On the brain structure heterogeneity of autism: Parsing out acquisition site effects with significance-weighted principal component analysis

Abstract

Neuroimaging studies have reported structural and physiological differences that could help understand the causes and development of Autism Spectrum Disorder (ASD). Many of them rely on multisite designs, with the recruitment of larger samples increasing statistical power. However, recent large-scale studies have put some findings into question, considering the results to be strongly dependent on the database used, and demonstrating the substantial heterogeneity within this clinically defined category. One major source of variance may be the acquisition of the data in multiple centres. In this work we analysed the differences found in the multisite, multi-modal neuroimaging database from the UK Medical Research Council Autism Imaging Multicentre Study (MRC AIMS) in terms of both diagnosis and acquisition sites. Since the dissimilarities between sites were higher than between diagnostic groups, we developed a technique called Significance Weighted Principal Component Analysis (SWPCA) to reduce the undesired intensity variance due to acquisition site and to increase the statistical power in detecting group differences. After eliminating site-related variance, statistically significant group differences were found, including Broca's area and the temporo-parietal junction. However, discriminative power was not sufficient to classify diagnostic groups, yielding accuracies results close to random. Our work supports recent claims that ASD is a highly heterogeneous condition that is difficult to globally characterize by neuroimaging, and therefore different (and more homogenous) subgroups should be defined to obtain a deeper understanding of ASD. Hum Brain Mapp 38:1208-1223, 2017. © 2016 Wiley Periodicals, Inc.

Keywords: autism spectrum disorder; structural heterogeneity; structural magnetic resonance imaging; voxel based morphometry.

Figure 1

Summary of the SWPCA algorithm, along with its context in the pipeline used in this article. Circles represent the input data, both images (green shading) and class (group and acquisition site, purple shading). Rectangles represent the different procedures applied, comprising the DARTEL normalization and registration, the different steps contained in SWPCA ‐PCA, ANOVA and obtaining the weighting function Λ(c)‐ and the suseqent analysis.

Figure 2

Box‐plot of the distribution of the component scores at each site in the four first components. We assume that bigger differences between distributions imply a bigger influence of the acquisition site on the portion of variance modelled by that component and therefore, to parse out those differences, the resulting weight will be smaller. [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 3

Brain t‐map (voxel‐based morphometry) of significant (P < 0.01, |t|>2.57) GM and WM between‐group differences using qT1, qT2, synT1, GM and WM modalities after applying SWPCA to remove site effects. [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 4

Brain t‐map (voxel‐based morphometry) of significant (P < 0.01, |t|>2.57) grey and white matter differences in ASD using qT1, qT2, synT1, GM and WM images before and after applying SWPCA to remove site effects. [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 5

Brain Z‐map (component‐based morphometry) of significant (P < 0.01, |Z|>2.57) grey and white matter differences in ASD using qT1, qT2, synT1, GM and WM images before and after applying SWPCA to remove site effects. [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 6

Location of the significant region that we have labelled D (posterior part of the superior temporal gyrus) within the MNI template.

Figure 7

The template used in this work compared to two of the participants with abnormal ventricle size (21016 and 21018). Atrophy of the cerebellum in participant 21016 can also be appreciated, responsible for some of the ‘highlighted’ areas in qT1, qT2 and synT1 t‐maps (see Fig. 4).