Resumo:
Autism Spectrum Disorder (ASD) is an age- and sex-related lifelong neurodevelopmental
disorder characterized primarily by social impairments. Current ASD prevalence indicates
that 1/59 children are diagnosed inside the spectrum. The Autism Diagnostic Observation
Schedule, Second Edition (ADOS-2) classifies ASD according to the disorder severity.
ADOS-2 classifies as ’autism’ cases that manifest more severe symptoms and as ’ASD
non-autism’ cases that exhibit milder symptoms. Many papers aimed to create algorithms
to diagnose ASD through Machine Learning (ML) and functional Magnetic Resonance
Images (fMRI). Such approaches evaluate the oxygen flow in the brain to classify the
subjects as ASD or typical development. However, most of these works, do not provided
information regarding the disorder severity. This paper aims to use ML and fMRI to
classify the disorder severity, aim to find brain regions potentially related to the disorder
severity. We used fMRI data of 202 subjects and their ADOS-2 scores available at the
ABIDE consortium to determine the correct ASD sub-class for each one. Our results
corroborate the initial hypothesis of functional differences within ASD, with some brain
regions where the functional difference was enough to create classification accuracy of 74%.
This paper has limitations regarding the total number of samples. However, it shows a
promising approach to ASD diagnosis.