Google Health has announced a novel, non-invasive method to screen for anaemia.
In a study published by Nature Biomedical Engineering, researchers applied an artificial intelligence (AI) algorithm to more than 114,000 retinal fundus images from about 57,000 participants. When combining this with patient data, such as age and gender, the model was able to detect anaemia with 88% accuracy, compared to 74% when using metadata alone.
The model, developed using a dataset of primarily Caucasian participants, was validated on a separate Asian dataset and achieved comparable results.
Lead researcher Dr Akinori Mitani said the team sought to understand which parts of the eye contained signs of anaemia, finding much of the information comes from the optic disc and surrounding blood vessels.
Using deep learning will help discover new biomedical knowledge and extend Google’s existing work on cardiovascular risk factors, refractive error and progression of macular degeneration, concluded Dr Mitani.