Nonlocal regularization for active appearance model: Application to medial temporal lobe segmentation

Abstract

The human medial temporal lobe (MTL) is an important part of the limbic system, and its substructures play key roles in learning, memory, and neurodegeneration. The MTL includes the hippocampus (HC), amygdala (AG), parahippocampal cortex (PHC), entorhinal cortex, and perirhinal cortex--structures that are complex in shape and have low between-structure intensity contrast, making them difficult to segment manually in magnetic resonance images. This article presents a new segmentation method that combines active appearance modeling and patch-based local refinement to automatically segment specific substructures of the MTL including HC, AG, PHC, and entorhinal/perirhinal cortex from MRI data. Appearance modeling, relying on eigen-decomposition to analyze statistical variations in image intensity and shape information in study population, is used to capture global shape characteristics of each structure of interest with a generative model. Patch-based local refinement, using nonlocal means to compare the image local intensity properties, is applied to locally refine the segmentation results along the structure borders to improve structure delimitation. In this manner, nonlocal regularization and global shape constraints could allow more accurate segmentations of structures. Validation experiments against manually defined labels demonstrate that this new segmentation method is computationally efficient, robust, and accurate. In a leave-one-out validation on 54 normal young adults, the method yielded a mean Dice κ of 0.87 for the HC, 0.81 for the AG, 0.73 for the anterior parts of the parahippocampal gyrus (entorhinal and perirhinal cortex), and 0.73 for the posterior parahippocampal gyrus.

Keywords: appearance modeling; label fusion; medial temporal lobe structures; nonlocal means; segmentation.

Figure 1

Processing pipeline for the appearance model‐based segmentation.

Figure 2

Impact of distance range ±d of φ on the performance of the two‐stage segmentation. Kappa (κ) values of 14 test subjects for d from 1.0 to 4.0 mm with steps of 0.5 mm. (a) AG, (b) HC, (c) EPC, and (d) PHC.

Figure 3

Impact of different SSIM values on the performance of the two‐stage segmentation. Kappa (κ) values of three randomly chosen test subjects for SSIM from 0.9 to 1.0. (a) AG, (b) HC, (c) EPC, and (d) PHC.

Figure 4

Impact of patch size on segmentation performance. Kappa (κ) values of 14 test subjects under different patch sizes. (a) HC, (b) EPC, and (c) PHC.

Figure 5

Impact of search window size on segmentation performance. Kappa (κ) values of 14 test subjects under different search window sizes. (a) HC, (b) EPC, and (c) PHC.

Figure 6

Two‐dimensional visualization of 3D segmentation results of three test subjects with average κ value: one test subject per row, and columns from left to right for test image and segmented contours from manual label and three automatic segmentation methods. κ values shown under each graph. The segmented contours of different structures rendered on top of the corresponding T1‐weighted test MR image with color coding: purple for HC, blue for AG, sky blue for EPC, and white for PHC.

Figure 7

Two examples showing the two‐stage segmentation mismatching the true structure boundary of the left HC: One example per row—upper row for example no. 1, lower row for example no. 2; Each example shows four sagittal slices through the medial temporal lobe. Color coding—purple for the manual labeled contour; sky blue for the automatically segmented contour.

Figure 8

Volumetric comparison between the two‐stage segmentation results and manual labels for the HC and AG (volumes normalized in stereotaxic space).

Figure 9

Volumetric comparison between the two‐stage segmentation results and manual labels for the EPC and PHC (volumes normalized in stereotaxic space).