Deep learning-based characterization of the retinal vessel pathology for prognostication of multiple sclerosis

• Applicants: Privatdozent Dr. Benjamin Knier; Dr. Martin Menten; Professor Dr. Daniel Rückert, Ph.D.
• Subject Area: Clinical Neurology; Neurosurgery and Neuroradiology ; Ophthalmology ; Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 528297171

Multiple sclerosis is an autoimmune inflammatory disease of the central nervous system. Besides inflammatory and neurodegenerative lesions in the brain and spinal cord, most patients reveal alterations of the retinal architecture. There is growing evidence that also changes of the retinal vasculature occur. By applying retinal optical coherence tomography angiography (OCTA), a novel, high-resolution and non-invasive imaging modality, we could recently show that a rarefication of retinal superficial vessels occurs during multiple sclerosis independent to a history of optic neuritis. These alterations are linked to a proinflammatory immune phenotype and may predict a worse disease course in patients with multiple sclerosis. The underlying pathology of retinal vessel alterations, however, is not yet understood. OCTA images contain a wealth of information and most of it is currently not available for research and clinical decision making due to missing analytical tools. Though OCTA generates three-dimensional imaging data, available analytical approaches only incorporate a two-dimensional en face image of the retina and depict rough estimates of the retinal vasculature. Moreover, the OCTA technique is susceptible to imaging artifacts limiting a standardized use in both neurological and ophthalmological diseases. Building upon our recent achievements, we will establish deep learning-based applications for automated OCTA quality control and will generate a pipeline to extract advanced quantitative metrics from three-dimensional OCTA images. Combined, these tools will enable us to in-depth characterize retinal vessel pathology during multiple sclerosis. As a final step, we will search for features within the retinal vasculature that predict the future disease activity and disease prognosis in patients with relapsing remitting multiple sclerosis. This project will generate novel tools for advanced analysis of OCTA images and automated quality control that will help to improve and expand the clinical and scientific application of OCTA. We are confident that these tools will enable us to understand fundamental novel aspects of the pathophysiology of multiple sclerosis. Ultimately, these findings may substantially improve clinical decision making and prognostication of affected patients. Beyond the application to autoimmune disease of the central nervous system, our developed tools may be of relevance and usefulness for neurodegenerative, neurovascular and ophthalmological disorders.

• DFG Programme: Research Grants