Image registration is an important enabling technology in medical image analysis. The current emphasis is on development and validation of application-specific non-rigid techniques, but there is already a plethora of techniques and terminology in use. In this paper we discuss the current state of the art of non-rigid registration to put on-going research in context and to highlight current and future clinical applications that might benefit from this technology. The philosophy and motivation underlying non-rigid registration is discussed and a guide to common terminology is presented. The core components of registration systems are described and outstanding issues of validity and validation are confronted.

Measures of overlap of labelled regions of images, such as the Dice and Tanimoto coefficients, have been extensively used to evaluate image registration and segmentation algorithms. Modern studies can include multiple labels defined on multiple images yet most evaluation schemes report one overlap per labelled region, simply averaged over multiple images. In this paper, common overlap measures are generalized to measure the total overlap of ensembles of labels defined on multiple test images and account for fractional labels using fuzzy set theory. This framework allows a single "figure-of-merit" to be reported which summarises the results of a complex experiment by image pair, by label or overall. A complementary measure of error, the overlap distance, is defined which captures the spatial extent of the nonoverlapping part and is related to the Hausdorff distance computed on grey level images. The generalized overlap measures are validated on synthetic images for which the overlap can be computed analytically and used as similarity measures in nonrigid registration of three-dimensional magnetic resonance imaging (MRI) brain images. Finally, a pragmatic segmentation ground truth is constructed by registering a magnetic resonance atlas brain to 20 individual scans, and used with the overlap measures to evaluate publicly available brain segmentation algorithms.

Volumetric magnetic resonance imaging analyses of 30 subjects were undertaken to quantify the global and temporal lobe atrophy in semantic dementia and Alzheimer's disease. Three groups of 10 subjects were studied: semantic dementia patients, Alzheimer's disease patients, and control subjects. The temporal lobe structures measured were the amygdala, hippocampus, entorhinal cortex, parahippocampal gyrus, fusiform gyrus, and superior, middle, and inferior temporal gyri. Semantic dementia and Alzheimer's disease groups did not differ significantly on global atrophy measures. In semantic dementia, there was asymmetrical temporal lobe atrophy, with greater left-sided damage. There was an anteroposterior gradient in the distribution of temporal lobe atrophy, with more marked atrophy anteriorly. All left anterior temporal lobe structures were affected in semantic dementia, with the entorhinal cortex, amygdala, middle and inferior temporal gyri, and fusiform gyrus the most severely damaged. Asymmetrical, predominantly anterior hippocampal atrophy was also present. In Alzheimer's disease, there was symmetrical atrophy of the entorhinal cortex, hippocampus, and amygdala, with no evidence of an anteroposterior gradient in the distribution of temporal lobe or hippocampal atrophy. These data demonstrate that there is a marked difference in the distribution of temporal lobe atrophy in semantic dementia and Alzheimer's disease. In addition, the pattern of atrophy in semantic dementia suggests that semantic memory is subserved by anterior temporal lobe structures, within which the middle and inferior temporal gyri may play a key role.

No abstract