SEARCH

SEARCH BY CITATION

Multiparameter Classification Approach to Structural Neuroimaging Data [Ecker et al., 2010]

  1. Top of page
  2. Multiparameter Classification Approach to Structural Neuroimaging Data [Ecker et al., ]
  3. Heterogeneity of 16p11.2 Microdeletion Clinical Presentation [Hanson et al., ]

The authors attempted to address the heterogeneity of previous ASD neuroanatomical findings by using an analysis that would allow the study of multiple parameters of cortical brain structure. The authors began by recruiting 20 adult control male subjects and 20 adult male subjects with ASD. All subjects were right-handed. There was an attempt to match for IQ but the mean full-scale, verbal, and performance IQs were 7, 4, and 12 points lower for the ASD group compared to the control group. The performance IQ was the only difference that was statistically significant. Although the mean ages did not significantly differ, the range of ages of ASD subjects was 20–68 years compared to 20–49 years for the control group. Diagnosis was made on a clinical basis (ICD-10 criteria) informed by ADI-R administration for 17 subjects and informed by ADOS administration for the three subjects whose parents were unavailable or refused to participate. Participants with ASD met ADI-R or ADOS criteria in all domains or missed by one point in no more than one domain. Sixteen of the 20 subjects with ASD did not have a delay in phrase speech (i.e. developed phrase speech by 36 months) and four of the subjects with ASD developed phrase speech after 36 months of age. Due to the small sample size of those with phrase speech delay, the ASD group was analyzed as a combined group, independent of history of phrase speech delay.

Images were acquired using a 1.5-Tesla system in the coronal plane. Details about acquisition are available in the article, as are details about image analysis and further data analysis. The novel aspect of the article was the application of a support vector machine (SVM) approach to five parameters of brain structure. The five parameters were well defined and illustrated in the article. There were two volumetric parameters, cortical thickness and pial area, and three geometric parameters, radial curvature, average convexity, and metric distortion.

The SVM approach developed a classifier that used a two-out procedure in which 20 classifications were determined, each dropping out a subject with ASD from the classification and then applying the classifier to that subject. This allowed the determination of whether the parameters derived from the other subjects could classify each subject. This was important to avoid the circularity that would have resulted from using subject A's data to determine subject A's classification. This allowed the determination of sensitivity and specificity of the five-parameter model and of each parameter by itself.

Interestingly, the classification using the left hemisphere parameters performed at a higher level than the right hemisphere parameters. However, the performance of left hemisphere compared to right hemisphere parameters were not statistically compared. For all left hemisphere parameters together, 85% of subjects were correctly classified with a sensitivity of 90% and specificity of 80%. In contrast, with all right hemisphere parameters using for classification, 65% were correctly classified with a sensitivity of 60% and specificity of 70%. Although not significantly higher than all parameters, the left hemisphere cortical thickness parameter had the highest percentage of correct classifications (90%) with a sensitivity of 90% and a specificity of 90%.

The authors acknowledged the limitation of the classification system, which was developed for a restricted set of subjects with ASD. In addition, they have not replicated the classifier in another set of subjects. However, they are correct that this article provides a first step in application of a multi-parametric approach to the study of morphometric differences in ASD. This is consistent with the continuing spotlighting of heterogeneity of ASD across multiple domains of investigation of ASD.

Heterogeneity of 16p11.2 Microdeletion Clinical Presentation [Hanson et al., 2010]

  1. Top of page
  2. Multiparameter Classification Approach to Structural Neuroimaging Data [Ecker et al., ]
  3. Heterogeneity of 16p11.2 Microdeletion Clinical Presentation [Hanson et al., ]

Previously, this series of reviews of the current autism literature highlighted a series of articles that used developments in microarray genetic technology to identify what may be one of the more common recurrent copy number variations in ASD, deletions of the 16p11.2 region. In review of this and other articles (e.g. the chromosome 1q21.2 microdeletion [Cook, 2008]), it was pointed out that although an appealing possibility to some, the replacement of clinical diagnosis of ASD with specific genetic diagnosis was insufficient because of the variation in clinical presentation of many of the genetic syndromes. Although one often finds a range of presentations within subjects with ASD with a given specific genetic finding such as 16p11.2 microdeletion, the study of collections of subjects with ASD will lead to the appearance of less heterogeneous presentation than is present with a less biased ascertainment. Conversely, settings with less expertise in ASD diagnosis will often have less clinical index of suspicion for ASD and make diagnoses such as “speech delay” when ASD is present.

With this background, the article by Hanson et al. reported on subjects who were found to have 16p11.2 microdeletion in the group of children studied with chromosomal microarray at Boston Children's Hospital. In other words, ascertainment was not restricted to an autism clinic or autism research study. Twenty-one children were found to have 16p11.2 microdeletion based on medical record review. Of these 21 children, medical records documented an ASD diagnosis in 7 of the 21 subjects and suspected ASD diagnosis in an eighth child.

Detailed assessment, including administration of the ADI-R and ADOS, was conducted for 11 of these 21 children after attempted recruitment of all the individuals with 16p11.2 microdeletion. Of these 11 children, cognitive testing revealed a range of verbal reasoning from less than 30 to 111 and a range of nonverbal reasoning from less than 49 to 131. Four of these 11 children had a clinical diagnosis of ASD based on medical record review. ADOS revealed that six of the 11 children had an autism classification. Of the two children with no clinical diagnosis of an ASD who had ADOS classification of autism, each ADI-R domain (reciprocal social, communication, and restricted and repetitive behaviors) did not have scores exceeding cutoffs for autism classification. Consistent with the parent report by ADI-R, the parent-reported SRS scores for these two children were not elevated. In examination of the table detailing the presentation and testing of these two individuals, estimates of cognitive functioning ranged from 71 to 131 and there was no elevation in maladaptive behaviors. One of the subjects was a 17-year-old adolescent with articulation difficulties and the subject with nonverbal reasoning of 131 was 3 years, 5 months old. Given the typically higher specificity of the ADOS, it is unclear why there was a discrepancy in ADOS and ADI-R in these two subjects.

Five of 11 subjects met the full criteria of autism classification on the ADI-R. Eight of 11 subjects met the full criteria for autism classification on the ADI-R or ADOS, leading the authors to comment that independent of diagnosis of ASD, assessment for ASD with structured instruments is likely to contribute to improved assessment of the range of difficulties of these patients to allow optimal development of individual intervention strategies.

This article contributes to the growing literature that demonstrates that although 16p11.2 microdeletion is found relatively commonly in ASD (0.5–1%), is associated with relatively intact cognitive functioning, and has been found to be significantly increased in ASD populations compared to control populations, the presence of ASD varies across individuals with 16p11.2 microdeletion. Until population-based samples are obtained, the predictive value of a 16p11.2 microdeletion for the development of ASD will not be able to be accurately estimated, but this article contributes to the literature demonstrating the variable expression of developmental impairment related to 16p11.2 microdeletion.

  • Cook, E.H. (2008). Adult structural MRI, Postmortem GABA-A protein, 1q21.1 Microdeletion/microduplication. Autism Research, 1, 308310.
  • Ecker, C., Marquand, A., Mourao-Miranda, J., Johnston, P., Daly, E.M., et al. (2010). Describing the brain in autism in five dimensions—magnetic resonance imaging-assisted diagnosis of autism spectrum disorder using a multiparameter classification approach. Journal of Neuroscience, 30, 1061210623.
  • Hanson, E., Nasir, R.H., Fong, A., Lian, A., Hundley, R., et al. (2010). Cognitive and behavioral characterization of 16p11.2 deletion syndrome. Journal of Developmental and Behavioral Pediatrics.