Spatially Constrained ICA for Classification of ASD with Multi-Site ABIDE Resting-State fMRI Data
Poster No:
320
Submission Type:
Abstract Submission
Authors:
Talha Imtiaz Baig1, Benjamin Klugah-Brown1, Jing Junlin2, Hu Peng1, Bochao Niu1, Hongzhou Wu1, Bharat Biswal3
Institutions:
1University of Electronic Science & Technology of China, Chengdu, Sichuan, 2Ludwig Maximilian University of Munich, Munich, Bavaria, 3New Jersey Institute of Technology, Newark, NJ
First Author:
Co-Author(s):
Introduction:
Autism Spectrum Disorder (ASD) is a persistent neurodevelopmental condition characterized by social interaction difficulties and repetitive behaviors (Rafiee et al. 2022). Neuroimaging techniques offer promising avenues for identifying biomarkers associated with ASD by analyzing functional connectivity; however, clinical diagnosis continues to depend predominantly on symptom observation (Huang et al., 2020). fMRI studies have illuminated the neurological underpinnings of ASD, revealing altered functional connectivity in key regions such as the cerebellum, amygdala, and prefrontal cortex. This study employed Semi-blind ICA, specifically the Spatial Constraint Reference ICA method, to analyze resting-state fMRI (rs-fMRI) data. This approach is particularly advantageous for large datasets, as it identifies spatially independent components with consistent activity patterns across subjects. Machine learning has become a powerful tool for predicting and classifying brain disorders using rs-fMRI data (Meijie et al. 2021). Subsequently, SVM classification was conducted to identify the brain networks most strongly associated with ASD. Results were analyzed to assess network connectivity coherence across individuals and grouped subjects, providing insights into ASD-related connectivity patterns and their implications for future diagnostic frameworks.
Methods:
We utilized publicly available resting-state MRI datasets including 38 sites with 2264 subjects. We implemented strategic data cleaning and preprocessing steps to enhance the quality of the dataset. This resulted in a final dataset comprising 996 participants from 11 sites, including 451 individuals with ASD and 545 healthy controls. Following the previous steps, a group average mask was generated across all the sites to equalize the number of features. This research employed semi-blind Constrained ICA (Spatial) algorithm, which provides 28 predefined independent components selected from the spatial reference files. Three prominent networks (DMN, SMN & VSN) were selected for analysis. For the components pairing, we combined identical components belonging to the same network from all the sites. ComBat algorithm was utilized to reduce site-specific variations in the dataset. ComBat is an empirical Bayesian method originally developed for removing batch effects in MRI studies (Acquitter et al. 2022). Following the harmonization stage, we combined two models: Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) to select optimal features. In this research, we used an SVM classifier to evaluate the classification (Oskoei, 2008). To validate the efficiency of the proposed classification approach, several performance indices were calculated, namely accuracy, area under the curve, sensitivity, and specificity (Qadri et al., 2022).
Results:
In the results, individual-site analysis shows remarkable accuracy for ASD classification, achieving up to 80.40% using spatial features (Figure 1). Multi-site analysis showed accuracies of 83.23%, 81.43%, and 80.52% for VSN, DMN, and SMN Networks, by combining spatial features of different sites. The VSN network achieved the highest combined site accuracy, with a specificity, sensitivity and AUC of 84.74%, 81.42%, and 87.90%. This demonstrates the optimism of our network analysis approach in improving ASD recognition. Results are visualized in (Figure 2).
·Classification Results of Individual Site Components Across All Networks
·Multi-site Classification Results of Combined Components Across All Networks
Conclusions:
A network analysis approach identified the three most significant brain networks: DMN, SMN, and VSN. For combined sites, our technique produced excellent results for the VSN network. In our study, we addressed inter-site heterogeneity using a harmonization technique and reduced feature variations with PCA+LDA, while other approaches neglect heterogeneity in multi-site data. These outcomes support the importance of standardizing rs-fMRI data in multi-site research. Results indicates that our network analysis method accurately classify ASD from HCs, regardless of the high number of datasets.
Disorders of the Nervous System:
Neurodevelopmental/ Early Life (eg. ADHD, autism) 1
Modeling and Analysis Methods:
Classification and Predictive Modeling 2
Keywords:
Autism
Data analysis
DISORDERS
FUNCTIONAL MRI
Machine Learning
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.
2. Huang. (2020). Identifying autism spectrum disorder from resting-state fMRI using deep belief network. IEEE Transactions on neural networks and learning systems, 32(7), 2847-2861.
3. Meijie. (2021). Autism spectrum disorder studies using fMRI data and machine learning: a review. Frontiers in Neuroscience, 15, 697870.
4. Acquitter, (2022). Radiomics-based detection of radionecrosis using harmonized multiparametric MRI. Cancers, 14(2), 286.
5. Oskoei, (2008). Support vector machine-based classification scheme for myoelectric control applied to upper limb. IEEE Transactions on Biomedical Engineering, 55(8), 1956-1965.
6. Qadri. (2022). SVseg: Stacked sparse autoencoder-based patch classification modeling for vertebrae segmentation. Mathematics, 10(5), 796.