AI-Enabled Non-Invasive Anaemia Detection: Techniques, Challenges, and Future Directions

Abstract

This review examines the use of artificial intelligence (AI) for non-invasive anaemia detection across multiple physiological modalities. It reviews image-based approaches, including convolutional neural networks (CNNs) and transfer learning applied to palpebral conjunctiva, retinal fundus images, and smartphone camera data, alongside signal-based approaches such as photoplethysmography (PPG). The paper synthesizes existing methodologies, data acquisition techniques, and performance measures used across studies. Recent research indicates promising results in terms of accuracy and sensitivity, particularly for conjunctival imaging, retinal fundus photography, PPG-based methods, and smartphone-based screening. However, challenges including limited datasets, variability across populations, generalizability, and regulatory requirements remain. The review highlights the potential of AI-enabled non-invasive methods to support accessible and early anaemia screening while emphasizing the need for further clinical validation before widespread healthcare deployment.