Automatic Classification of Macular Diseases from OCT Images Using CNN Guided with Edge Convolutional Layer

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Science

Background:

• Optical Coherence Tomography (OCT) is crucial for diagnosing retinal pathologies.
• Accurate segmentation of retinal layers is vital for disease diagnosis.
• Manual classification of OCT images is time-consuming and labor-intensive.

Purpose of the Study:

• To develop an automated method for classifying retinal pathologies from OCT images.
• To introduce a novel Edge Convolutional Layer (ECL) for improved retinal boundary extraction.
• To propose an ECL-guided Convolutional Neural Network (ECL-CNN) for enhanced OCT image classification.

Main Methods:

• Developed a new Edge Convolutional Layer (ECL) for precise retinal boundary extraction.
• Proposed an ECL-guided Convolutional Neural Network (ECL-CNN) for automated OCT image classification.
• Evaluated the ECL-CNN on a dataset of 45 OCT volumes (15 normal, 15 AMD, 15 DME).

Main Results:

• The ECL accurately extracts retinal boundaries with fewer parameters than conventional layers.
• The ECL-CNN achieved an outstanding average precision of 99.43% for three-class classification.
• Demonstrated superior performance in classifying age-related macular degeneration (AMD) and diabetic macular edema (DME) from OCT images.

Conclusions:

• The proposed ECL-CNN method offers a highly accurate and efficient automated solution for retinal OCT image classification.
• This approach can significantly reduce the manual workload for ophthalmologists.
• The ECL-CNN holds promise for clinical applications in diagnosing various retinal diseases.