Estimating Insulin Dose for Diabetic Patients from Food Images Using Artificial Intelligence

Abstract

This article presents an intelligent system designed to assist type 1 diabetic patients in accurately estimating their insulin dosage by analyzing food images. Leveraging advanced artificial intelligence models, including computer vision techniques and deep learning, the system identifies food components, calculates nutritional values, and provides personalized insulin dose recommendations based on various health factors. This research automates a critical aspect of diabetes management, thereby improving patient quality of life and mitigating risks associated with manual calculations. Experimental results demonstrate high accuracy in food recognition (96% classification accuracy), nutritional estimation (MAE of 4.5g for carbohydrates), and insulin dose prediction (RMSE of 1.2 units), outperforming traditional methods.