Visual acuity assessment from optical coherence tomography images using the foundation model RETFound

Visual acuity is the most important ophthalmologic measure of eye function. Its standard assessment method relies on a vision test on eye charts, while there is currently no method in clinical routine to derive visual acuity information from medical eye images. This can be explained by the lack of clearly defined structure-function correlations for all biomarkers visible in these images. Prior works already showed that deep learning can be used to predict visual impairment from medical eye images without the need for single biomarker identification. Beyond that, we show that fine-tuning an ophthalmic foundation model with a comparatively small dataset from clinical routine enables us to predict visual acuity based on only a single image. To this end, we adapt the recently published foundation model RETFound such that it outputs one of three visual impairment levels from optical coherence tomography images taken of patients with one of two different macular diseases. In this way, we achieve a satisfactory image-based prediction of no/mild, moderate or severe visual impairment.