Automatic Detection of Genetics and Genomics of Eye Disease Using Deep Assimilation Learning Algorithm

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Science

Background:

• Diabetic retinopathy (DR) is a leading cause of vision loss, often undetected due to a lack of experts.
• Current image processing methods for DR diagnosis face challenges with noise removal, segmentation, and classification accuracy.
• Cotton wool spots (CWS) and hard exudates (HE) are key indicators of retinal diseases, including DR.

Purpose of the Study:

• To develop an advanced image processing method for accurate automatic diagnosis of diabetic retinopathy (DR).
• To improve the detection and classification of retinal abnormalities like CWS and HE.
• To overcome limitations in existing DR diagnostic techniques.

Main Methods:

• Utilized feature-based medical image retrieval (FBMIR) datasets for analysis.
• Applied histogram filtering for noise removal and conversion to grayscale images.
• Proposed a Super Iterative Clustering Algorithm (SICA) for CWS and HE detection.
• Employed a Deep Assimilation Learning Algorithm (DALA) for feature-based classification.

Main Results:

• The proposed SICA method effectively identified CWS and HE, eliminating irrelevant image areas.
• DALA achieved high performance in classification, evaluated using recall, precision, and F-measure.
• The overall method demonstrated an accurate detection rate of 98.5%, surpassing conventional approaches.

Conclusions:

• The developed image processing and deep learning approach significantly enhances the accuracy of automatic DR detection and classification.
• This method offers a promising tool for early diagnosis and management of diabetic retinopathy.
• Improved accuracy in identifying retinal markers like CWS and HE can aid in preventing vision loss.