Detection of atypical sulcal patterns via deep graph-based normative modeling
Poster No:
1561
Submission Type:
Abstract Submission
Authors:
Hyeokjin Kwon1,2,3, P. Ellen Grant1,2,3,4, Jane Newburger3,5, Jong-Min Lee6,7,8, Kiho Im1,2,3
Institutions:
1Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital, Boston, MA, 2Division of Newborn Medicine, Boston Children’s Hospital, Boston, MA, 3Department of Pediatrics, Harvard Medical School, Boston, MA, 4Department of Radiology, Harvard Medical School, Boston, MA, 5Department of Cardiology, Boston Children’s Hospital, Boston, MA, 6Department of Artificial Intelligence, Hanyang University, Seoul, Korea, Republic of, 7Department of Biomedical Engineering, Hanyang University, Seoul, Korea, Republic of, 8Department of Electronic Engineering, Hanyang University, Seoul, Korea, Republic of
First Author:
Hyeokjin Kwon
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital|Division of Newborn Medicine, Boston Children’s Hospital|Department of Pediatrics, Harvard Medical School
Boston, MA|Boston, MA|Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital|Division of Newborn Medicine, Boston Children’s Hospital|Department of Pediatrics, Harvard Medical School
Boston, MA|Boston, MA|Boston, MA
Co-Author(s):
P. Ellen Grant
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital|Division of Newborn Medicine, Boston Children’s Hospital|Department of Pediatrics, Harvard Medical School|Department of Radiology, Harvard Medical School
Boston, MA|Boston, MA|Boston, MA|Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital|Division of Newborn Medicine, Boston Children’s Hospital|Department of Pediatrics, Harvard Medical School|Department of Radiology, Harvard Medical School
Boston, MA|Boston, MA|Boston, MA|Boston, MA
Jane Newburger
Department of Pediatrics, Harvard Medical School|Department of Cardiology, Boston Children’s Hospital
Boston, MA|Boston, MA
Department of Pediatrics, Harvard Medical School|Department of Cardiology, Boston Children’s Hospital
Boston, MA|Boston, MA
Jong-Min Lee
Department of Artificial Intelligence, Hanyang University|Department of Biomedical Engineering, Hanyang University|Department of Electronic Engineering, Hanyang University
Seoul, Korea, Republic of|Seoul, Korea, Republic of|Seoul, Korea, Republic of
Department of Artificial Intelligence, Hanyang University|Department of Biomedical Engineering, Hanyang University|Department of Electronic Engineering, Hanyang University
Seoul, Korea, Republic of|Seoul, Korea, Republic of|Seoul, Korea, Republic of
Kiho Im
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital|Division of Newborn Medicine, Boston Children’s Hospital|Department of Pediatrics, Harvard Medical School
Boston, MA|Boston, MA|Boston, MA
Fetal Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital|Division of Newborn Medicine, Boston Children’s Hospital|Department of Pediatrics, Harvard Medical School
Boston, MA|Boston, MA|Boston, MA
Introduction:
Analyzing the global patterning and altered arrangement of sulcal folds provides deeper understanding of the mechanisms of neurodevelopmental differences in psychiatric and neurological disorders (Barkovich et al., 2012). Our previous analytic tool has been performed to assess deviations of target individuals from normative sulcal patterns by using spectral graph matching of sulcal pit-based graphs (Im et al., 2011). However, key challenges include the absence of a standard criterion for defining a typical reference set, time-consuming cost of graph matching, user-defined feature weight sets, and assumptions about uniform node distribution. To address these issues, we adopted unsupervised reconstruction-based anomaly detection framework for normative sulcal pattern analysis on the sulcal pattern graphs.
Methods:
We trained the graph variational auto-encoder (GVAE) model with information bottleneck using vector-quantization (VQ) strategy on healthy control subjects without any defective global patterns of primary sulci and applied it to both healthy and disease patients to quantitatively analyze the abnormal sulcal patterns (Van Den Oord & Vinyals, 2017). A graph neural networks (GNN) encoder extracted hidden representations for each node of an input sulcal pattern graph, and then M learnable codebook vectors were used to quantize the node representations. The quantized embeddings were fed into GNN and dot-product decoders to reconstruct both node attributes and adjacency matrix, respectively. T1w structural MRIs from 1096 subjects (n [male/female] = 498/598, age [mean ± SD, range]: 28.81±3.69) of the Human Connectome Project (HCP)'s S1200 release were used to train the GVAE (Glasser et al., 2013). For evaluation, a total of 345 CHD patients (n = 220/125, age: 15.83±4.67) and 94 typical controls (n = 46/48, age: 15.42±1.87) were included from the existing data at the Boston Children's Hospital and approved by the institutional review board (Maleyeff et al., 2024; Morton et al., 2023). The MRI scans were processed to extract cortical surfaces using the FreeSurfer pipeline (Dale et al., 1999). The whole-brain sulcal pattern graphs were represented with sulcal pits using the algorithm from our previous study, for subjects from HCP and CHD cohorts (Im et al., 2011).
Results:
CHD patients showed significantly increased abnormality scores of sulcal depth in left hemisphere (FDR-corrected P=0.030), right temporal and right parietal lobar regions, compared with healthy control subjects (FDR- corrected P=0.048, and 0.018, respectively) (Fig. 1). However, the sulcal pattern abnormality scores of combined features were not found to have significant difference between the CHD and control participants. When considering the coordinate features, CHD patients showed significantly higher abnormality scores in left parietal lobar region (FDR-corrected P=0.023). Moreover, we visualized the group-level heatmaps for each abnormality score by aligning individual surfaces to MNI ICBM152 template surface (Fig. 2). We observed that specific regions including left and right paracentral lobule exhibiting the high abnormality scores of sulcal depth. High abnormality scores of sulcal coordinates and area were also found in subregions of superior frontal and superior temporal gyrus.
Conclusions:
In this study, unsupervised anomaly detection framework was designed using VQ GVAE-based reconstruction model on sulcal pattern graphs. This provides versatility and potential to be applied to various target disease conditions while only leveraging normal data for training process. We believe that the proposed methods empower further related applications by providing a more sensitive and interpretable AI-based sulcal pattern analytic tool.
Disorders of the Nervous System:
Neurodevelopmental/ Early Life (eg. ADHD, autism)
Lifespan Development:
Normal Brain Development: Fetus to Adolescence
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural) 2
Methods Development 1
Keywords:
Computational Neuroscience
Cortex
Data analysis
Development
Machine Learning
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?
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:
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.
Dale, A. M., Fischl, B., & Sereno, M. I. (1999). Cortical surface-based analysis: I. Segmentation and surface reconstruction. Neuroimage, 9(2), 179-194.
Glasser, M. F., Sotiropoulos, S. N., Wilson, J. A., Coalson, T. S., Fischl, B., Andersson, J. L., Xu, J., Jbabdi, S., Webster, M., & Polimeni, J. R. (2013). The minimal preprocessing pipelines for the Human Connectome Project. Neuroimage, 80, 105-124.
Im, K., Pienaar, R., Lee, J.-M., Seong, J.-K., Choi, Y. Y., Lee, K. H., & Grant, P. E. (2011). Quantitative comparison and analysis of sulcal patterns using sulcal graph matching: a twin study. Neuroimage, 57(3), 1077-1086.
Maleyeff, L., Park, H. J., Khazal, Z. S., Wypij, D., Rollins, C. K., Yun, H. J., Bellinger, D. C., Watson, C. G., Roberts, A. E., & Newburger, J. W. (2024). Meta-regression of sulcal patterns, clinical and environmental factors on neurodevelopmental outcomes in participants with multiple CHD types. Cerebral Cortex, 34(6).
Morton, S. U., Norris-Brilliant, A., Cunningham, S., King, E., Goldmuntz, E., Brueckner, M., Miller, T. A., Thomas, N. H., Liu, C., & Adams, H. R. (2023). Association of potentially damaging de novo gene variants with neurologic outcomes in congenital heart disease. JAMA network open, 6(1), e2253191-e2253191.
Van Den Oord, A., & Vinyals, O. (2017). Neural discrete representation learning. Advances in neural information processing systems, 30.