The Neurosubtyping of Autism Spectrum Disorders Based On Structural MRI Data
G Liu*, W Lu, J Qiu, Shandong First Medical University & Shandong Academy of Medical Sciences, Taian, CN
Presentations
PO-GePV-M-37 (Sunday, 7/25/2021) [Eastern Time (GMT-4)]
Purpose: Clinical and etiological heterogeneity remain major obstacles to biomarker identification in Autism spectrum disorders (ASD). Recently, neuroimaging raised new hope in our ability to identify the subtypes of ASD for further understand the biological mechanisms. The purpose of this study is to utilize a novel semi-supervised machine learning method termed HYDRA (heterogeneity through discriminative analysis) to identify neurosubtyping of ASD.
Methods: In this study, brain structural MRI data and clinical measures from 261 subjects with ASD and 314 healthy controls were selected from 6 sites of the Autism Brain Image Data Exchange database (ABIDE). A novel semi-supervised clustering method was utilized to divide individuals with ASD into several subtypes by the regional volumetric measures of grey matter, white matter, and CSF. This method can identify true disease subtypes by removing the influence of confounding variations introduced by age, sex, site, and other factors. The ASD subtypes obtained in this study were contrasted by voxel-wise volumetric, clinical measures and whole brain functional connectivity. To assess the reproducibility of subtypes, we take a series of analyses, including permutation tests, split-sample tests and leave-one-site-out cross validation.
Results: Two reproducible neuroanatomical subtypes were found. The ASD subtypes with difference voxel-wise volumetric patterns were revealed. Besides, the IQ and ADOS scores is significantly different, atypical brain functional connectivity patterns were observed in two subtypes. By using HYDRA to divide the ASD into 2 subtypes, the classification accuracies were significantly improved between ASD and healthy controls.
Conclusion: The method revealed distinct neurosubtyping of ASD and build stable brain-behavior relationships which could facilitate better understanding neuroanatomical heterogeneity of this disease and potentially be used to improve clinical decision-making and optimize treatment in the future.
Funding Support, Disclosures, and Conflict of Interest: Funding Support: Taishan Scholars Program of Shandong Province(TS201712065), Academic promotion programme of Shandong First Medical University No. 2019QL009. Disclosures and Conflict of Interest: The authors declare no competing financial interests.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email