A. Akhondi-asl, L. Hoyte, M. E. Lockhart, and S. K. Warfield, A logarithmic opinion pool based staple algorithm for the fusion of segmentations with associated reliability weights, IEEE Transactions on Medical Imaging, vol.33, 1997.

R. Bellman, Adaptive Control Processes: A Guided Tour, 1961.

O. Commowick, A. Istace, M. Kain, B. Laurent, F. Leray et al., Objective Evaluation of Multiple Sclerosis Lesion Segmentation using a Data Management and Processing Infrastructure, Scientific Reports, vol.8, 2018.
URL : https://hal.archives-ouvertes.fr/inserm-01895603

O. Commowick, N. Wiest-daesslé, and S. Prima, Block-Matching Strategies for Rigid Registration of Multimodal Medical Images, 9th IEEE International Symposium on Biomedical Imaging (ISBI'2012, pp.700-703, 2012.
URL : https://hal.archives-ouvertes.fr/inserm-00681610

P. Coupe, P. Yger, S. Prima, P. Hellier, C. Kervrann et al., An optimized blockwise nonlocal means denoising filter for 3-d magnetic resonance images, IEEE Transactions on Medical Imaging, vol.27, pp.425-441, 2008.
URL : https://hal.archives-ouvertes.fr/inserm-00169658

N. Courty, R. Flamary, A. Habrard, and A. Rakotomamonjy, Joint Distribution Optimal Transportation for Domain Adaptation, 2017.
URL : https://hal.archives-ouvertes.fr/hal-01620589

N. Courty, R. Flamary, D. Tuia, and A. Rakotomamonjy, Optimal transport for domain adaptation, 2015.
URL : https://hal.archives-ouvertes.fr/hal-01170705

B. B. Damodaran, B. Kellenberger, R. Flamary, D. Tuia, and N. Courty, Deepjdot: Deep joint distribution optimal transport for unsupervised domain adaptation, 2018.
URL : https://hal.archives-ouvertes.fr/hal-01956356

R. M. Dudley, The speed of mean glivenko-cantelli convergence, Ann. Math. Statist, vol.40, pp.40-50, 1969.

R. Flamary and N. Courty, Pot python optimal transport library, 2017.

F. Galassi, O. Commowick, E. Vallee, and C. Barillot, Voxel-wise comparison with a-contrario analysis for automated segmentation of multiple sclerosis lesions from multimodal mri, Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries, vol.11383, p.11, 2018.
URL : https://hal.archives-ouvertes.fr/inserm-01888928

F. Galassi, O. Commowick, E. Vallee, and C. Barillot, Deep learning for multi-site ms lesions segmentation: two-step intensity standardization and generalized loss function, IEEE International Symposium on Biomedical Imaging hal-02052250, 2019.
URL : https://hal.archives-ouvertes.fr/hal-02052250

D. García-lorenzo, S. Francis, S. Narayanan, D. L. Arnold, and D. L. Collins, Review of automatic segmentation methods of multiple sclerosis white matter lesions on conventional magnetic resonance imaging, Medical Image Analysis, vol.17, pp.1-18, 2013.

A. Genevay, L. Chizat, F. Bach, M. Cuturi, and G. Peyré, Sample complexity of sinkhorn divergences, of Proceedings of Machine Learning Research, vol.89, pp.1574-1583, 2019.

R. Gopalan, R. Li, C. , and R. , Domain adaptation for object recognition: An unsupervised approach, 2011 International Conference on Computer Vision, pp.999-1006, 2011.

F. Isensee, P. Kickingereder, W. Wick, M. Bendszus, and K. H. Maier-hein, Brain tumor segmentation and radiomics survival prediction: Contribution to the BRATS 2017 challenge, 2018.

L. V. Kantorovich, On the transfer of masses, Dokl. Acad. Nauk, vol.37, pp.7-8, 1942.

W. M. Kouw and M. Loog, A review of single-source unsupervised domain adaptation, 2019.

A. Segmentation, . Multiple, . Lesions, and . Mri-data,

K. Kushibar, S. Valverde, S. González-villà, J. Bernal, M. Cabezas et al., Supervised domain adaptation for automatic sub-cortical brain structure segmentation with minimal user interaction, 2019.

J. V. Manjón and P. Coupé, Volbrain: An online mri brain volumetry system, Frontiers in Neuroinformatics, vol.10, p.30, 2016.

J. A. Onofrey, D. I. Casetti-dinescu, A. D. Lauritzen, S. Sarkar, R. Venkataraman et al., , 2019.

, Generalizable multi-site training and testing of deep neural networks using image normalization, 2019 IEEE 16th International Symposium on Biomedical Imaging, pp.348-351, 2019.

O. Ronneberger, P. Fischer, and T. Brox, U-net: Convolutional networks for biomedical image segmentation, 2015.

S. Sankaranarayanan, Y. Balaji, C. D. Castillo, and R. Chellappa, Generate to adapt: Aligning domains using generative adversarial networks, 2017.

K. J. Smith and W. I. Mcdonald, The pathophysiology of multiple sclerosis the mechanisms underlying the production of symptoms and the natural history of the disease, Philosophical Transactions of the Royal Society B: Biological Sciences, vol.354, pp.1649-1673, 1999.

N. J. Tustison, B. B. Avants, P. A. Cook, Y. Zheng, A. Egan et al., N4itk: Improved n3 bias correction, IEEE Transactions on Medical Imaging, vol.29, 2010.

E. Tzeng, J. Hoffman, K. Saenko, D. , and T. , Adversarial discriminative domain adaptation, 2017.

S. Valverde, M. Cabezas, E. Roura, S. González-villà, D. Pareto et al., Improving automated multiple sclerosis lesion segmentation with a cascaded 3d convolutional neural network approach, NeuroImage, vol.155, pp.159-168, 2017.

C. Villani, Optimal transport -Old and new, vol.338, 2008.

A. Segmentation, . Multiple, . Lesions, and . Mri-data,

, Figure 3: Variability in MS lesion volume and number. Lesion load per patient per site (left) and Number of lesions per patient per site (right)

A. Segmentation, . Multiple, . Lesions, and . Mri-data, , p.7

A. Segmentation, . Multiple, . Lesions, and . Mri-data, Target, p.8

, Each row corresponds to a combination of source and target. Dice score (left column) and F1 score (right column) are computed with no adaptation (No Seg-JDOT) and with Seg-JDOT, where the direction of the domain adaptation is indicated (07, 08, or 01). For each combination of source and target, performances are given for all the four testing sites. Each point is a patient of a given site; performances of a patient with and without Seg-JDOT are tracked, Performance of Seg-JDOT with single-site source and single-site target domain adaptation, vol.4

A. Segmentation, . Multiple, . Lesions, and . Mri-data, , p.1

, Each row corresponds to a combination of source and target. Dice score (left column) and F1 score (right column) are computed with no adaptation (No Seg-JDOT) and with Seg-JDOT, where the direction of the domain adaptation is indicated (07, 08, or 01). For each combination of source and target, performances are given for all the four testing sites. Each point is a patient of a given site; performances of a patient with and without Seg-JDOT are tracked, Performance of Seg-JDOT with multi-site source and single-site target domain adaptation, vol.7