Klasifikasi Kadar Kolesterol Total melalui Citra Iris Mata Menggunakan Algoritma AlexNet

Abstract

The World Health Organization (WHO) noted that the death rate caused by cardiovascular disease (including coronary heart disease, hypertension, and stroke) reached 7.9% worldwide in 2022. One of the causes of cardiovascular disease is excess cholesterol levels in the blood. Reported from the statement of the Ministry of Health, Directorate General of Health Services in 2022, cholesterol sufferers in Indonesia reached 28% of the total population. so that action is needed to find out cholesterol levels as early as possible in order to prevent cardiovascular disease. However, cholesterol checks are currently still widely carried out by taking blood which causes pain from needle pricks and limited equipment, and not everyone has the equipment and knows how to use it. This study implements the AlexNet model to develop a non-invasive method for classifying total cholesterol levels quickly and accurately based on the branch of iridology. The data used are iris image data from Telkom University Dataverse and the results of independent collection at the Barus Health Center. The best model was obtained with a combination of hyperparameters including epoch 50, batch size 16 and learning rate 0.00001 with an average training time of 10 seconds per epoch. The results of the model training were implemented into a mobile application with an accuracy of 93% on the test data. The system can be used as a tool or can be used by the wider community as an easy and fast early detection method.