Fast and Efficient Method for Optical Coherence Tomography Images Classification Using Deep Learning Approach

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Optical Coherence Tomography (OCT) is widely used in medical diagnostics.
• Increasing volumes of OCT data necessitate automated support systems for medical professionals.
• Accurate and efficient classification of OCT images is crucial for timely diagnosis and treatment.

Purpose of the Study:

• To develop and present a novel convolutional neural network (CNN) based approach for classifying eye diseases from OCT B-scans.
• To create an automated support system to assist medical staff in interpreting large datasets of OCT images.
• To classify OCT B-scans into four categories: Diabetic Macular Edema (DME), Choroidal Neovascularization (CNV), Drusen, and Normal.

Main Methods:

• Utilized a publicly available dataset of over 84,000 OCT B-scan images.
• Developed and tested a 5-layer convolutional neural network (CNN) model for image classification.
• Implemented and detailed image data augmentation techniques to improve classification performance.
• Evaluated multiple derived CNN architectures and compared performance against existing solutions.

Main Results:

• The proposed 5-layer CNN model achieved promising classification results.
• The algorithm demonstrated high quality when compared to other available solutions.
• The model's limited size contributed to reduced computational time, enhancing its practical applicability.
• Data augmentation significantly impacted the classification outcomes, leading to improved accuracy.

Conclusions:

• The developed CNN-based algorithm offers a high-quality and efficient method for classifying eye diseases from OCT B-scans.
• This automated system can serve as a valuable tool for speeding up diagnosis during screening tests, complementing the role of clinicians.
• The research highlights the potential of machine learning in improving medical treatment processes through enhanced diagnostic support.