Early Diabetic Retinopathy Detection from OCT Images Using Multifractal Analysis and Multi-Layer Perceptron...

Area of Science:

• Ophthalmology
• Medical Imaging
• Computational Biology

Background:

• Diabetic retinopathy (DR) is a leading cause of preventable blindness globally.
• Early detection and management are crucial to prevent irreversible vision loss.
• Automated, non-invasive screening tools are vital for modern ophthalmology.

Purpose of the Study:

• To develop a novel framework for early diabetic retinopathy detection.
• To utilize multifractal analysis of Optical Coherence Tomography (OCT) images for DR screening.
• To evaluate machine learning algorithms for classifying DR based on OCT data.

Main Methods:

• A novel framework employing multifractal analysis of OCT images was developed.
• Multifractal features were extracted using a box-counting approach.
• Several machine learning algorithms were evaluated for classification performance.

Main Results:

• The Multi-Layer Perceptron (MLP) algorithm demonstrated the highest predictive accuracy at 98.02%.
• MLP achieved excellent performance metrics: 98.24% precision, 97.80% recall, and 98.01% F1-score.
• The study successfully quantified structural irregularities in retinal tissue associated with DR.

Conclusions:

• Combining OCT imaging, multifractal geometry, and deep learning offers a robust approach for DR screening.
• The proposed method shows significant potential for improving early diagnosis and patient outcomes in diabetic eye care.
• This automated system can aid clinical decision-making and enhance the scalability of DR screening programs.