Deep Learning based Automatic Detection of Cholesterol
High cholesterol level remains the primary source of critical illness like Coronary Heart Disease (CHD), stroke and Peripheral Arterial Disease (PAD). A level of cholesterol greater than 240 mg/dL in the blood is considered a high cholesterol condition. In testing the cholesterol level, an individual had to fast for 9 to 12 hours before giving their blood samples for the cholesterol test. To overcome the difficulty, there is a need for alternative diagnostic techniques. Iridology is a branch of science in which the defective organs can be identified by analyzing the structure and pattern of the iris. So, iridology can be considered as an alternative diagnostic tool. In iridology, the presence of arcus senilis in the boundary of the iris confirms the high cholesterol condition. In this work, a CNN-based deep learning network is created to classify the eye image into three categories: high cholesterol, individual having cholesterol risk, and normal. The proposed system is tested and trained using the images collected from a private clinic, publicly available arcus senilis images, and CASIA iris interval. The proposed CNN architecture is trained from scratch and achieved an accuracy of 99.21%.
Authors: Sruthi K; Vijayakumar J; Thavamani S
Abstract link: https://ieeexplore.ieee.org/abstract/document/9588825