Anemia Detection Using Convolutional Neural Network Based on Palpebral Conjunctiva Images
Endah Purwanti, Helsani Amelia, Winarno Winarno, Muhammad Arief Bustomi, Marcella Aurelia Yatijan, Revita Novianti Putri
Abstract
Anemia is a condition where the level of hemoglobin in the blood is below normal limits. Anemia can cause disruption of the oxygen transport system in the body which will affect the work of important organs such as the heart, kidneys and other organs. One of the factors causing the delay in treating anemia is that invasive procedures for examination are still considered frightening for some people. Patients with anemia, generally will experience pallor in the palpebral conjunctiva which indicates a decrease in hemoglobin levels in the blood. Palpebral conjunctival examination has the potential to be developed as a non-invasive and inexpensive alternative for anemia diagnosis. This study aims to detect anemia through image classification of the palpebral conjunctiva using a convolutional neural network (CNN). There are 3 CNN architectures used, namely AlexNet, ResNet-50 and MobileNetV2. The results showed that the performance accuracy of the AlexNet, ResNet-50 and MobileNetV2 architectures were 89.93%, 97.94%, 97.19%, respectively.