CLASSIC: Consistent Longitudinal Alignment and Segmentation for Serial Image Computing
MR brain image segmentation is a key processing step in many brain image analysis applications, e.g. morphometry, automatic tissue labeling, tissue volume quantification, image registration, and computer integrated surgery. Analysis of a series of 3-D data of the same subject captured at different time-points, i.e. of a 4-D image, is important in many neuroimaging studies that concentrate on normal development, aging, and evolution of pathology. Consistent segmentation is particularly important in the literature of aging and Alzheimer's Disease (AD) since subtle brain changes that might be indicative of early stages of underlying pathology must be estimated from serial MR images. However, existing 3-D segmentation algorithms may not provide adequate longitudinal stability for serial brain images since they process each image at a time. We have proposed a 4-D segmentation method that overcomes this limitation and significantly improves longitudinal stability of segmentation, referred to as CLASSIC: Consistent Longitudinal Alignment and Segmentation for Serial Image Computing. CLASSIC builds upon previous 3D methods for brain image segmentation, especially FANTASM, and extends them to 4D as well as to simultaneous estimation of segmentation and CLASSIC not only jointly segments longitudinal 3-D MR brain images of the same subject, but also estimates the longitudinal deformations in the image series, e.g. tissue atrophy. It iteratively performs two steps:
It jointly segments serial 3-D images using a 4-D Image-A daptive Clustering Algorithm based on the current estimate of the longitudinal deformations in the image series, It then refines these longitudinal deformations using the 4-D elastic warping algorithm ---4-D HAMMER.
In this way, we obtain both a longitudinally-consistent segmentation result and an estimate of longitudinal deformation of anatomy in a series of 3-D images. The 4-D image-adaptive clustering algorithm has the following three advantages: First, a new temporal consistency constraint term on the fuzzy membership functions is used in order to obtain temporally-consistent segmentation results. Second, the spatiotemporal constraints of fuzzy membership functions are made adaptive to the smoothness of the image, i.e. they are stronger in the regions that have more uniform image intensities, and vice versa, thus fuzzy membership functions are not necessarily overly smooth across tissue boundaries. Third, the clustering centers at each voxel location are adaptive to local image intensity variations. In this way, the proposed algorithm not only provides temporally-consistent segmentation results, but also adapts to local image intensity variations.
Results obtained from CLASSIC:
Segmentated images at year 1 (left) and st year 5(right) using 3-D segmentation given as inputs to CLASSIC. This was segmented using adpkmeans.
Consistent Segmented Images delivered as output by CLASSIC.
To download please visit our CLASSIC NITRC page.
- Zhong Xue, Dinggang Shen, Christos Davatzikos, "CLASSIC: Consistent Longitudinal Alignment and Segmentation for Serial Image Computing", Neuroimage, 388-399, Vol. 30, No. 2, 2006