The AI tool RETFound examines retinal scans to identify abnormalities, enabling disease diagnosis.
Artificial intelligence (AI) is ushering in a transformation across various domains, encompassing corporate and medical sectors alike. AI assists humans in task management and furnishes valuable analytical insights. In a recent breakthrough within the realm of artificial intelligence, scientists have unveiled an AI tool capable of diagnosing and predicting the risk of several health conditions, including heart failure.
This remarkable tool, known as RETFound, harnesses the power of AI to diagnose and forecast the likelihood of various health issues by scrutinizing individuals’ retinal images. The groundbreaking aspect of RETFound lies in its deployment of an innovative self-supervised learning method, eliminating the need for manual labeling of the 1.6 million retinal images used for training. This approach not only enhances efficiency but also reduces costs associated with the development of AI tools for medical diagnosis.
Much akin to how ChatGPT and similar large language models predict forthcoming words in a sentence based on contextual cues, RETFound delves into an extensive dataset of retinal photographs. It scans and predicts missing portions of images with remarkable accuracy. In simpler terms, over the course of millions of images, RETFound learns to discern the intricate features of retinas, laying the groundwork for its adaptability to a wide array of applications.
Pearse Keane, an ophthalmologist at Moorfields Eye Hospital NHS Foundation Trust in London and a co-author of the paper, elucidated that “Over the course of millions of images, the model somehow learns what a retina looks like and what all the features of a retina are.” This foundational knowledge endows the model with versatility, rendering it amenable to customization for diverse applications.
RETFound employs retinal scans to shed light on an individual’s cardiovascular health. Retinas offer a unique perspective as they provide a direct view of the body’s smallest blood vessels, comprising an intricate capillary network. Consequently, conditions affecting the circulatory system, such as hypertension, become observable through retinal images.
Additionally, retinas share similarities with the central nervous system, resembling the brain in certain aspects. This resemblance permits the assessment of neural tissue using retinal images. Nevertheless, interpreting these scans often necessitates specialized expertise, a gap that AI adeptly bridges.
The AI-driven tool, RETFound, has already assimilated the characteristics of a normal retina from unlabeled data. Leveraging this knowledge, it evaluates retinal scans to detect abnormalities, subsequently enabling disease diagnosis.
Scientists acknowledge that while the AI tool demonstrates impressive performance, particularly in the realm of ocular diseases like diabetic retinopathy, its accuracy in predicting systemic conditions such as heart attacks, heart failure, stroke, and Parkinson’s is not flawless but surpasses many other AI models. Continued refinement through the accumulation of data will likely enhance its accuracy over time.
