Research on K-means clustering segmentation method for MRI brain image based on selecting multi-peaks in gray histogram.
- Author:
Zhaoxue CHEN
;
Haizhong YU
;
Hao CHEN
- Publication Type:Journal Article
- MeSH:
Algorithms;
Brain;
anatomy & histology;
Cluster Analysis;
Humans;
Magnetic Resonance Imaging;
Neuroimaging
- From:
Journal of Biomedical Engineering
2013;30(6):1164-1170
- CountryChina
- Language:Chinese
-
Abstract:
To solve the problem of traditional K-means clustering in which initial clustering centers are selected randomly, we proposed a new K-means segmentation algorithm based on robustly selecting 'peaks' standing for White Matter, Gray Matter and Cerebrospinal Fluid in multi-peaks gray histogram of MRI brain image. The new algorithm takes gray value of selected histogram 'peaks' as the initial K-means clustering center and can segment the MRI brain image into three parts of tissue more effectively, accurately, steadily and successfully. Massive experiments have proved that the proposed algorithm can overcome many shortcomings caused by traditional K-means clustering method such as low efficiency, veracity, robustness and time consuming. The histogram 'peak' selecting idea of the proposed segmentootion method is of more universal availability.