Optical Coherence Tomography Image Classification Using Hybrid Deep Learning and Ant Colony Optimization.

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Manual detection of retinal diseases from Optical Coherence Tomography (OCT) images is subjective and error-prone.
• Existing automated methods require further accuracy improvements for reliable early-stage detection.
• Deep learning approaches show promise for enhancing automated retinal disorder diagnosis.

Purpose of the Study:

• To develop and evaluate a deep learning-based automated method for detecting and classifying multiple retinal diseases using OCT images.
• To improve the accuracy and reliability of automated retinal disease diagnosis.
• To compare the performance of the proposed method with and without feature optimization techniques.

Main Methods:

• Utilized three pretrained deep learning models (DenseNet-201, InceptionV3, ResNet-50) for feature extraction via transfer learning.
• Employed Ant Colony Optimization (ACO) to enhance extracted features and select the most relevant ones.
• Classified retinal diseases using K-Nearest Neighbors (KNN) and Support Vector Machine (SVM) algorithms with the optimized features.

Main Results:

• The proposed deep learning method achieved a high accuracy of 99.1% when incorporating Ant Colony Optimization (ACO).
• Without ACO, the method achieved an accuracy of 97.4%.
• The system demonstrated state-of-the-art performance, outperforming existing techniques in accuracy for retinal disease classification.

Conclusions:

• The developed deep learning model effectively detects and classifies retinal diseases from OCT images with high accuracy.
• Ant Colony Optimization significantly improves the performance of the automated diagnostic system.
• This automated approach offers a reliable and accurate alternative to manual diagnosis, aiding in early detection of conditions like diabetic macular edema and age-related macular degeneration.