Deep Learning Based Real Time Diabetes Recognition

In the modern era, diabetes is one of a group of common diseases that affect people of different age groups, as diabetics suffer from the possibility of developing many heart diseases. This results in chest pain, which is known as angina pectoris, heart attacks, and stroke, as well as narrowing and hardening of the arteries. In this paper, a Convolutional Neural Network (CNN) type algorithm is proposed to detect diabetes by skin color, and characterizing the disease by RGB skin image. Several morphological operations were performed by shining laser light in colors (White, Green, and Red) on the skin to detect and extract features. Using a data set that was collected from patients with diabetes, the patients were divided into four categories to determine the percentage of high diabetes in the patient. An algorithm in the classification model for diabetes has achieved that the white led light channel has a largely excellent performance in classifying diabetic patients with a typical accuracy value of (92.24%) and the green led light with accuracy value of (81.12 %) and the red led light achieved an accuracy of (79.31%). Because of the limited penetration depth of the white led light channel in skin tissue, the blue channel is more effective compared to other channels such as green and red channels.