SLOTMFound: Foundation-Based Diagnosis of Multiple Sclerosis Using Retinal SLO Imaging and OCT Thickness-maps

Multiple Sclerosis (MS) is a chronic autoimmune disorder of the central nervous system that can lead to significant neurological disability. Retinal imaging-particularly Scanning Laser Ophthalmoscopy (SLO) and Optical Coherence Tomography (OCT) provides valuable biomarkers for early MS diagnosis through non-invasive visualization of neurodegenerative changes. Thisstudy proposes a foundation-based bi-modal classification framework that integrates SLO images and OCT-derived retinal thickness maps for MS diagnosis. To facilitate this, we introduce two modality-specific foundation models-SLOFound andTMFound-fine-tuned fromtheRETFound-Fundus backbone using an independent dataset of 203 healthy eyes, acquired at Noor Ophthalmology Hospital with the Heidelberg Spectralis HRA+OCT system. This dataset, which contains only normal cases, was used exclusively for encoder adaptation and is entirely disjoint from the classification dataset. For the classification stage, we use a separate dataset comprising IR-SLO imagesfrom 32 MS patients and 70 healthy controls, collected at the Kashani Comprehensive MS Center in Isfahan, Iran. We first assess OCT-derived maps layer-wise and identify the Ganglion Cell-Inner Plexiform Layer (GCIPL) as the most informative for MS detection. Allsubsequent analyses utilize GCIPL thickness maps in conjunction with SLO images. Experimental evaluations on the MS classification dataset demonstrate that our foundation-based bi-modal model outperforms unimodal variants and a prior ResNet-based state-of-the-art model, achieving a classification accuracy of 97.37%, with perfect sensitivity (100%). These results highlight the effectiveness of leveraging pretrained foundationmodels, even when fine-tuned on limited data, to build robust, efficient, and generalizable diagnostic toolsforMS in medical imaging contexts where labeled datasets are often scarce.