A Machine Learning Method to Screen Anemia From Conjunctiva Images Taken by Smartphone

Anemia is one of the most common health issues in third world countries. According to the World Health Organization (WHO), nearly a quarter of the human population suffers from anemia. The prevalence of anemia as indicated by hemoglobin was found among more than a third of preschool children in Egypt. Symptoms associated with anemia make them more at risk of illness and infection. With the world′s current situation due to the pandemic; it’s very difficult to get access to medical facilities since they could be vulnerable to any diseases, not to mention the high costs and dangers of invasive methods, which are the current standard. The paper aims to provide a remote, non-invasive standardized approach that enables a quick screening to detect hemoglobin levels using smartphones and AI techniques. The image of the eye is captured and the eye conjunctiva is automatically extracted from the image as a Region of Interest (ROI). The ROI is then processed and features are extracted from it to train a machine-learning algorithm to determine if the patient is anemic or not. The model was run over 200 subjects and reached an accuracy of 85%, precision of 86%, and recall of 81%.