Artificial Intelligence Powered Fundus Image Enhancement with Smart Error Detection and Correction for Accurate Diabetic Retinopathy Analysis

Diabetic Retinopathy is a prevalent eye condition and a significant precursor to blindness in individuals with diabetes. It affects 80–85% of those who have lived with diabetes for over a decade. Retinal fundus images are frequently utilized in clinical settings to identify and examine DR. Analysing these images manually is a labor intensive process that demands considerable expertise, which can occasionally result in incorrect diagnoses. Thus, it is crucial to accurately identify and categorize DR for effective clinical diagnosis and treatment. The research was carried out using AGGHE and ORS. Kirsch edge detection, fuzzy clustering, morphological distance and watershed transformation algorithms were utilized to extract features from blood vessels, exudates, MA and the optic disc, respectively. A novel classification system called DenseRetinoNet-169 was used to classify the different stages of DR. When numerous medical images and data are exchanged among physicians via a public network, there is a possibility of transmission errors, such as the distortion of the affected areas through enlargement or reduction. This leads to incorrect medical diagnosis. To overcome this problem, an effective transmission using HEED code is developed. This work mainly focuses on detecting and correcting the changes in the transmitted images so that it enables the exact diagnosis of patients. The models were developed using Matlab 2024 platform. The assessment outcomes indicate that the proposed method exhibits excellent imperceptibility, robustness, security and recovery accuracy when subjected to image processing attacks. Consequently, this method is suitable for the efficient transmission of medical images.