Effects of registration regularization and atlas sharpness on segmentation accuracy

Abstract

In non-rigid registration, the tradeoff between warp regularization and image fidelity is typically determined empirically. In atlas-based segmentation, this leads to a probabilistic atlas of arbitrary sharpness: weak regularization results in well-aligned training images and a sharp atlas; strong regularization yields a "blurry" atlas. In this paper, we employ a generative model for the joint registration and segmentation of images. The atlas construction process arises naturally as estimation of the model parameters. This framework allows the computation of unbiased atlases from manually labeled data at various degrees of "sharpness", as well as the joint registration and segmentation of a novel brain in a consistent manner. We study the effects of the tradeoff of atlas sharpness and warp smoothness in the context of cortical surface parcellation. This is an important question because of the increasingly availability of atlases in public databases, and the development of registration algorithms separate from the atlas construction process. We find that the optimal segmentation (parcellation) corresponds to a unique balance of atlas sharpness and warp regularization, yielding statistically significant improvements over the FreeSurfer parcellation algorithm. Furthermore, we conclude that one can simply use a single atlas computed at an optimal sharpness for the registration-segmentation of a new subject with a pre-determined, fixed, optimal warp constraint. The optimal atlas sharpness and warp smoothness can be determined by probing the segmentation performance on available training data. Our experiments also suggest that segmentation accuracy is tolerant up to a small mismatch between atlas sharpness and warp smoothness.

Fig. 1

Generative model for registration and segmentation. A is an atlas used to generate the label map L′ in some universal atlas space. The atlas A and label map L′ generate image I′. S is the smoothness parameter that generates random warp field R. This warp is then applied to the label map L′ and image I′ to create the label map L and the image I. We assume the label map L is available for the training images, but not for the test image. The image I is observed in both training and test cases.

Fig. 2

Strategies for exploring space of atlas sharpness and warp smoothness of a new image.

Fig. 3

Example of manual parcellation shown on a partially inflated cortical surface. In our data set, the neuroanatomist preferred gyral labels to sulcal labels. There are also regions where sulci and gyri are grouped together as one label, such as the superior and inferior parietal complexes.

Fig. 4

Parcellation accuracy as a function of warp smoothness. S is plotted on a log scale.

Fig. 5

Histogram of optimal warp smoothness S across subjects (MAMS).

Fig. 6

(a) Typical plot of Dice against smoothness S. (b) A noisy plot of Dice against smoothness S.

Fig. 7

Histogram of optimal S across structures (MAMS).

Fig. 8

Overall Dice versus smoothness. S is plotted on a log scale.

Fig. 9

Percentage improvement of SASS over FreeSurfer. The boundaries between parcellation regions are set to reddish-brown so that the different regions are more visible.

Fig. 10

Structure-specific parcellation accuracy for the left hemisphere. First column (dark blue) corresponds to FreeSurfer. Second (light blue), third (yellow) and fourth (brown) columns correspond to MAMS, SAMS and SASS respectively. (S = 1, α = 1). ^* indicates structures where SASS shows statistically significant improvement over FreeSurfer. There is no structure that becomes worse.

Fig. 11

Structure-specific parcellation accuracy for the right hemisphere. First column (dark blue) corresponds to FreeSurfer. Second (light blue), third (yellow) and fourth (brown) columns correspond to MAMS, SAMS and SASS respectively. (S = 1, α = 1). ^* indicates structures where SASS shows statistically significant improvement over FreeSurfer. There is no structure that becomes worse.