An effective technique for diabetic retinopathy using hybrid machine learning technique

Area of Science:

• Ophthalmology and Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Diabetic retinopathy (DR) is a progressive complication of diabetes, damaging retinal blood vessels and potentially causing vision loss.
• Early detection and timely treatment are crucial for managing DR and preventing severe visual impairment.
• Identifying abnormal retinal blood vessels (RBVs) is key to understanding DR progression.

Purpose of the Study:

• To develop an automated approach for the detection of diabetic retinopathy by analyzing retinal blood vessels.
• To improve the accuracy and efficiency of DR diagnosis through advanced image analysis techniques.

Main Methods:

• A two-step method involving segmentation and classification of retinal blood vessels.
• Segmentation utilizes Kinetic Gas Molecule Optimization with Fuzzy C-means Clustering.
• Classification employs a hybrid Convolutional Neural Network (CNN) with Bidirectional Long Short-Term Memory (Bi-LSTM) enhanced by a self-attention mechanism.

Main Results:

• The proposed hybrid algorithm achieved higher accuracy, specificity, and sensitivity compared to existing methods.
• The self-attention mechanism refined the classification accuracy of the Bi-LSTM model.
• The method effectively segments and classifies affected retinal blood vessels for DR detection.

Conclusions:

• The developed automated approach shows significant potential for accurate and efficient diabetic retinopathy detection.
• The integration of CNN, Bi-LSTM, and self-attention offers a promising direction for medical image analysis in ophthalmology.
• Early and precise diagnosis of DR can be facilitated, leading to better patient outcomes.