segmentation of ms lesions using entropy-based em algorithm and markov random fields分割的病变女士使用entropy-based em算法和马尔可夫随机域.pdf

J. Biomedical Science and Engineering, 2011, 4, 552-561 JBiSE doi:10.4236/jbise.2011.48071 Published Online August 2011 (http://www.SciRP.org/journal/jbise/). Segmentation of MS lesions using entropy-based EM algorithm and Markov random fields Ahmad Bijar1 1 2 1 , Mahdi Mohamad Khanloo , Antonio Peñalver Benavent , Rasoul Khayati 1Department of Biomedical Engineering, Shahed University, Tehran, Iran; 2Department de Estadística, Matemáticas e Informática, Universidad Miguel Hernández, Elche, Spain. Email: ahmad_bijar@; mahdi.khanlo@; a.penalver@umh.es; khayati@shahed.ac.ir Received 25 June 2011; revised 28 July 2011; accepted 8 August 2011. ABSTRACT inflammatory, demyelinating disease of the Central Nervous System (CNS) that primarily affects the white This paper presents an approach for fully automatic matter (WM) of the central nervous system. MS lesions segmentation of MS lesions in fluid attenuated inver- exhibit hypersignals in T2 weighted and hyposignals in sion recovery (FLAIR) Magnetic Resonance (MR) T1 weighted, with respect to normal white matter inten- images. The proposed method estimates a gaussian sities. Previous research has shown that the FLAIR mixture model with three kernels as cerebrospinal (Fluid Attenuated Inversion Recovery) sequence con- fluid (CSF), normal tissue and Multiple Sclerosis le- tains the most distinctive lesion-healthy tissue differen- sions. To estimate this model, an automatic En