ARTMED-1043; No of Pages 15 Please cite this article in press as: Scherrer B, et al. Agentification of Markov model-based segmentation: Application to magnetic resonance brain scans. Artif Intell Med (2008), doi:10.1016/j.artmed.2008.08.012 Agentification of Markov model-based segmentation: Application to magnetic resonance brain scans Benoit Scherrer a,b,c , Michel Dojat a,c, * , Florence Forbes c,d , Catherine Garbay b,c a Grenoble Institut des Neurosciences (GIN), Centre de Recherche Institut national de la sante´ et de la recherche medicale (Inserm) U 836, E5 Chemin Fortune´ Ferrini, La Tronche, BP 170, 38042 Grenoble Cedex 09, France b Centre National de la Recherche Scientifique (CNRS), Grenoble Informatics Laboratory (LIG), Batiment IMAG B, 385 avenue de la Bibliothe`que, 38400 Saint Martin d’He`res, France c Universite´ Joseph Fourier, BP 53, 38041 Grenoble Cedex 9, France d Laboratoire Jean Kuntzmann, MISTIS, ZIRST, 655 avenue de l’Europe, Montbonnot, 38334 Saint Ismier Cedex, France Received 14 December 2007; received in revised form 22 August 2008; accepted 22 August 2008 Artificial Intelligence in Medicine (2008) xxx, xxx—xxx http://www.intl.elsevierhealth.com/journals/aiim KEYWORDS Medical imaging; Markov random field; Multiagents system; Distributed expectation maximization; Magnetic resonance brain scan segmentation Summary Objective: Markov random field (MRF) models have been traditionally applied to the task of robust-to-noise image segmentation. Most approaches estimate MRF para- meters on the whole image via a global expectation—maximization (EM) procedure. The resulting estimated parameters are likely to be uncharacteristic of local image features. Instead, we propose to distribute a set of local MRF models within a multiagent framework. Materials and methods: Local segmentation agents estimate local MRF models via local EM procedures and cooperate to ensure a global consistency of local models. We demonstrate different types of cooperations between agents that lead to additional levels of regularization compared to the standard label regularization provided by MRF. Embedding Markovian EM procedures into a multiagent paradigm shows interesting properties that are illustrated on magnetic resonance (MR) brain scan segmentation. * Corresponding author at: Grenoble Institut des Neurosciences (GIN), Centre de Recherche Institut national de la sante ´ et de la recherche medicale (Inserm) U 836, E5 Chemin Fortune ´ Ferrini, La Tronche, BP 170, 38042 Grenoble Cedex 09, France. Tel.: +33 4 56 52 06 01; fax: +33 4 56 52 05 98. E-mail addresses: benoit.scherrer@imag.fr (B. Scherrer), michel.dojat@ujf-grenoble.fr (M. Dojat), florence.forbes@inrialpes.fr (F. Forbes), catherine.garbay@imag.fr (C. Garbay). 0933-3657/$ — see front matter # 2008 Published by Elsevier B.V. doi:10.1016/j.artmed.2008.08.012