A Generative Model Framework for Task-based fMRI: Integrating Activation Patterns and Brain Network
Presented During:
Friday, June 27, 2025: 11:30 AM - 12:45 PM
Brisbane Convention & Exhibition Centre
Room:
P2 (Plaza Level)
Poster No:
1038
Submission Type:
Abstract Submission
Authors:
Rongquan Zhai1, Tianye Jia1
Institutions:
1Fudan University, Shanghai, Shanghai
First Author:
Co-Author:
Introduction:
Research Background:
Functional Magnetic Resonance Imaging (fMRI) is a pivotal technique in understanding brain responses associated with cognitive functions or during resting states. Traditional studies often assume that a set of homogeneous cognitive functions is attached to specific task paradigms, expecting similar responses across all participants. However, this fundamental hypothesis remains largely untested due to the lack of personalized cognitive processing data, such as personalized brain activation templates.
Brain Measurements in fMRI:
Brain activations and functional connectivity (FC) are the two primary measurements in fMRI studies. While brain activation reflects cognition-specific responses, functional connectivity represents the general network of brain interactions. Previous studies have rarely analyzed these two measurements together.
Motivation and Objective:
Inspired by the similarity between the diffusion model's generation process and the signal transmission and generation in the brain, this study aims to address whether combining brain activation and functional connectivity information can enhance our understanding of brain responses to different task conditions.
Functional Magnetic Resonance Imaging (fMRI) is a pivotal technique in understanding brain responses associated with cognitive functions or during resting states. Traditional studies often assume that a set of homogeneous cognitive functions is attached to specific task paradigms, expecting similar responses across all participants. However, this fundamental hypothesis remains largely untested due to the lack of personalized cognitive processing data, such as personalized brain activation templates.
Brain Measurements in fMRI:
Brain activations and functional connectivity (FC) are the two primary measurements in fMRI studies. While brain activation reflects cognition-specific responses, functional connectivity represents the general network of brain interactions. Previous studies have rarely analyzed these two measurements together.
Motivation and Objective:
Inspired by the similarity between the diffusion model's generation process and the signal transmission and generation in the brain, this study aims to address whether combining brain activation and functional connectivity information can enhance our understanding of brain responses to different task conditions.
Methods:
Experimental Paradigms and Data:
The study utilized fMRI data from the IMAGEN cohort, which includes extensive neuropsychological, behavioral, clinical, and environmental assessments, along with T1-weighted structural MRI, task-based fMRI, and genetic data. Specifically, the Emotional Face Task (EFT) and Stop Signal Task (SST) were used.
Preprocessing Methods:
Functional and structural data underwent rigorous preprocessing using Statistical Parametric Mapping software (SPM8) and other tools. This included slice timing correction, realignment, normalization, and spatial smoothing. Resting-state functional connectivity was calculated using both traditional seed-based methods and novel patch-based methods.
Diffusion Generative Model:
The study employed a U-ViT model based on the Transformer block, modified for 3D denoising generative architecture. The model starts from Gaussian noise and gradually denoises it to generate brain activations that closely match the target distribution.
Cognitive Relevance Map Construction:
Cognitive relevance maps were constructed by sequentially removing brain regions and measuring the decrease in prediction similarity. This process yields a map of cognitive relevance across the entire brain, indicating which regions are critical for the prediction tasks.
The study utilized fMRI data from the IMAGEN cohort, which includes extensive neuropsychological, behavioral, clinical, and environmental assessments, along with T1-weighted structural MRI, task-based fMRI, and genetic data. Specifically, the Emotional Face Task (EFT) and Stop Signal Task (SST) were used.
Preprocessing Methods:
Functional and structural data underwent rigorous preprocessing using Statistical Parametric Mapping software (SPM8) and other tools. This included slice timing correction, realignment, normalization, and spatial smoothing. Resting-state functional connectivity was calculated using both traditional seed-based methods and novel patch-based methods.
Diffusion Generative Model:
The study employed a U-ViT model based on the Transformer block, modified for 3D denoising generative architecture. The model starts from Gaussian noise and gradually denoises it to generate brain activations that closely match the target distribution.
Cognitive Relevance Map Construction:
Cognitive relevance maps were constructed by sequentially removing brain regions and measuring the decrease in prediction similarity. This process yields a map of cognitive relevance across the entire brain, indicating which regions are critical for the prediction tasks.
·fMRI diffusion model combining network and activation
Results:
Generative Model for EFT and SST:
This study used a transformer-based diffusion model to predict brain activations in specific task conditions. The model could handle various imaging data, such as task-specific activations, resting-state functional networks, structural connectivity, and gray matter density.
Prediction Performance:
The model captured individual brain data variability and performed best with both brain activation and resting-state functional connectivity. Additional connectivity or gray matter data helped in transitioning between Emotional Face Task (EFT) and Stop Signal Task (SST) states.
Cross-task & Individual Analysis:
Cross-task predictions using structural or functional information failed, highlighting task differences. Cognitive relevance analysis showed that specific cognitive maps ("emotional" or "inhibition") were better than group averages in predictions. Individualized maps revealed unique features in key brain regions like the left inferior frontal cortex, anterior cingulate cortex, and orbitofrontal cortex.
This study used a transformer-based diffusion model to predict brain activations in specific task conditions. The model could handle various imaging data, such as task-specific activations, resting-state functional networks, structural connectivity, and gray matter density.
Prediction Performance:
The model captured individual brain data variability and performed best with both brain activation and resting-state functional connectivity. Additional connectivity or gray matter data helped in transitioning between Emotional Face Task (EFT) and Stop Signal Task (SST) states.
Cross-task & Individual Analysis:
Cross-task predictions using structural or functional information failed, highlighting task differences. Cognitive relevance analysis showed that specific cognitive maps ("emotional" or "inhibition") were better than group averages in predictions. Individualized maps revealed unique features in key brain regions like the left inferior frontal cortex, anterior cingulate cortex, and orbitofrontal cortex.
Conclusions:
Combining brain activation and functional connectivity yields a model predicting individual brain activations and cognitive maps. This reveals task engagement variations, hinting at personalized cognitive activations. The brain's task-switching ability ensures efficient processing. These findings aid personalized medicine, training, and brain function understanding.
Brain Stimulation:
Non-Invasive Stimulation Methods Other
Disorders of the Nervous System:
Psychiatric (eg. Depression, Anxiety, Schizophrenia) 2
Emotion, Motivation and Social Neuroscience:
Emotional Learning
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 1
Connectivity (eg. functional, effective, structural)
Keywords:
Cognition
Computational Neuroscience
Computing
Data analysis
Emotions
fMRI CONTRAST MECHANISMS
Motor
Multivariate
Psychiatric Disorders
Other - BOLD fMRI
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.
Bao, F., Li, C., Cao, Y., & Zhu, J. (2022). All are worth words: a vit backbone for score-based diffusion models. NeurIPS 2022 Workshop on Score-Based Methods,
Bari, A., & Robbins, T. W. (2013). Inhibition and impulsivity: behavioral and neural basis of response control. Progress in neurobiology, 108, 44-79.
Grosbras, M.-H., & Paus, T. (2006). Brain networks involved in viewing angry hands or faces. Cerebral cortex, 16(8), 1087-1096. https://doi.org/10.1093/CERCOR/BHJ050
Huang, C.-C., Rolls, E. T., Feng, J., & Lin, C.-P. (2022). An extended Human Connectome Project multimodal parcellation atlas of the human cortex and subcortical areas. Brain Structure and Function, 227(3), 763-778.
Huang, C.-C., Rolls, E. T., Hsu, C.-C. H., Feng, J., & Lin, C.-P. (2021). Extensive cortical connectivity of the human hippocampal memory system: beyond the “what” and “where” dual stream model. Cerebral cortex, 31(10), 4652-4669.
Mascarell Maričić, L., Walter, H., Rosenthal, A., Ripke, S., Quinlan, E. B., Banaschewski, T., Barker, G. J., Bokde, A. L., Bromberg, U., & Büchel, C. (2020). The IMAGEN study: a decade of imaging genetics in adolescents. Molecular psychiatry, 25(11), 2648-2671.
Schumann, G., Loth, E., Banaschewski, T., Barbot, A., Barker, G., Büchel, C., Conrod, P. J., Dalley, J., Flor, H., & Gallinat, J. (2010). The IMAGEN study: reinforcement-related behaviour in normal brain function and psychopathology. Molecular psychiatry, 15(12), 1128-1139.
Song, J., Meng, C., & Ermon, S. Denoising Diffusion Implicit Models. International Conference on Learning Representations,
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems, 30.
Yarkoni, T., Poldrack, R. A., Nichols, T. E., Van Essen, D. C., & Wager, T. D. (2011). Large-scale automated synthesis of human functional neuroimaging data. Nature Methods, 8(8), 665-670.