Atlas-based hippocampus segmentation in Alzheimer's disease and mild cognitive impairment

Abstract

This study assesses the performance of public-domain automated methodologies for MRI-based segmentation of the hippocampus in elderly subjects with Alzheimer's disease (AD) and mild cognitive impairment (MCI). Structural MR images of 54 age- and gender-matched healthy elderly individuals, subjects with probable AD, and subjects with MCI were collected at the University of Pittsburgh Alzheimer's Disease Research Center. Hippocampi in subject images were automatically segmented by using AIR, SPM, FLIRT, and the fully deformable method of Chen to align the images to the Harvard atlas, MNI atlas, and randomly selected, manually labeled subject images ("cohort atlases"). Mixed-effects statistical models analyzed the effects of side of the brain, disease state, registration method, choice of atlas, and manual tracing protocol on the spatial overlap between automated segmentations and expert manual segmentations. Registration methods that produced higher degrees of geometric deformation produced automated segmentations with higher agreement with manual segmentations. Side of the brain, presence of AD, choice of reference image, and manual tracing protocol were also significant factors contributing to automated segmentation performance. Fully automated techniques can be competitive with human raters on this difficult segmentation task, but a rigorous statistical analysis shows that a variety of methodological factors must be carefully considered to insure that automated methods perform well in practice. The use of fully deformable registration methods, cohort atlases, and user-defined manual tracings are recommended for highest performance in fully automated hippocampus segmentation.

Figure 1

Schematic view of atlas-based segmentation. An intensity transformation and geometric transformation are estimated to register the atlas image to the subject image; the geometric transformation is applied to the atlas mask in order to estimate the subject mask.

Figure 2

Example image deformations produced by fully-deformable, semi-deformable, and affine registration techniques. The moving image is registered to the stationary image using each of the 7 algorithms we analyze. The colored dots show the geometric positions of voxels in the shown slice of the moving image before and after deformation by each of the methods. The transformation produced by the AIR affine method and SPM affine method were almost identical to that of the FSL affine method.

Figure 3

Evaluating consistency between masks using overall and sectional overlap. A ground-truth subject mask and estimated subject mask are shown in light and dark gray. Figure 3d) : Voxels in red overlap between the ground-truth and the estimate. Overlap ratio measures the ratio between the volume of the red region and the volume of the combined red and gray regions. Figure 3f) : The green bars split the hippocampus voxels into axis-parallel sections. In sectional analysis, overlap ratio is computed for each section independently.

Figure 4

Points on the left hippocampus in all 19 MCI subjects are shown projected onto the XZ plane of the image. Note that all the hippocampi share the same rough initial orientation in this plane.

Figure 5

Overlap ratio as a function of disease state, registration method category, and side of the brain for the 54 images using cohort atlases.

Figure 6

Overlap ratio as a function of disease state, registration method, manual tracing, and side of the brain for the 54 images using standard atlases.

Figure 7

Left: Overlap ratio for cohort-atlas-based and standard-atlas-based segmentation using Chen’s fully-deformable registration method. Right: p values and the contrast correlation r_contrast for F tests between factor levels in the mixed-effects model.

Figure 8

Overlap ratio between manual and automated segmentations (automatic vs. manual) and between pairs of manual segmentations (manual vs. manual).

Figure 9

Overlap ratio delineated along posterior-anterior line (left), inferior-superior line (middle), and medial-lateral line (right), for the Chen fully-automated registration method on the right hippocampus in MCI images. Similar patterns of overlap ratio distribution are seen for other registration methods, other disease states, and the left hippocampus. See text for details.