Anemia Estimation Using Eye Conjunctiva Image: A Comparative Study Of Deep Learning Algorithms

Worldwide, anemia is a prevalent medical ailment that is defined by a lack of hemoglobin or red blood cells. Effective treatment and the avoidance of consequences depend on the early recognition and management of anemia. In this work, we propose comparing the Convolutional Neural Network, also called the CNN and EfficientNet models for estimation of anemia from eye conjunctival image. The conjunctiva is a readily accessible tissue that reflects systemic hemoglobin levels and is thus suitable for non-invasive anemia assessment. We collected a dataset of eye conjunctiva images from patients with varying degrees of anemia and healthy individuals. Our experimental results demonstrate the efficacy of both CNN and EfficientNet architectures in accurately estimating anemia from conjunctiva images, with the CNN approximately 85% for the model and roughly 90% for the EfficientNet model in terms of accuracy. However, we observed that the EfficientNet model outperformed the CNN model in terms of accuracy and computational efficiency. Our findings suggest that deep learning models, particularly EfficientNet, hold promise for automated anemia estimation, offering a valuable tool for healthcare professionals in resource limited settings where traditional diagnostic methods may be inaccessible or impractical.