Adaptive registration and atlas-based segmentation.

With the rapid developments in image registration techniques, registrations are applied not only as linear transforms but also as warping transforms with increasing frequency. The latter is especially suitable for soft tissue organs in the human body. When using Thin Plate Spline (TPS) as the warping transform of choice, a high degree of freedom (DOF) warping can be either manually specified by the placement of control points or implemented by using a dense grid of control points. The former leads to problems of operator bias, whereas the latter is very computationally expensive. Instead, we propose to automate the determination of DOF by locally increasing the density of control points in regions where they are needed rather than globally increasing the density of control points. Local estimates of Mutual Information (MI) and entropy are used to identify local regions requiring higher DOF. There have been significant efforts to build a probabilistic atlas of the brain and to use it for many common applications like segmentation and registration. Though the work related to brain atlases can be applied to non-brain organs, less attention has been paid to actually building an atlas for organs other than the brain. We present a method to construct a probabilistic atlas of an abdomen consisting of 4 organs (i.e., liver, kidneys and spinal cord). Using 32 non-contrast abdominal CT scans, 31 are mapped onto one individual scan using TPS as the warping transform and MI as the similarity measure. Except for an initial coarse placement of 4 control points by the operators, the MI based registration is automatic. Additionally, the four organs in each of the 32 CT data sets are manually segmented. The manual segmentations are warped onto the standard patient space using the same transform computed from their gray scale CT data set and a probabilistic atlas is calculated. Then the atlas is used to aid the segmentation of low contrast organs in additional 20 CT data sets not included in the atlas. By incorporating the atlas information into the Bayesian framework, segmentation results clearly showed improvements over a standard unsupervised segmentation method.Ph.D.Applied SciencesBiomedical engineeringComputer scienceElectrical engineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/123666/2/3096167.pd