Publikation
Visual Acuity Assessment from Uni-and Multimodal Retinal Image Data using the Foundation Model MIRAGE
Caroline v Dresky; Claus von der Burchard Caroline v Dresky; Monty Santarossa; Julia Andresen; Marc S Seibel; Timo Kepp; Johann Roider; Heinz Handels
In: Handels H. et al. (Hrsg.). Bildverarbeitung für die Medizin 2026. Workshop Bildverarbeitung für die Medizin (BVM-2026), March 15-17, Lübeck, Germany, ISBN 978-3-658-51099-2, Springer, Wiesbaden, 3/2016.
Zusammenfassung
This study investigates image-based prediction of visual acuity, the most important measure of eye function,using the ophthalmic foundation model MIRAGE based on uni- and multimodal retinal image input. The three imaging modalities optical coherence tomography(OCT), infrared reflectance scanning laser ophthalmoscopy and fundus autofluorescence as wellas multi-modal combinations there of turn out to be suitable for image-based prediction of no/mild,moderate,or severe visual impairment for patients suffering from central serouschorioretinopathy(CSCR). Among these,OCT-based inputs achieve the best performance across evaluation metrics.Multimodal combinations, however, provide no additional benefit over the best unimodal approach, which suggests that OCT captures the most relevant information for estimating visual function in the presence of CSCR.
