3-D HYBRID SEGMENTATION OF MEDICAL VOLUMES AND ITS VISUALIZATION

This paper demonstrates a 3-D hybrid algorithm for segmentation and visualization of medical volumes. The algorithm combines both K-Means and region growing technique in three dimensions to produces accurate segmentation results depending on a joint decision based on both color and spatial information. Although the 3-D K-means does not have any spatial information, it has been applied in 3-D domain to ensure that all K-means Output results of volume slices belong to the same clusters. 3-D region growing has been used to determine the region of interest of the volume. Sequential images of the visible man project have been used for testing the algorithm. The algorithm has been applied to the segmentation of 3D white matter of the brain as a region of interest. The performance of the algorithm are studied and compared with other hybrid algorithm results using Scatter Matrices and Separability criteria.