Classification of Retinal Diseases in Optical Coherence Tomography Images Using Artificial Intelligence and Firefly...

Area of Science:

• Biomedical engineering
• Medical imaging analysis
• Artificial intelligence in healthcare

Background:

• Automatic diagnosis of biomedical diseases is increasingly reliant on Deep Learning (DL).
• DL offers high accuracy but demands substantial data and computational resources, posing challenges for systems with limited processing power.
• Traditional Machine Learning (ML) is faster and less computationally intensive but generally yields lower accuracy compared to DL.

Purpose of the Study:

• To develop a hybrid system for biomedical disease diagnosis that balances high accuracy with reduced computational load and time.
• To specifically address the diagnosis of retinal diseases using this novel hybrid approach.

Main Methods:

• Image preprocessing for retinal layer extraction.
• Integration of traditional feature extractors with pre-trained Deep Learning feature extractors.
• Utilizing the Firefly algorithm for optimal feature selection.
• Employing multiple binary classifications with Machine Learning classifiers instead of a single multiclass classification.

Main Results:

• The hybrid system achieved a mean accuracy of 0.957 on the first public dataset.
• The system demonstrated a mean accuracy of 0.954 on the second public dataset.
• The developed method effectively reduces computational requirements while maintaining high diagnostic accuracy.

Conclusions:

• The proposed hybrid system offers a promising solution for efficient and accurate biomedical disease diagnosis.
• This approach effectively overcomes the limitations of using solely Deep Learning or traditional Machine Learning for such tasks.
• The method shows significant potential for clinical application in diagnosing retinal diseases and potentially other biomedical conditions.