ESPEYB21 15. Editors’ Choice Artificial Intelligence (4 abstracts)
15.10. A foundation model for generalizable disease detection from retinal images
Zhou Y , Chia MA , Wagner SK , Ayhan MS , Williamson DJ , Struyven RR , Liu T , Xu M , Lozano MG , Woodward-Court P , Kihara Y , Eye UKB , Vision C , Altmann A , Lee AY , Topol EJ , Denniston AK , Alexander DC & Keane PA.
Nature 622(7981): 156–163 (2023). PubMed: 37704728
In Brief: The authors present ‘RETFound’, a self-supervised learning approach that has so far analysed 1.6 million retinal images to enable disease detection. RETFound shows good accuracy for diagnosis and prognosis of sight-threatening eye diseases. It also contributes to incident prediction of complex systemic disorders such as heart failure and myocardial infarction.
The well-known saying ‘the eyes are the window to the soul’ means that you can guess a person’s emotions, thoughts, or inner self by looking into their eyes. This saying is hereby extended to ‘the eyes are a window to your complex disease risks’! ‘Oculomics’ is the concept that retinal images illustrate physical changes associated with (non-ophthalmological) systemic diseases. For example, changes in the optic nerve and inner retinal may share common determinants to central nervous system pathogenesis and neurodegeneration, and changes in retinal vascular geometry may be shared with vascular disorders of the heart and kidneys.
It is impressive enough that RETFound showed good ability to diagnose and classify eye diseases, such as diabetic retinopathy (AUROC =0.943, 0.822 and 0.884 on 3 separate datasets). More remarkably, it also showed good prediction for future diagnosis of ischaemic stroke, myocardial infarction, heart failure and Parkinson’s disease. Its self-supervised learning basis means that RETFound can continue to develop itself with additional input data. These academic researchers hope that this type of ‘medical foundation model’ will democratize access to medical artificial intelligence and accelerate progress towards its widespread clinical implementation.
Article tools
My recent searches
My recently viewed abstracts
Authors
ESPE Yearbook of Paediatric Endocrinology
ISSN 1662-4009 (Online)
© Bioscientifica 2024 |
Privacy policy |
Cookie settings
BiosciAbstracts
Bioscientifica Abstracts is the gateway to a series of products that provide a permanent, citable record of abstracts for biomedical and life science conferences.
Find out more