Anisotropic kernels for coordinate-based meta-analyses of neuroimaging studies.

Radua J, Rubia K, Canales-Rodríguez EJ, Pomarol-Clotet E, Fusar-Poli P, Mataix-Cols D. · 2014

Read Anisotropic kernels for coordinate-based meta-analyses of neuroimaging studies. by Radua J, Rubia K, Canales-Rodríguez EJ, Pomarol-Clotet E, Fusar-Poli P, Mataix-Cols D. (2014) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc3919071.txt.

Introduction

Joaquim Radua and colleagues propose a better way to combine brain-scan results across studies in Frontiers in Psychiatry in 2014. Coordinate-based meta-analyses rebuild a brain map for each study by spreading a small kernel of effect around every reported peak. Standard kernels are round: a nearby voxel's value depends only on its distance from the peak, as if effects spread evenly in all directions.

The authors argue that real brain effects follow the brain's own wiring, so voxels that vary together with the peak across people are likelier to share its effect. Their anisotropic kernels weight each neighbor by its spatial correlation with the peak instead. The paper builds the needed correlation templates and implements the method in the ES-SDM software, while noting it could be added to other meta-analysis tools.

Catalog record

CREATORRadua J, Rubia K, Canales-Rodríguez EJ, Pomarol-Clotet E, Fusar-Poli P, Mataix-Cols D.
YEAR2014
COLLECTIONPAPERS
FORMATTXT
FILE SIZE33 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC3919071

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Anisotropic Kernels for Coordinate-Based Meta-Analyses of Neuroimaging Studies By Joaquim Radua, Katya Rubia, Erick Jorge Canales-Rodríguez, Edith Pomarol-Clotet, Paolo Fusar-Poli, David Mataix-Cols Published in: Frontiers in Psychiatry DOI: 10.3389/fpsyt.2014.00013 ABSTRACT Peak-based meta-analyses of neuroimaging studies create, for each study, a brain map of effect size or peak likelihood by convolving a kernel with each reported peak. A kernel is a small matrix applied in order that voxels surrounding the peak have a value similar to, but slightly lower than that of the peak. Current kernels are isotropic, i.e., the value of a voxel close to a peak only depends on the Euclidean distance between the voxel and the peak. However, such perfect spheres of effect size or likelihood around the peak are rather implausible: a voxel that correlates with the peak across individuals is more likely to be part of the cluster of significant activation or difference than voxels uncorrelated with the peak. This paper introduces anisotropic kernels, which assign different values to the different neighboring voxels based on the spatial correlation between them. They are specifically developed for effect-size signed differential mapping (ES-SDM), though might be easily implemented in other meta-analysis packages such as activation likelihood estimation (ALE). The paper also describes the creation of the required correlation templates for gray matter/BOLD response, white matter, cerebrospinal fluid, and fractional anisotropy. Finally, the new method is validated by quantifying the accuracy of the recreation of effect size maps from peak information. This empirical validation showed that the optimal degree of anisotropy and full-width at half-maximum (FWHM) might vary largely depending on the specific data meta-analyzed. However, it also showed that the recreation substantially improved and did not depend on the FWHM when full anisotropy was used. Based on these results, we recommend the use of fully anisotropic kernels in ES-SDM and ALE, unless optimal meta-analysis-specific parameters can be estimated based on the recreation of available statistical maps. The new method and templates are freely available at http://www.sdmproject.com/. INTRODUCTION In order to help summarize and integrate the results of the ever-growing number of neuroimaging studies, some groups have developed methods to conduct voxel-based meta-analyses solely relying on the information reported in the papers, namely the peaks of the clusters where there were statistically significant activations or where patients and controls showed statistically significant differences. Activation likelihood estimation (ALE) (1–4), (effect-size) signed differential mapping (ES-SDM) (5–7) and (multilevel) kernel density analysis (M-KDA) (8, 9) are commonly used methods that have already been applied to meta-analyze a wide range of normal brain functions (10–12) and abnormalities in neurological (13–15) and psychiatric disorders (16–18). As briefly introduced in Figure 1, these methods differ substantially in their algorithms [for a deeper review, see Ref. (19)], but one characteristic they share is that all convolve an isotropic kernel with the peak. In this context, a “kernel” is a small matrix convolved with the peaks in order that voxels surrounding a peak have a value similar to, but slightly lower than that of the peak. An “isotropic kernel” is identical in all directions. In simple terms, the effect size of a voxel close to a peak would only depend on the effect size of the peak and the Euclidian distance between the voxel and the peak (Figure 2). All voxels at 1 cm of a peak would have the same effect size, independently of whether they are in the same brain region or not. [Figure: Main steps of activation likelihood estimation (ALE) and effect-size signed differential mapping (ES-SDM). ALE (left approach) aims to estimate the likelihood that a peak lies in any given voxel. To this end, it first applies a Gaussian kernel so that the likelihood is high in the voxel where the peak is reported and similar but slightly lower in the close voxels. Afterward, it calculates the probability of the union of the likelihoods estimated from the different peaks and studies. ES-SDM algorithms (middle and right approach) are different, as this method aims to estimate the effect size rather than the peak likelihood. However, the first step also consists in applying an (un-normalized) Gaussian kernel, this time to achieve that voxels around a reported peak have an estimated effect size which is similar but slightly smaller to that of the peak. Afterward, effect-sizes recreated from the different peaks of a study are combined using a weighted average, i.e., when a voxel is close to two peaks, it has an effect size that depends on both peaks. Finally, the effect size maps as well as their variance maps are introduced in a meta-analytic random-effects general linear model.] [Figure: Recreation of clusters using isotropic kernels in previous versions of effect-size signed differential mapping (ES-SDM) and activation likelihood estimation (ALE). Note that the recreation of the effect size (or the estimation of the activation likelihood) does not depend on the strength of the spatial correlations, but only on the Euclidean distance between each voxel and the peak.] However, such perfect spheres of effect size around the peak are probably implausible. Independently on the distance, voxels in the same brain region as the peak are more likely to be part of the cluster of significant activation or difference. Conversely, voxels in other brain regions, or separated from the peak by cerebrospinal fluid, are less likely to be part of the cluster. Applying an isotropic kernel may thus underestimate the effect size of voxels in the same brain region as the peak, whereas it may overestimate the effect size of voxels from other brain regions. Some groups have recommended the use of large kernels, thus minimizing the underestimation of the effect size, though at the cost of a potential overestimation in some voxels (6, 20). Other developers have recommended the use of narrow kernels, thus minimizing the overestimation of the effect size, though at the cost of potential underestimation in some other voxels (4). The aim of this study was to develop anisotropic kernels for coordinate-based meta-analyses, which would assign different values to the different neighboring voxels based on the spatial correlation between them. This was specifically developed for ES-SDM, although it might be easily implemented in other widely used meta-analytical programs such as ALE. The paper also includes the creation of new tissue-specific templates and a validation of the new method. We hypothesized that the recreation of effect size maps using anisotropic kernels would be more accurate than using isotropic kernels. THEORY Previous versions of ES-SDM adopted the ALE Gaussian kernel with the aim that, in the recreated statistical map, the voxels close to a peak have slightly smaller effect sizes than that of the peak, and progressively further voxels have progressively smaller effect sizes. Specifically, the effect size of a voxel close to a peak depended on the effect size of the peak and on the Euclidean distance between the voxel and the peak by means of an un-normalized Gaussian function: (1)dvoxel=exp−D22⋅σ2⋅dpeak where d is the effect size, D is the distance, and σ is the standard deviation of the kernel (approximately 0.425 of its full-width at half-maximum, FWHM). The new method described here is based on the correlation between close voxels in the underlying structural image. Note that correlated voxels (e.g., individuals with much gray matter

Read the complete text