Anaemia Detection using Deep Learning by processing Conjunctiva, Fingernails, Palm and Tongue Images

Abstract

Anemia poses a significant global public health challenge, particularly impacting children and pregnant women. According to the WHO, anemia is a highly prevalent condition, especially among patients in the emergency department. It occurs when the hemoglobin level falls below its normal threshold or when red blood cells are weakened or destroyed. This decreased number of red blood cells will lead to the paleness of the skin especially in the conjunctiva, palpable palm & fingernails. In current scenario anemia is being diagnosed by using invasive method. In this project, we propose the use of non-invasive techniques employing deep learning algorithms to aid in the diagnosis and detection of clinical diseases, specifically focusing on anemia detection. This approach aims to simplify the process of detection by analyzing images of specific areas such as the conjunctiva of the eye, palpable palm, fingernails, and tongue from diverse individuals. The processing and classification of these images are conducted using various deep learning algorithms including Convolutional Neural Networks (CNN), Mobilenet, and others. Our project paves way to identify the best algorithm that gives greatest accuracy rate and minimum time consumption for the detection of anemia. We have got the maximum accuracy rate of 99% in mobilenet and 83% in CNN.