TReND: Transformer derived features and Regularized NMF for Delineation of Functional Networks
Presented During:
Saturday, June 28, 2025: 11:30 AM - 12:45 PM
Brisbane Convention & Exhibition Centre
Room:
P2 (Plaza Level)
Poster No:
1646
Submission Type:
Abstract Submission
Authors:
Sovesh Mohapatra1,2, Minhui Ouyang2,3, Lianglong Sun4, Yong He4, Hao Huang2,3
Institutions:
1Department of Bioengineering, University of Pennsylvania, Philadelphia, PA, 2Department of Radiology, Children's Hospital of Philadelphia, Philadelphia, PA, 3Department of Radiology, University of Pennsylvania, Philadelphia, PA, 4Beijing Normal University, Beijing, Beijing
First Author:
Sovesh Mohapatra
Department of Bioengineering, University of Pennsylvania|Department of Radiology, Children's Hospital of Philadelphia
Philadelphia, PA|Philadelphia, PA
Department of Bioengineering, University of Pennsylvania|Department of Radiology, Children's Hospital of Philadelphia
Philadelphia, PA|Philadelphia, PA
Co-Author(s):
Minhui Ouyang
Department of Radiology, Children's Hospital of Philadelphia|Department of Radiology, University of Pennsylvania
Philadelphia, PA|Philadelphia, PA
Department of Radiology, Children's Hospital of Philadelphia|Department of Radiology, University of Pennsylvania
Philadelphia, PA|Philadelphia, PA
Hao Huang
Department of Radiology, Children's Hospital of Philadelphia|Department of Radiology, University of Pennsylvania
Philadelphia, PA|Philadelphia, PA
Department of Radiology, Children's Hospital of Philadelphia|Department of Radiology, University of Pennsylvania
Philadelphia, PA|Philadelphia, PA
Introduction:
Neuroscientists have long attempted to subdivide the human brain into a mesh of anatomically and functionally distinct, contiguous regions (Huang, 2005, Yeo, 2011). This challenge become particularly complex in the neonatal brain, where functional organization differs markedly from that of adults (Peng, 2020). During the third trimester, the neonatal brain undergoes a critical phase of enhanced functional segregation, primarily driven by the rapid development of functional connectivity and the formation of hubs in primary regions (Cao, 2017). However, achieving accurate and reliable parcellation of specific functional networks (FNs) in newborns presents unique challenges. The combined effects of rapid functional segregation, low imaging quality, and the absence of established functional atlases complicate this process. In this study, we developed a four-stage bottom-up approach to parcellate neonatal brain FNs, providing a foundational neonatal functional brain atlas. This approach integrates a transformer-based autoencoder architecture for extracting novel features, coupled with innovative regularized NMF clustering algorithm.
Methods:
Data details: The study comprised of two datasets: simulated dataset and term infants. The simulated dataset comprised 200 simulated 2D images, each with dimensions of 100x100 voxels and 2300 timepoints, including 15 FNs to replicate real brain connectivity (Fig.1A). The neonatal dataset consists of 300 term infants (Mean scan age: 41.16GA weeks and SD: 1.7) from the dHCP cohort (Edwards, 2022). All data were preprocessed and registered to the 40 week neonatal surface. Feature extraction: We employed a self-supervised transformer-based autoencoder model architecture to capture temporal dependencies and interactions within the BOLD signals at voxel level. To effectively train the model, we introduced a composite loss function combing mean absolute error helping the model to reduce variance and mean squared error to decrease bias by penalizing larger errors (Fig.1B). Correlation and clustering: Following that, we apply Pearson correlation on the voxel-wise features to obtain the functional connectivity matrix (Fig.1C). This connectivity matrix serves as the basis for clustering. We improved the NMF initialization by implementing non-negative double singular value decomposition with averaging for greater stability. Regularization and smoothing parameters were added to ensure robust parcellation, and the NMF-derived features were clustered using KMeans to generate FN labels. (Fig.1D/E). For constructing a confidence map, we utilized the silhouette measure (Rousseeuw, 1987).
·Fig 1. Schematic representation of end-to-end functional parcellation framework.
Results:
Due to the lack of functional atlases in neonates, we validated our transformer-based autoencoder feature extraction method using the simulated dataset, maintaining a consistent clustering approach. This validation yielded high average Dice (0.92) and Intersection over Union (IoU, 0.89) scores, with a low percentage of mislabeled vertices (5.82%), outperforming traditional mathematically constrained methods (Fig.2A). Additionally, our coupled regularized NMF and KMeans clustering approach demonstrated superior stability and accuracy compared to NMF and other clustering algorithms, further validating its effectiveness (Fig.2A). The cluster stability analyses indicate that 7 and 19 FNs are suitable starting points for parcellating the neonatal cortex. The confidence maps, as illustrated in the 7 FNs map (Fig. 2B), show lower confidence in boundary regions, suggesting potential for further subdivision with a higher number of networks. Notably, we observe that the emergence of the default mode network at this age coincides with reduced confidence in corresponding regions of the map.
·Fig 2. Comparative evaluation of TReND method against traditional techniques on simulated fMRI and 7 functional networks neonatal parcellation using TReND framework.
Conclusions:
Our study shows that TReND framework, with transformer-autoencoder and RNMF-KMeans clustering, outperforms traditional methods, enabling reliable and robust neonatal functional brain parcellation, thereby establishing a foundational neonatal functional brain atlas.
Lifespan Development:
Normal Brain Development: Fetus to Adolescence
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
Segmentation and Parcellation 1
Task-Independent and Resting-State Analysis 2
Neuroinformatics and Data Sharing:
Brain Atlases
Keywords:
Atlasing
Development
FUNCTIONAL MRI
Machine Learning
PEDIATRIC
Other - Brain Connectivity
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?
Are you Internal Review Board (IRB) certified? Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
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:
Provide references using APA citation style.
Edwards, A. D. (2022). The developing human connectome project neonatal data release. Frontiers in neuroscience, 16, 886772.
Huang, H. (2005). DTI tractography based parcellation of white matter: application to the mid-sagittal morphology of corpus callosum. Neuroimage, 26(1), 195-205.
Peng, Q. (2020). Regularized-Ncut: Robust and homogeneous functional parcellation of neonate and adult brain networks. Artificial intelligence in medicine, 106, 101872.
Rousseeuw, P. J. (1987). Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. Journal of computational and applied mathematics, 20, 53-65.
Yeo, B. T. (2011). The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of neurophysiology.