Detection of Anemia Using Machine Learning: An analysis Using Eye Conjunctiva and Fingernail Images

Anemia, marked by a scarcity of hemoglobin, can impede the transportation of oxygen to vital organs, potentially resulting in diverse health issues. Conventional diagnosis with a full blood count is an intrusive, expensive, and time-consuming procedure. Non-invasive techniques, such as assessing the palpebral conjunctiva and nail color, provide alternate approaches but are subjective and susceptible to variation. This work suggests a thorough and automated method for identifying anemia without the need for intrusive procedures. It does this by utilizing machine learning algorithms on photographs of the eye conjunctiva and fingernails. We analyzed a deep neural network model specifically convolutional neural network (CNN) that includes sophisticated picture augmentation techniques to overcome the constraints of minimal datasets. In addition, computer vision methods were used for preprocessing and extracting features, while theoretically defined regularizers were applied to optimize the performance of the neural network. The model demonstrated a remarkable accuracy rate of ~86 and ~ 82% for eye conjunctiva images and fingernail images, respectively. This comprehensive strategy not only improves the precision and dependability of non-invasive anemia diagnosis but also establishes a foundation for uniform, expandable screening techniques.