To quantify structural connectivity integrity in multiple sclerosis (MS) patients with different clinical phenotypes. To simulate a disconnection due to T2 visible lesions and to verify its effect on network based measures.MS is a disease characterized by focal and diffuse damage to the white matter (WM). T2 lesions are disseminated in space and time. This makes the idea of modelling the brain as a network of roads and treating lesions as disconnecting events very attractive. Diffusion tensor imaging (DTI) and tractography enable to define and quantify structural damage in brain networks (1); it was already applied in MS proving that there are effects already in the early phase of the disease (2). With this study we wished to control for the variability coming from the application of tractography directly to damaged brain, to simulate the effect of lesions on the network and to explore difference between MS phenotypes.DT and dual echo turbo spin echo (TSE) MRI scans were obtained using a 3 tesla scanner from 239 MS patients (12 with a clinically isolated syndrome [CIS], 111 relapsing-remitting [RR] MS, 45 benign MS and 71 secondary progressive [SP] MS) and 131 healthy controls (HC). DT MRI scans were processed using FSL software (3) for distortions and motion correction (topup tool), tensor estimate and non-linear registration into the MNI space (fnirt tool). From the whole group of healthy controls, 80 were selected to create a reference group, with the aim of producing an atlas of WM tracts connecting cortical areas: after averaging the transformed DT components, tractography was run from all WM voxels (Diffusion Toolkit) and Trackvis software (www.trackvis.org) (4) was used to select those fibers connecting pairs of cortical areas defined by the AAL atlas (5). This allowed us to define edges and nodal degrees of a structural connectivity matrix. For 51 HCs and 239 MS patients, after the application of the WM tract atlas, the average fractional anisotropy (FA) value of each connection was calculated and saved within the corresponding element of the connectivity matrix. A similar procedure was used to produce the matrix where lesion load within each connection considered was saved. Such a lesional matrix was used to artificially produce disconnection in the connectivity matrix, so that when the lesion load was higher than 25 mm3, the FA values was set to 0 (Figures 1-2). Global and nodal network metrics were calculated (6).Compared to HC, MS patients showed significantly decreased (p<0.001) strength, assortativity, transitivity, global efficiency and increased average path length of the whole network. The nodes with the highest strength (hubs of the network) were the same for patients and controls. Among these, the postcentral, superior parietal, precuneus and cerebellar crus 1 e 2 had a significant reduced strength in SPMS patients compared to RRMS. When the disconnected matrices were considered, more nodes, including frontal superior, orbital, middle occipital areas, the thalamus and caudate nuclei were found to be decreased in strength in SPMS patients in comparison with RRMS (Figure 3) and the calcarine, middle occipital, superior parietal regions, precuneus and paracentral lobule in SPMS compared to benign MS.Global measures of structural connectivity were significantly different in MS patients compared with HC, showing an extended disruption of structural connectivity integrity in MS patients, thus confirming that MS can be considered as a disconnection syndrome. The pattern of hubs distribution appeared to be preserved in MS patients, with a preserved architecture of brain structural networks and a reduced strength of connections of the nodes identified as hubs. The simulation of disconnection due to T2 lesions allowed us to identify the role of lesions in determining abnormalities in structural connectivity not only comparing MS patients with healthy controls, but also assessing the differences across MS phenotypes.