Image Preprocessing in Classification and Identification of Diabetic Eye Diseases

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Science

Background:

• Diabetic eye disease (DED) is a significant complication of diabetes, potentially leading to vision loss.
• Early identification of DED through retinal fundus imaging is critical for timely intervention and preventing visual impairment.
• The accuracy of DED diagnostic models relies heavily on the quality and quantity of retinal fundus images.

Purpose of the Study:

• To investigate the importance of image processing techniques in classifying DED from retinal fundus images.
• To develop and evaluate an automated classification framework for DED.
• To assess the performance of a novel convolution neural network (CNN) architecture combined with traditional image processing methods for DED classification.

Main Methods:

• A systematic approach involving image quality enhancement, segmentation of regions of interest, and geometric transformation-based image augmentation.
• Development of a new convolution neural network (CNN) architecture tailored for DED classification.
• Integration of traditional image processing techniques with the proposed CNN model.

Main Results:

• The combined approach of traditional image processing and the novel CNN architecture yielded optimal results for DED classification.
• The automated framework demonstrated adequate accuracy, specificity, and sensitivity in identifying DED.
• The study highlights the effectiveness of the proposed method in improving diagnostic performance.

Conclusions:

• The integration of advanced image processing with a custom CNN architecture offers a promising automated solution for early DED detection.
• Enhanced image quality and data augmentation are crucial for building robust DED diagnostic models.
• This methodology provides a strong foundation for improving the accuracy and efficiency of DED classification in clinical settings.