Diabetes Diagnosis Using Pre-Trained Deep Learning Model

Diabetes Diagnosis Using Pre-Trained Deep Learning Model

S. Navya1, B. Keerthana2, G. Kiran3, D. Harshitha4, V. Vineeth5

1Assistant Professor in Department of CSE, Raghu Engineering College 2B.Tech Computer Science and Engineering (CSE-AIML), Raghu Institute of Technology 3B.Tech Computer Science and Engineering (CSE-AIML), Raghu Institute of Technology 4B.Tech Computer Science and Engineering (CSE-AIML), Raghu Institute of Technology 5B.Tech Computer Science and Engineering (CSE-AIML), Raghu Institute of Technology

                                                                          ***

Abstract: Since prevention is always preferable to treatment, disease prediction is essential for efficient prevention and management. One of the most serious diseases in the world is diabetes. Significant predictive potential of diabetes is shown by BMI, waist circumference, serum urate levels, GGT, waist-hip ratio, HDL cholesterol, family history of diabetes, and plasma glucose. Human stress can impact blood glucose regulation, metabolic activity, and insulin resistance, all of which can have an impact on diabetic levels. This article discusses how to anticipate the disease using pre-trained deep learning models.  ResNet is among them. It is an image classification model that is specifically designed to analyze retinal images and find indications of diabetic retinopathy, a diabetes condition that can cause visual problems. To improve the quality of training images, methods like augmentation, normalization, and scaling are applied. A classification or regression score showing the degree of severity is the model's output. Its deeply layered structure facilitates the efficient learning of intricate diabetic retinopathy patterns. The program predicts if a patient has diabetes and lowers the chance of visual loss through early detection based on the patient's current medical record.

Keywords: diabetic retinopathy, deep layered structure, deep learning, ResNet, diabetic symptoms, diabetes prediction, and pre-trained model.