Spatial normalization for voxel-based lesion symptom mapping: impact of registration approaches

Abstract

Background: Voxel-based lesion symptom mapping (VLSM) assesses the relation of lesion location at a voxel level with a specific clinical or functional outcome measure at a population level. Spatial normalization, that is, mapping the patient images into an atlas coordinate system, is an essential pre-processing step of VLSM. However, no consensus exists on the optimal registration approach to compute the transformation nor are downstream effects on VLSM statistics explored. In this work, we evaluate four registration approaches commonly used in VLSM pipelines: affine (AR), nonlinear (NLR), nonlinear with cost function masking (CFM), and enantiomorphic registration (ENR). The evaluation is based on a standard VLSM scenario: the analysis of statistical relations of brain voxels and regions in imaging data acquired early after stroke onset with follow-up modified Rankin Scale (mRS) values.

Materials and methods: Fluid-attenuated inversion recovery (FLAIR) MRI data from 122 acute ischemic stroke patients acquired between 2 and 3 days after stroke onset and corresponding lesion segmentations, and 30 days mRS values from a European multicenter stroke imaging study (I-KNOW) were available and used in this study. The relation of the voxel location with follow-up mRS was assessed by uni- as well as multi-variate statistical testing based on the lesion segmentations registered using the four different methods (AR, NLR, CFM, ENR; implementation based on the ANTs toolkit).

Results: The brain areas evaluated as important for follow-up mRS were largely consistent across the registration approaches. However, NLR, CFM, and ENR led to distortions in the patient images after the corresponding nonlinear transformations were applied. In addition, local structures (for instance the lateral ventricles) and adjacent brain areas remained insufficiently aligned with corresponding atlas structures even after nonlinear registration.

Conclusions: For VLSM study designs and imaging data similar to the present work, an additional benefit of nonlinear registration variants for spatial normalization seems questionable. Related distortions in the normalized images lead to uncertainties in the VLSM analyses and may offset the theoretical benefits of nonlinear registration.

Keywords: VLSM; image registration; neuroimaging; spatial normalization; stroke.

Figure 1

(A) Illustration of imprecise alignment of atlas and patient brain after affine registration (see, for instance, the midline and the lesioned periventricular tissue). (B) After nonlinear registration, the alignment was improved. However, for the patient shown in (C, D), the nonlinear registration introduces a major midline shift (D), compared to the alignment after affine registration (C).

Figure 2

Top row: examples for midline shift scores [(A) score 1, midline shift limited to an area near the parietofrontal border, maintaining the original topology in other parts of the brain; (B) score 2, the thalami have shifted so far that involvement of other subcortical structures cannot be excluded; (C) score 3, wide-ranging topological distortions]. (D) Example measurements for the maximum midline shift (measurement 1) and the septum pellucidum shift (measurement 2). Bottom row: examples of abnormally small (E), normal-sized (F), and abnormally large (G) ventricles and a case showing a frontal blob (H). All examples show cases after nonlinear normalization.

Figure 3

Illustration of the VLSM maps for the four registration approaches [affine, nonlinear, nonlinear with cost function masking (CFM) and enantiomorphic registration] for two axial slices (z = 5 and z = 20) of the GIN atlas. (Top panel) Overlay of the stroke lesions on days 2–3 after stroke onset from all patients (N = 122; color bar: number of overlapping lesions, with a lower threshold of 10, that is, the minimum number of lesions required to include a voxel in the statistical analysis). An overlay without lower threshold is shown in the Supplementary Figure S1. (Middle) Results for univariate statistical analysis (color bar: Brunner-Munzel z-score after multiple comparison correction). (Bottom panel) Corresponding results for multivariate analysis (SCCAN; color bar: normalized negative correlation with multiple comparison correction, as returned by LESYMAP).

Figure 4

Illustration of the differences between CFM (A, C) and NLR (B, D). The atlas mask for the insular cortex is shown in red. While in NLR peripheral structures such as the Sylvian fissure are moved into the mask area, the mask covers a lesioned area after CFM-based registration. (A, B) as well as (C, D) show registration results from the same patient.

Figure 5

Illustration of the atlas right lateral ventricle mask being located inside the lesioned area after affine (A, C) and enantiomorphic registration (B, D). (A, B) as well as (C, D) show registration results from the same patient.

Figure 6

Comparison of FLAIR patient images after normalization using nonlinear FLAIR patient data to atlas registration (panels A–D) and nonlinear T1-weighted patient data to atlas registration (panels E–H). Distortions like frontal blobs and midline shifts after normalization are clearly reduced after using the T1-weighted data.