BrainyNet: An ImageNet-like resource for brain-reading AI
Poster No:
1826
Submission Type:
Abstract Submission
Authors:
Peng Gao1,2, Xi-Nian Zuo1,2
Institutions:
1State Key Lab of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, 100875, China, 2National Basic Science Data Center, Beijing, 100190, China
First Author:
Peng Gao
State Key Lab of Cognitive Neuroscience and Learning, Beijing Normal University|National Basic Science Data Center
Beijing, 100875, China|Beijing, 100190, China
State Key Lab of Cognitive Neuroscience and Learning, Beijing Normal University|National Basic Science Data Center
Beijing, 100875, China|Beijing, 100190, China
Co-Author:
Xi-Nian Zuo
State Key Lab of Cognitive Neuroscience and Learning, Beijing Normal University|National Basic Science Data Center
Beijing, 100875, China|Beijing, 100190, China
State Key Lab of Cognitive Neuroscience and Learning, Beijing Normal University|National Basic Science Data Center
Beijing, 100875, China|Beijing, 100190, China
Introduction:
The ImageNet database has driven advancements in deep learning and computer vision, but no equivalent resource exists for mapping human brain function. While fMRI has significantly advanced our understanding of brain activity, a large-scale, publicly available dataset capturing the complexity of brain responses to real-world stimuli is still lacking. This gap impedes progress in brain decoding and constrains AI's ability to replicate human cognition and perception.
To address this, we introduce BrainyNet, a multimodal dataset containing fMRI and structural MRI data from 50 healthy adult participants across two different sites. The dataset includes brain scans taken while participants viewed 4,441 unique images from seven major categories, offering a resource for both deep learning and neuroscience research. The data underwent standardized preprocessing and is publicly available, supporting test-retest studies and reproducibility efforts.
To address this, we introduce BrainyNet, a multimodal dataset containing fMRI and structural MRI data from 50 healthy adult participants across two different sites. The dataset includes brain scans taken while participants viewed 4,441 unique images from seven major categories, offering a resource for both deep learning and neuroscience research. The data underwent standardized preprocessing and is publicly available, supporting test-retest studies and reproducibility efforts.
Methods:
BrainyNet consists of structural (T1-weighted) and functional MRI (fMRI) data from participants who completed both retinal localization and image-viewing tasks. Data was collected on 3T MRI scanners at two sites: the Institute of Biophysics (CAS) and the Institute of Psychology (CAS). Specifically, 30 participants at the Biophysics Institute and 20 at the Psychology Institute completed the retinal localization task followed by seven or five image-viewing rounds. Each round involved 10 image categories with seven images per category, totaling 70 images per round. The image categories include scenes, buildings, faces, man-made objects, animals, plants, and food.
Imaging parameters for structural scans (T1w, T2w) and functional scans (EPI BOLD) were optimized for each site. The study employed BIDS standards for data organization, and data quality was assessed using the MRIQC toolkit. Data preprocessing utilized the Connectome Computation System (CCS), including denoising, motion correction, normalization, and brain extraction. Surface-based functional brain mapping was performed, and data were processed in CIFTI format for cortical surface analysis.
Imaging parameters for structural scans (T1w, T2w) and functional scans (EPI BOLD) were optimized for each site. The study employed BIDS standards for data organization, and data quality was assessed using the MRIQC toolkit. Data preprocessing utilized the Connectome Computation System (CCS), including denoising, motion correction, normalization, and brain extraction. Surface-based functional brain mapping was performed, and data were processed in CIFTI format for cortical surface analysis.
Results:
BrainyNet includes data from 520 fMRI sessions across 50 participants. The dataset was quality-checked using MRIQC, and the preprocessed data are available in BIDS format, allowing easy integration with other neuroscience tools. Preprocessing steps included motion correction, ICA-AROMA for noise removal, intensity normalization, and brain extraction. The cortical surface models were created and resampled to 32,000 vertices, with functional data aligned across participants. The processed data are publicly shared via the Science Data Bank.
Conclusions:
BrainyNet represents a critical resource for advancing research in both neuroscience and artificial intelligence. The dataset supports studies on the reproducibility of brain imaging data and the development of deep learning models to decode brain activity. By providing high-quality fMRI data in response to natural stimuli, BrainyNet can facilitate research on brain-computer interfaces (BCIs), cognitive neuroscience, and AI-driven brain modeling.
This dataset is designed to bridge the gap between fMRI-based neuroscience and AI-driven brain decoding, offering a platform for developing more accurate models of human cognition. By making these data publicly available, we aim to foster interdisciplinary collaboration and accelerate advances in both fields, enhancing our understanding of brain function and the potential for AI models to replicate it.
BrainyNet is an essential step in integrating deep learning with neuroscience, enabling more accurate models of brain activity and advancing our knowledge of human cognition, brain function, and brain-machine interfaces. The availability of large-scale datasets like BrainyNet will be crucial in overcoming current limitations and accelerating progress in both neuroscience and AI.
This dataset is designed to bridge the gap between fMRI-based neuroscience and AI-driven brain decoding, offering a platform for developing more accurate models of human cognition. By making these data publicly available, we aim to foster interdisciplinary collaboration and accelerate advances in both fields, enhancing our understanding of brain function and the potential for AI models to replicate it.
BrainyNet is an essential step in integrating deep learning with neuroscience, enabling more accurate models of brain activity and advancing our knowledge of human cognition, brain function, and brain-machine interfaces. The availability of large-scale datasets like BrainyNet will be crucial in overcoming current limitations and accelerating progress in both neuroscience and AI.
Neuroinformatics and Data Sharing:
Databasing and Data Sharing 1
Novel Imaging Acquisition Methods:
BOLD fMRI 2
Keywords:
ADULTS
Design and Analysis
FUNCTIONAL MRI
1|2Indicates the priority used for review
·Figure.1 Overview of BrainyNet: (a) Experimental Design of BrainyNet Dataset; (b) The process of image task scanning
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?
Provide references using APA citation style.
[2] Xu, T., Yang, Z., Jiang, L. L., Xing, X. X. & Zuo, X. N. A Connectome Computation System for discovery science of brain. Science Bulletin 60, 86–95 (2015).
[3] Xing, X. X., Xu, T., Jiang, C., Wang, Y. S. & Zuo, X. N. Connectome Computation System: 2015–2021 updates. Science Bulletin 67, 448–451 (2022).