Enhanced Diabetes Detection Through CNN-SVM Hybrid Architecture

Corresponding and unconventional medicine approaches have shown capacity for the treatment and diagnosis of chronic illnesses such as diabetes and arthritis. Concurrently, computer image treating approaches for illness detection have emerged as a reliable and quickly expanding sector in biomedical research. The purpose of this suggested model is to use soft computing approaches to assess the investigative validity of iridology, an ancient complementary and substitute medicine practice, for the diagnosis of type II diabetes. Methods: The study included a mixed population of diabetic and non-diabetic participants. Infrared pictures of both eyes were taken. In the iris picture the area of interest (ROI), which matches to the location of the pancreatic organ on the iridology chart, was clipped. CNN and CNN were used to extract features such as accuracy, sensitivity, specificity, difference variance, contrast, homogeneity, correlation, dissimilarity, and entropy. The collection included both a test iris database and real-time iris photographs. The CNN model obtained an extraordinary validation accuracy of 98.7 %, while the SVM model achieved 68.39 %. The CNN model’s performance measures were 9 8. 7 % accuracy, 1 0 0 % recall (sensitivity), and 1 0 0 % F1-score. The false negative and false positive ratios were both 0.0, demonstrating that the CNN model is very reliable and precise.