Non-Invasive Blood Cholesterol Monitoring Using Artificial Neural Networks on Smartphone

Abstract

Artificial Intelligence is increasingly utilized, particularly in creating medical devices. This research aims to establish a model for an application that can identify and monitor cholesterol levels in a non-invasive manner. The research presents the innovative cholesterol detection (Choldet) application, which operates on smartphones to keep track of cholesterol levels. Cholesterol levels are identified through image processing of images captured from the skin on the hands. The research involved collecting hand skin images alongside corresponding lab blood cholesterol measurements. Samples were gathered from male and female participants aged between 20 and 65 years, amounting to a total of 57 sample images. To identify cholesterol levels, artificial neural networks (ANN) were employed to analyze hand textures, thereby generating a database. The training outcomes from the ANN were incorporated into the Choldet application for smartphones, utilizing the Application Programming Interface (API). The development of the Choldet application, designed for Android, can be installed on smartphones, allowing for real-time non-invasive monitoring of blood cholesterol levels. The efficacy of this novel evaluation method was validated using confusion matrix analysis, leading to an F-1 score of 80%, and analysis with Clarke-Error Grid Analysis (C-EGA) resulted in 92.98% in zone A indicating good accuracy. This application significantly contributes to facilitating early monitoring of health indicators, enabling individuals to assess their health status at any time, and aids in progressing health equipment technology.