetic Retinopathy Based on Extreme Learning Machine

Area of Science:

• Ophthalmology
• Medical Informatics
• Machine Learning

Background:

• Diabetic retinopathy (DR) is a leading cause of blindness in type 2 diabetes (T2D) patients.
• Early detection through fundus examinations prevents vision loss but faces healthcare system challenges.
• Developing efficient screening tools is crucial for managing diabetes-related eye complications.

Purpose of the Study:

• To develop a diabetic retinopathy (DR) prediction model using the extreme learning machine (ELM) algorithm.
• To compare the performance of the ELM-based DR model against Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and Artificial Neural Network (ANN) models.

Main Methods:

• Data collected from electronic inpatient medical records (January 2020 - November 2021).
• An ELM algorithm was employed to build the prediction model using demographic, blood, and urine test data.
• Model performance evaluated using accuracy (ACC), sensitivity, specificity, precision, negative predictive value (NPV), training time, and AUC.

Main Results:

• ELM and SVM models outperformed ANN, KNN, and RF in ACC, sensitivity, specificity, precision, NPV, and AUC.
• The ELM model demonstrated superior performance in ACC (84.45%), precision (83.93%), specificity (93.16%), training time (1.24s), and AUC (88.34%).
• The SVM model achieved the highest sensitivity (70.82%) and NPV (85.60%).

Conclusions:

• The ELM-based model exhibits excellent performance for diabetic retinopathy prediction.
• This model offers valuable technological support for DR screening in clinical settings.
• Machine learning approaches can address the challenges of routine ophthalmic screenings for diabetic patients.