HDL-ACO hybrid deep learning and ant colony optimization for ocular optical coherence tomography image classification

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Optical Coherence Tomography (OCT) is vital for diagnosing eye diseases.
• Conventional Convolutional Neural Networks (CNNs) struggle with computational load, noise, and imbalanced data in OCT analysis.

Purpose of the Study:

• To introduce HDL-ACO, a Hybrid Deep Learning (HDL) framework integrating CNNs and Ant Colony Optimization (ACO).
• To enhance classification accuracy and computational efficiency for OCT image analysis.

Main Methods:

• Pre-processing OCT data with discrete wavelet transform and ACO-optimized augmentation.
• Utilizing multiscale patch embedding and ACO for hyperparameter optimization.
• Employing a Transformer-based feature extraction module with content-aware embeddings and multi-head self-attention.

Main Results:

• HDL-ACO achieved 95% training accuracy and 93% validation accuracy.
• Outperformed state-of-the-art models like ResNet-50, VGG-16, and XGBoost.
• Demonstrated improved feature selection and training efficiency.

Conclusions:

• HDL-ACO offers a scalable and resource-efficient solution for real-time OCT image classification.
• The framework significantly enhances diagnostic capabilities in ophthalmology.
• Addresses limitations of traditional CNN models in medical image analysis.