Small vessel diseases: Ultra-realistic Microstructure computational Model to refine Individual Treatment – SUMMIT

Small vessel disease (SVD) accounts for 25% of strokes and is the second most common cause of dementia after Alzheimer's disease. Unlike other causes of stroke, SVD manifests itself years before the stroke by the accumulation of tissue damage. Although heterogeneous, these lesions appear on MRI as white matter hypersignals. Our objective is to characterise these lesions in vivo in order to develop new markers in the early stages of stroke, which are essential for testing new therapeutic approaches. The project will be carried out in 4 stages.
The first step will aim at establishing a computational model useful to characterise the microstructure of white matter hypersignals. This computational model will be based on large-scale numerical simulations of the microstructure of normal and pathological white matter and the associated diffusion MRI signal. Large learning and test databases will be built to train deep learning methods to establish a tool for decoding microstructural parameters from diffusion MRI data. The choice of suitable generative microstructural parameters will be based on a priori obtained from a histological study of normal and pathological white matter.
The second step will therefore lead to the collection and ex vivo scanning of human samples corresponding to regions of white matter of normal appearance or corresponding to hypersignals on FLAIR imaging sequences in order to carry out, firstly, an acquisition using ultra-high field MRI (17.2 Tesla) with a diffusion imaging protocol adapted for the SUMMIT model, and, secondly, a microscopy carried out on histological sections after cutting of the samples and immunostaining. The analysis of the diffusion MRI data using the SUMMIT computational model will be compared with the histological ground truth obtained by segmentation of the cells of the various populations constituting the white matter and reconstruction of the mesoscopic mapping of the cytoarchitectural parameters targeted by the SUMMIT model. This second step will thus provide a first ex vivo validation of the SUMMIT model.
In a third step, a joint in vivo and ex vivo validation will be conducted on the FibrAtlas cohort of 100 healthy elderly subjects who have donated their bodies to science, targeting 20 subjects with specific white matter hypersignals. The subjects will be follow a 3 Tesla MRI acquisition using the SUMMIT protocol to provide diffusion MRI data that will allow in vivo assessment of white matter microstructure at the level of hypersignals. At the death of the subjects, after collection of their brain and selection of 3 target subjects, post-mortem acquisitions will be conducted at 7T on the whole brain and at 17.2 Tesla on samples taken at the level of regions of the white matter presenting hypersignals. The microstructural maps obtained ex vivo from the SUMMIT model at high resolution will be compared to the maps obtained in vivo, thus completing the validation of the computational model.
The fourth phase of the project aims to evaluate the relevance of the SUMMIT model for the prediction of the clinical status of 50 patients recruited in the DHU-LAC cohort after an ischemic stroke related to SVD. These patients will undergo two 3 Tesla MRI acquisitions at a one year distance with the SUMMIT protocol in order to characterise the microstructure of white matter regions with hypersignals and to assess the prognostic potential of the SUMMIT microstructural markers identified from the first imaging session on the occurrence of new lesions observed during the second imaging session.