4D segmentation of brain MR images with constrained cortical thickness variation

Abstract

Segmentation of brain MR images plays an important role in longitudinal investigation of developmental, aging, disease progression changes in the cerebral cortex. However, most existing brain segmentation methods consider multiple time-point images individually and thus cannot achieve longitudinal consistency. For example, cortical thickness measured from the segmented image will contain unnecessary temporal variations, which will affect the time related change pattern and eventually reduce the statistical power of analysis. In this paper, we propose a 4D segmentation framework for the adult brain MR images with the constraint of cortical thickness variations. Specifically, we utilize local intensity information to address the intensity inhomogeneity, spatial cortical thickness constraint to maintain the cortical thickness being within a reasonable range, and temporal cortical thickness variation constraint in neighboring time-points to suppress the artificial variations. The proposed method has been tested on BLSA dataset and ADNI dataset with promising results. Both qualitative and quantitative experimental results demonstrate the advantage of the proposed method, in comparison to other state-of-the-art 4D segmentation methods.

Figure 1. The proposed framework for the 4D brain segmentation.

Figure 2. Illustration of temporal cortical thickness variation constraint.

The solid red (or blue) curves are the zero-level surface of (or ). is the cortical thickness measured from the inner surfaces (red curves). The dashed green arrows denote the registration operation to warp the corresponding thickness from the temporal neighborhoods to the current time-point. The dashed blue curves in the middle is the reasonable surface determined after measuring the cortical thickness difference between current time-point and two neighboring time-points.

Figure 3. Rows from top to bottom show the simulated intensity images, ground-truth segmentation, and segmentation results by CLASSIC and the proposed method, respectively.

Figure 4. Rows from top to bottom show the cortical thickness maps of ground truth (1st row), thickness maps by CLASSIC (2nd row), and thickness maps by the proposed method (3rd row).

Figure 5. Comparison of cortical thickness and tissue overlap with the ground truth by CLASSIC and the proposed method.

Left: The thickness maps. Right: Dice ratios of CLASSIC and the proposed method for WM and GM, respectively.

Figure 6. Cortical thickness maps derived by CLASSIC (the 1st row) and the proposed method (the 2nd row) on a randomly selected subject from the BLSA dataset.

The last two rows show the zoomed views of the first two rows.

Figure 7. The first row shows the original intensity images, and the next two rows show the segmentation results by CLASSIC and the proposed method, respectively.

The last two rows show the zoomed views of the 2nd and 3rd rows.

Figure 8. The average cortical thicknesses on 4 lobes of 10 elderly subjects from the BLSA dataset measured by CLASSIC (left column) and our proposed method (right column).

One curve indicates for one subject.

Figure 9. Cortical thickness maps derived by CLASSIC and the proposed method on the 4 reprehensive subjects from a) NC, b) S-MCI, c) P-MCI, and d) AD groups.

In each group, the upper row shows the results of CLASSIC and the lower row shows the proposed results. Circles indicate the region with dramatic thickness changes by CLASSIC, while consistent measurement achieved by our proposed method.

Figure 10. The average cortical thickness on 4 lobes derived by CLASSIC (the left column), the FreeSurfer (the middle column) and the proposed method (the right column) on all subjects from NC, S-MCI, P-MCI and AD groups.

Figure 11. Cortical thickness maps derived by FreeSurfer (the 1st row) and the proposed method (the 2nd row) on a randomly selected normal subject from the ADNI dataset.

Regions indicated by the dotted curves show dramatic longitudinal changes of cortical thickness by FreeSurfer, while much consistent results by the proposed method.

Figure 12. Average cortical thickness in 10 representative small cortical regions (ROIs) on all NC subjects.

These 10 ROIs include the left (L) and right (R) parts of Precentral, Frontal Sup, Postcentral, Temporal Sup, and Occipital Sup regions, where Sup denotes superior gyrus.