@techreport{TR-IC-10-07,
   number = {IC-10-07},
   author  =  {Fábio  Augusto Menocci Cappabianco and Alexandre Xavier
                   Falcão  and  Clarissa  Lin  Yasuda  and  Jayaram K.
                   Udupa},
   title  =  {{MR-Image  Segmentation  of  Brain Tissues based on Bias
                   Correction and Optimum-Path Forest Clustering}},
   month = {March},
   year = {2010},
   institution = {Institute of Computing, University of Campinas},
   note = {In English, 23 pages.
    \par\selectlanguage{english}\textbf{Abstract}
       Magnetic resonance (MR) image segmentation of brain tissues has
       become  crucial  to  advance  research, diagnosis and treatment
       procedures  in Neurology. Despite the large number of published
       papers,  there is no standard method suitable for all cases. We
       present   here  a  fast,  accurate  and  robust  to  anatomical 
       variations  approach based on inhomogeneity/bias correction and
       clustering  by  optimum-path  forest  (OPF). The method assumes
       skull stripping and corrects voxel values in the brain based on
       local  estimations  of the white-matter intensities. We use two
       segmentation  steps  with  different  parameters  (e.g.,  graph 
       topologies   and  image  features).  The  first  separates  the 
       cerebrospinal  fluid  (CSF)  from  the  white-matter  (WM)  and 
       gray-matter  (GM)  tissues  and the second separates WM from GM
       voxels. In each step, random samples are uniformly estimated to
       form  a small unlabeled training set. Groups of training voxels
       with  similar  features are obtained by OPF clustering, a class
       label  is  assigned to each group and the labels are propagated
       to the remaining voxels. Each step is also repeated a few times
       (e.g.,  3)  to  take  the majority vote as the final label. The
       method  was  evaluated  using  several  data  sets  from  three 
       protocols;    control    subjects,   phantoms   and   patients;  
       quantitative  and qualitative evaluation methodologies; and two
       state-of-the-art approaches. The results show that it can solve
       brain  tissue segmentation in less than 5 minutes on modern PCs
       with   higher   accuracy   and  robustness  than  the  baseline 
       approaches.
  }
}