Abstract—This work presents a multi-classification model of brain tissue in a simulation stage magnetic resonance imaging (MRI). Its purpose is to improve the quantification of brain pathologies and the planning of neurosurgeries. This paper shows the development and evaluation of the multi-class classification methods one-versus-one (1-v-1) and one-versus-all (1-v-r), based on support vector machines, selecting four classes of brain tissues in the sequences T1, T2, and DP (multispectral) MRI. The classified tissues were gray matter, white matter, cerebrospinal fluid (CSF) and a group of tissues called ‘the rest’, composed of bone, skin, muscle, fat, connective tissue and background. Finally, the performance of the classifier on different MRI slices was evaluated and showed an accuracy rate of 99.01 % using the one-versus-one model, and an average of 96.65% using the one-versus-all model
Index Terms—Magnetic resonance imaging (MRI), support vector machines, brain tissue.
The authors are with Espíritu Santo University of Brazil, Brazil (e-mail: msuarezb@ ecci.edu.co, cacifuentes@gmail.com, suarez.juan@ecci.edu.co, salinas.kathleen@gmail.com).
Cite: Suárez B. Marco, Cifuentes G. Carlos, Suárez B. Juan, and Salinas V. Kathleen, "Brain Tissue Model Classification for Telesurgery Navigation," International Journal of Machine Learning and Computing vol. 5, no. 1, pp. 68-72, 2015.