Multiple sclerosis (MS) is a disease of the central nervous system characterized by spatial and temporal dissemination of demyelination, axonal loss, and gliosis. It is the primary cause of non-traumatic disability in young patients, and its exact pathophysiology remains unknown. Characteristically, lesions in the white matter develop around a central vein¹.

Susceptibility-weighted imaging (SWI) can directly depict cerebral veins by exploiting venous blood oxygenation². No study has presented the dynamic interaction between the veins and the occurrence and early evolution of new lesions in MS. We describe dynamic changes in appearance of both the central vein and the surrounding new lesions in a unique cohort of patients followed at weekly intervals with a multimodal MR protocol^3,4.

Furthermore, we use an automated analysis method⁵ to quantify the observed dynamic changes in the central vein.

Five untreated relapsing-remitting MS patients underwent 8 weekly 3T MR acquisitions⁴. New lesions were detected on gadolinium-enhanced T1-weighted MRI as well as on SWI MRI. SWI was used to describe changes in the appearance of lesion and central vein during lesion formation.

Selection criteria for studied MS lesions: (1) Contrast-enhancement appears after the baseline acquisition and is no longer visible at the 8^th time-point, to ensure complete history of enhancement; (2) The veins on which the lesions are centred must have a minimal diameter of 2 pixels.

The selected SWI ROIs were processed with an unsupervised spatiotemporal clustering method⁵ in order to group pixels sharing similar intensity evolution patterns. No prior knowledge about data behavior was used to initialize or guide the analysis.

101 lesions out of 212 active lesions were enhancing after baseline and no longer enhancing at week 8. Out of these, only three new enhancing lesions in 2 of the 5 patients met the full inclusion criteria, enabling their characterization. In all three lesions transient hyperintensity was observed on SWI within the central vein at the time of enhancement. The automatic clustering obtained on each ROI identified several evolution patterns within the lesions, with one occurring at the lesion core between the veins and the white matter as shown in Fig. 1.

Moreover, it is shown in Figures 2, 3 and 4 that the specific grey level behaviors at the vein/lesion (V/L) interface occur either simultaneously to the lesion enhancement (ROI A) or before lesion enhancement in the SWI acquisitions (ROI B, C). These observations show the centrifugal development around a central vein of the observed MS lesions on highly resolved temporal data, which is consistent with previous work⁶.

The development of MS lesions around the vein was observed in histology long time ago especially by^7,8. They first hypothesised a centrifugal lesion development from the veins but they did not have access to the dynamic evolution of MS lesions. With the advent of MRI, longitudinal measurements demonstrating that lesions grow centrifugally were acquired⁶. Recently, vein size modification concomitant to plaque formation was shown⁹.

Our results suggest that a stenosis of the vein could be observed on SWI at the time of enhancement on post-gadolinium T1. These findings are also supported by another recent study making the hypothesis that the small apparent size of intralesional MS veins may reflect compression by the perivascular inflammatory cuff within active lesions or hardening of the vascular wall in chronic lesions⁹. Given that the intralesional vein narrowing does reverse at later time points, when the lesion appears larger, it is unlikely that partial-voluming would determine the intravascular hyperintensity.

Our study shows that vein narrowing is an early event that appears to precede blood-brain barrier disruption signified by contrast-enhancement. The observed transient vein narrowing and intraveinous hyperintensity in active lesions is consistent with previous cross-sectional findings⁹.

Intralesional vein stenosis is reversible and consistent with focal hypercellularity in the context of T-cell aggregation during MS lesion formation. Spatiotemporal clustering enables automated detection of venular and lesional changes over time. Our findings are not generalizable due to insufficient spatial resolution, yielding only 3 analyzable lesions. Future work at high temporal and spatial resolution is needed. Nevertheless, our findings are a further step in the understanding of MS plaque formation, and might guide future research into disrupting lesion formation at its earlier stages.