Article summary with The neurodevelopment of autism from infancy through toddlerhood by Girault & Piven - 2020

What is autism spectrum disorder?

Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by heterogeneous symptom profiles associated with varying levels of severity in social communication deficits and restricted and repetitive behaviors. The diagnostic symptoms emerge during the end of the first year of life. Differences in other developmental domains are detectable in the first year of life, including motor skills, response to name, visual reception, attention to faces and social scenes, and visual orienting. In the beginning of the second year there are also differences in language skills and disengagement of visual attention.

How is autism spectrum disorder related to brain development?

The above mentioned behaviors develop during a highly dynamic period of postnatal brain growth that is marked by cortical expansion, functional organization of neural circuitry, and fiber myelination and maturation. Atypical brain phenotypes emerge during infancy, with altered developmental trajectories that precede the consolidation of symptoms that begins in the second year of life. Presymptomatic magnetic resonance imaging (fMRI) in infants may be used to predict diagnostic outcomes in toddlerhood.

What is characteristic about brain growth in infants and children with ASD?

ASD is characterized by brain overgrowth. Brain overgrowth is not present at birth, but emerges at the end of the first year of life, and is present by two years of age in children with ASD. The rate of change in total brain volume during the second year of life is linked to the severity of ASD-related social deficits. Faster rates of cortical surface area growth from six to twelve months of age precedes brain overgrowth in the second year of life in infants who later developed ASD, supporting the hypothesis that cortical hyper-expansion drives brain overgrowth in ASD.

What have MRI studies revealed about cortical surface area, cortical thickness, and gyrification in children with ASD?

The results support the pathological hyper-expansion of cortical surface area in ASD:

• Increases in the surface area of the frontal, temporal, and parietal lobes in two year olds with ASD.
• Accelerated rates of total cortical surface area expansion.
• Regionalized expansion in areas in the occipital, temporal, and frontal lobes in infants who later developed ASD, with robust rates of expansion in the visual cortex.
• Results with regards to cortical thickness are mixed, but it seems that aberrant patterns of cortical thickness in ASD emerge sometime after age three and then follow a dynamic developmental pattern.
• Increased gyrification in older children and adults with ASD.

How does the development of the amygdala relate to the development of ASD?

There have been only few studies that investigated the development of the amygdala in relation to the development of ASD and the results vary slightly between them. They mostly indicate that an increased amygdala size is correlated with the severity of social and communication deficits. One research reported that amygdala enlargement was present and stable across the preschool period, but, contrasting previous research, that increased amygdala volume conferred better joint attention among children with ASD.

What other noteworthy results have been found by MRI studies investigating the brain and ASD development?

• Cerebellar structural abnormalities are frequently reported in older children and adults with ADS, but the direction of the effect varies. More research is necessary, and studies should carefully control for overall brain size to ensure that findings of volumetric enlargement are specific to the cerebellum.
• The development of the corpus callosum reflects a dynamic process. The size of the corpus callosum in individuals with ASD is increased in the first year of life, normalizes by age two, and becomes smaller in the third year of life.
• Extra-axial fluid is the cerebrospinal fluid that occupies the subarachnoid space surrounding the cortical surface of the brain. It is a robust brain biomarker of ASD in early life, as increased volumes of extra-axial fluid are present in the first year of life in infants who go on to develop ASD.
• Resting-state connectivity is an ASD biomarker. Two atypical circuits were found in young children with ASD: brain regions involved in social cognition exhibited under-connectivity, whereas sensory-motor and visual brain regions showed over-connectivity.

What is fractional anisotropy and how does it relate to ASD?

Fractional anisotropy (FA) reflects the degree of directed water diffusion in the brain, indicative of more mature white matter properties, including myelination, axonal density, and fiber packaging. ASD is characterized by increased FA in the first year of life. Then maturation slows down, which may ultimately result in reduced FA values observed in older children and adults. 

How do brain and behavioral phenotypes associated with ASD emerge during the prodromal period before the second birthday?

Aberrant white matter development (indicated by fractional anisotropy and corpus callosum size) and increased extra-axial cerebrospinal fluid volumes are detectable by six months of age. This coincides with motor delays, aberrant attention to social stimuli, and atypical visual orienting. Surface area hyper-expansion in the first year precedes brain overgrowth in the second year. Infants who develop ASD show altered response to name, beginning at nine months and continuing through twenty-four months. That coincides with differential trajectories in attention to eyes and the emergence of ASD symptoms.

What possible neurobiological mechanisms underlie the development of ASD?

ASD likely arises from multiple pre- and postnatal pathogenic mechanisms involving neural proliferation and migration, pruning, synaptogenesis, myelination, and axonal development and connectivity. Each of these processes contribute independently and interactively to brain development.

• Neural progenitor proliferation and neurogenesis may play a role in the development of ASD. Neural progenitor cells derived from individuals with ASD show excess proliferation, with the level of proliferation relating to the degree of brain overgrowth.
• Cortical hyper-expansion from six to twelve months, especially in the visual cortex, may underlie deficits in visual orienting behaviors. These in turn may alter experience-dependent neuronal development and result in inefficiently pruned circuits, brain overgrowth, and the emergence of ASD traits.
• Cerebrospinal fluid contains growth factors with age-dependent effects on neuronal proliferation. Increased volumes of extra-axial fluid suggest a disruption in the circulation of cerebrospinal fluid and an accumulation of brain metabolites that impact brain function.
• Alterations in corpus callosum morphology and in the development of white matter microstructure in early ASD implicates processes governing myelination, axon caliber, density, and axonal connectivity. White matter integrity and connectivity may also be altered through experience-dependent myelination (when oligodendrocytes selectively myelinate axons which receive more input from neurons).

How can MRI help predict ASD?

MRI has the potential to predict ASD diagnosis at twenty-four months using infant MRI scans that have been collected in the first year of life. Scientists have made a classification algorithm that relied primarily on measures of regional cortical surface area growth from six to twelve months of age to predict ASD diagnostic outcome at twenty-four months. Another study found that a support vector regression machine using whole-brain functional connectivity matrices could predict ASD diagnostic outcome.

How can support vector machines and deep learning in combination with MRI dataset assist in diagnosing ASD?

Support vector machines work by finding the optimal linear plane separating diagnostic groups using the original data or data mapped into a new feature space using pre-defined kernel functions where classes become linearly separable. Deep learning algorithms can automatically identify the optimal data representation in a data-driven manner, making the need for prior selection of appropriate non-linear mapping obsolete. Deep learning methods have a few advantages. Input features are learning from the data and not derived, and thus less prone to overfitting. It also allows for the detection of more subtle patterns in the data.