Arthur W. Toga and Paul M. Thompson
Laboratory of Neuro Imaging, Department of Neurology, Division of Brain Mapping, UCLA School of Medicine, Los Angeles, California 90095
The ability to measure and understand the rich complexity of brain structure and function often requires comparison against some index, standard or alternative representation. In this chapter, we review the current mathematical and computational research on image warping algorithms. Warping algorithms are widely used in the analysis of brain image data, since they can be used to integrate brain data from subjects whose brain geometry is different.
Applications. The variety of warping approaches is equaled by their range of applications. We describe classes of algorithms each suited to handling particular kinds of information. Within classes of warps, there are several mathematical strategies for achieving a solution. Furthermore, there is no obvious hierarchy with one better than another. Each has strengths and weaknesses, each may perform particularly well given certain kinds of data, and each often complements others in performance, application and strategy.
Warping algorithms also permit the generation of brain atlases and brain maps. The statistical product is often best accompanied by visualization. Incorporating interim displays and animating the spatial transformation has remarkable power for elucidating complex and significant morphological change. Finally, we describe how warping across modalities and with multiple dimensions has enabled the synthesis of comprehensive reference systems that describe brain structure and function within whole populations.
Key Words: Brain Mapping, 3D, image registration, deformable templates, elastic matching, Magnetic Resonance Imaging
|RESUME| E-MAIL ME| PERSONAL HOMEPAGE| PROJECTS|