DR-NASNet: Automated System to Detect and Classify Diabetic Retinopathy Severity Using Improved Pretrained NASNet...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Diabetes mellitus is a prevalent condition leading to severe visual impairment through Diabetic Retinopathy (DR).
• DR is a leading cause of blindness in diabetic patients, characterized by irreversible damage to the retina.
• Early detection and classification of DR severity are critical for timely intervention and management.

Purpose of the Study:

• To develop a reliable automated system, DR-NASNet, for detecting and classifying the severity of Diabetic Retinopathy.
• To improve the accuracy and efficiency of DR classification using an enhanced deep learning model.

Main Methods:

• Utilized a pretrained NASNet model integrated with dense blocks for DR severity classification.
• Employed preprocessing techniques (Ben Graham, CLAHE) to enhance image quality and lesion emphasis.
• Applied data augmentation to address class imbalance and prevent overfitting in the dataset.

Main Results:

• The DR-NASNet system achieved state-of-the-art performance with a smaller model size and reduced complexity.
• Achieved a high accuracy of 96.05% on a challenging DR dataset across five severity levels.
• Demonstrated improved model performance and learning capabilities compared to existing methods.

Conclusions:

• The DR-NASNet system offers an effective solution for automated DR severity classification.
• This automated system can assist ophthalmologists in the early detection and management of Diabetic Retinopathy.
• The proposed method shows significant potential for improving patient outcomes by enabling timely diagnosis.