Meeting Banner
Abstract #3015

A Robust Brain Segmentation of Multispectral MRI Using a Supervised Hybrid Classifier

Jyh-Wen Chai1, 2, Clayton Chi-Chang Chen, Hsian-Min Chen3, Yi-Ying Wu1, Pei-Hua Lo1, Chu-Jing Song1, Yi-Hsin Tsai2, San-Kan Lee, Yen-Chien Ouyang4, Chein-I Chang5

1Department of Radiology, Taichung Veterans General Hospital, Taichung, Taiwan, Taiwan; 2College of Medicine, China Medical University, Taichung, Taiwan, Taiwan; 3Department of Biomedical Engineering, HungKuang University, Taichung, Taiwan, Taiwan; 4Department of Electrical Engineering, National Chung Hsin University, Taichung, Taiwan, Taiwan; 5Department of Computer Science and Electrical Engineering, University of Maryland, Baltimore, Baltimore, MD, United States

With no need of a prior knowledge about the tissue intensity or anatomical information, the supervised hybrid classifier was utilized for tissue classification of multispectral MRI in the native coordinate space by using only one small set of training samples. The preliminary results demonstrated that the proposed method can perform an accurate and reproducible brain volume morphometry of multispectral 3D high spatial resolution MRI in synthetic image data and in different groups of human subjects. This supervised method has shown potential in clinical applications, particularly promising for longitudinal studies of brain morphometry with multispectral MRI.