, ? Lesion load and lesion count crucial in MS ? Part of diagnosis

, Subject to intra-/ inter-individual variability è Automatic segmentation is key

A. Thompson, Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria, The Lancet, vol.17, issue.2, pp.162-173, 2017.

, ? A huge number of automatic segmentation methods ? Tissue classification & outlier detection ? Machine learning (random forests, deep

?. T1, . T2, . Flair, and . Pd?,

?. Large-variety-of-implementations, ?. Gpu, . Matlab, and . Python, , 2008.

, ? Main drawbacks ? Possibility to adapt parameters to each patient ? Ground truth not well defined

. Styner, 3D Segmentation in the Clinic: A Grand Challenge II: MS lesion segmentation. Insight journal. Carass et al., 2017. Longitudinal multiple sclerosis lesion segmentation: Resource and challenge, An OFSEP and FLI challenge @ MICCAI, vol.148, pp.77-102, 2008.

, Evaluation objectives ? Evaluate algorithms developed in the community ? In a well defined computational framework (FLI)

F. Kremer, S. Hannoun, S. Vukusic, S. Dousset, and V. , ? Same set of parameters for all images ? With respect to a solid ground truth ? Additional objectives (OFSEP) ? Evaluate lesion segmentation algorithms for MS ? Fully automatic, Journal of Neuroradiology, vol.42, issue.3, pp.133-140, 2015.

, France Life Imaging-IAM node