Machine-Learning-Identified Brain Regions for Diagnosing Autism and Their Clinical Correlations
Poster No:
316
Submission Type:
Abstract Submission
Authors:
Yen-Chin Wang1, Chung-Yuan Cheng2, Chi-Chun Lee3, Susan Shur-Fen Gau1
Institutions:
1National Taiwan University Hospital and College of Medicine, Department of Psychiatry, Taipei, Taiwan, 2National Taiwan University College of Medicine, Taipei, Taiwan, 3Department of Electrical Engineering, National Tsing Hua University, Hsinchu, Taiwan
First Author:
Yen-Chin Wang
National Taiwan University Hospital and College of Medicine, Department of Psychiatry
Taipei, Taiwan
National Taiwan University Hospital and College of Medicine, Department of Psychiatry
Taipei, Taiwan
Co-Author(s):
Susan Shur-Fen Gau
National Taiwan University Hospital and College of Medicine, Department of Psychiatry
Taipei, Taiwan
National Taiwan University Hospital and College of Medicine, Department of Psychiatry
Taipei, Taiwan
Introduction:
Accumulating evidence revealed different brain development among autistic population compared with their typical developing comparisons (Bedford et al., 2024; Pretzsch et al., 2024; Shen et al., 2024). Brain imaging data can be utilized to establish machine-learning models for classifying autism (Hyde et al., 2019; Song et al., 2021). However, the specific brain regions that most significantly contribute to the models' explainability and their clinical relevance remain underexplored.
This study employed explainable artificial intelligence methods to identify important brain regions associated with a machine learning-based diagnostic model for autism using cortical thickness data. Additionally, it aimed to explore the correlations between these brain regions and clinical symptoms.
This study employed explainable artificial intelligence methods to identify important brain regions associated with a machine learning-based diagnostic model for autism using cortical thickness data. Additionally, it aimed to explore the correlations between these brain regions and clinical symptoms.
Methods:
Structural MRI data were collected from 162 autistic participants (94.4% male, mean age ± standard deviation, SD: 13.4 ± 2.6 years) and 227 typically developing individuals (70% male, mean age ± SD: 12.2 ± 2.9 years). Cortical thickness data were processed using Freesurfer and parcellated with the Destrieux atlas, yielding 150 features from both hemispheres. These features were used to train features an XGBoost classifier. Feature importance was assessed using SHapley Additive exPlanations (SHAP) values to identify the most important brain regions contributing to the model's explainability. Correlations between the cortical thickness in these most important brain regions and clinical measures-including the Social Responsiveness Scale (SRS) and Autism Spectrum Quotient (AQ)-were analyzed within the autistic group, adjusting for age, sex, and full-scale IQ.
Results:
The XGBoost model achieved high accuracy, yielding an area under the curve (AUC) of 0.883, sensitivity of 0.822, and specificity of 0.781. The highest average SHAP values and the best discrimination between autistic participants and controls identified the top four brain regions. These included the right superior segment of the circular sulcus of the insula, right short insular gyri, right posterior ramus of the lateral sulcus, and right intraparietal sulcus and transverse parietal sulci. Subsequent brain-behavior correlation analyses revealed consistently negative correlations with various autistic features. For example, the AQ subscales for patterns and attention to detail showed significant negative correlations with the cortical thickness of the right short insular gyri. In contrast, the SRS subscale for social awareness revealed a limited correlation with these brain regions. The fourth important region, the right intraparietal sulcus and transverse parietal sulci, exhibited weaker correlations with autistic features than the other regions.
Conclusions:
This study provides a machine-learning-based explanatory model for identifying key brain regions relevant to autism diagnosis and their clinical correlates. The findings enhance our understanding of brain-behavior relationships in autism and offer potential insights for improving diagnostic models and exploring therapeutic strategies to address the heterogeneity of autism.
Disorders of the Nervous System:
Neurodevelopmental/ Early Life (eg. ADHD, autism) 1
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Cortical Anatomy and Brain Mapping
Novel Imaging Acquisition Methods:
Anatomical MRI
Keywords:
Autism
Cortex
Machine Learning
MRI
STRUCTURAL MRI
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Hyde, K. K., Novack, M. N., LaHaye, N., Parlett-Pelleriti, C., Anden, R., Dixon, D. R., & Linstead, E. (2019). Applications of Supervised Machine Learning in Autism Spectrum Disorder Research: a Review. Review Journal of Autism and Developmental Disorders, 6(2), 128-146. https://doi.org/10.1007/s40489-019-00158-x
Pretzsch, C. M., Arenella, M., Lerch, J. P., Lombardo, M. V., Beckmann, C., Schaefer, T., Leyhausen, J., Gurr, C., Bletsch, A., Berg, L. M., Seelemeyer, H., Floris, D. L., Oakley, B., Loth, E., Bourgeron, T., Charman, T., Buitelaar, J., McAlonan, G., Murphy, D., . . . Group, E.-A. L. (2024). Patterns of Brain Maturation in Autism and Their Molecular Associations. JAMA Psychiatry, 81(12), 1253-1264. https://doi.org/10.1001/jamapsychiatry.2024.3194
Shen, L., Zhang, J., Fan, S., Ping, L., Yu, H., Xu, F., Cheng, Y., Xu, X., Yang, C., & Zhou, C. (2024). Cortical thickness abnormalities in autism spectrum disorder. European Child and Adolescent Psychiatry, 33(1), 65-77. https://doi.org/10.1007/s00787-022-02133-0
Song, D.-Y., Topriceanu, C.-C., Ilie-Ablachim, D. C., Kinali, M., & Bisdas, S. (2021). Machine learning with neuroimaging data to identify autism spectrum disorder: a systematic review and meta-analysis. Neuroradiology, 63, 2057-2072.