Autism: Project Description
Resting state functional MRI (rs-fMRI) measures spontaneous fluctuations in blood oxygen level dependent (BOLD) signal, thought to reflect fluctuation in underlying neuronal activity. Whole-brain rs-fMRI of normal people has characterized many long-range and short-range neural networks that demonstrate reproducible temporal synchrony of resting state BOLD signal. Applications of this rs-fMRI technique to several types of cognitive pathology (including autism, schizophrenia, bipolar disorder, and depression) have demonstrated consistent perturbations of this temporal synchrony related to the underlying pathology.
Regarding the autism of disease, my research group has discovered several features of rs-fMRI temporal synchrony that are perturbed at the population level, including long-range interhemispheric connectivity and short-range regional homogeneity. These findings certainly advance our understanding of the pathology underlying the autism spectrum of disease, but the clinical utility of this information is currently limited. A more clinically relevant issue is whether the features in a single rs-fMRI data set can be used to determine the population from which the data originated. That is, can you make the diagnosis of autism based on features in the rs-fMRI data? Recent application of machine learning classification to fMRI data sets suggests this to be a realistic possibility.
The current project proposes to develop a classifier that discriminates autism patients from typically developing controls on the basis of their rs-fMRI data. A support vector machine classifier will be developed from an existing large rs-fMRI data set obtained from a well-characterized population of autism patients and typically developing control patients. The classifier will be internally assessed for metrics of clinical validity, such as sensitivity, specificity, and accuracy. Then the classifier will be externally validated with rs-fMRI data obtained from a separate population of autism patients, at a different institution.