Patterns of coordinated cortical thinning over normal aging over large data using nonnegative matrix
Poster No:
925
Submission Type:
Abstract Submission
Authors:
Sung Min Ha1, Abdalla Bani1, Aristeidis Sotiras1
Institutions:
1Washington University in St. Louis, Saint Louis, MO
First Author:
Co-Author(s):
Introduction:
Cortical thinning is a well-established part of normal aging, and its accelerated rate of thinning is an important pattern observed in many neurodegenerative disease. However, patterns of cortical thinning is not homogeneous and different regions of the brain can have disparate patterns of thinning. Such regional heterogeneity has been understudied largely due to the conventional assumption of gyral based parcellation of the brain structure(Sotiras, 2017). Furthermore, difficulty of analyzing large scale, high resolution neuroimaging data restricted the statistical power of analysis arising from larger sample size. Nonnegative matrix factorization (NMF) is a dimensionality reduction tool widely used in the neuroimaging community to study patterns of covariance in a hypothesis-free, data-driven way.(Sotiras, 2015) Using a variant of NMF tailored towards high resolution surface data, multiresolution orthonormal projective NMF (m-opNMF), (Ha, 2024) we derived non-overlapping, interpretable components from large healthy cohort from UK Biobank that group cortical regions based on their rate of thinning. Using NMF, we acquired patterns that suggest that the patterns of cortical thinning does not align with structural features, and that there may be asymmetry between the hemispheres in certain regions.
Methods:
opNMF (Yang, 2010) is an iterative matrix factorization method that approximates a nonnegative matrix X as a multiplication of two smaller nonnegative matrices, component W and coefficient H, by minimizing the frobenius norm of X-WH. Due to the high memory constraint and computational cost of opNMF, we leveraged graph coarsening method with preserved spectral properties (Loukas, 2019) to reduce the surface data to lower resolution and performed the multiplicative update hierarchically at lower resolutions before acquiring components at original resolution.
Results:
We used cortical thickness from baseline T1 MRI scans of 5992 subjects (2997 female, age 62.22 ± 7.20 ranging from 46 to 82) from UK Biobank. These scans were preprocessed with Freesurfer pipeline (Fischl, 2012), registered to fsaverage6, smoothed with 4mm FWHM kernel, then vectorized and concatenated to generate matrix X of size 81924 by 5992. m-opNMF was run with rank k=20, lowest resolution at 1/8th of original, with 5 intermediate resolutions to reduce computational burden.
Visualizing individual 20 components of W in Fig 1, the components group vertices that covary and decrease in thickness in a similar rate within the region. Notably, while most components exhibit a degree of symmetry between the two hemispheres, the components emphasizing the temporal region diverge into two distinct components (component 2 predominantly on the left and component 3 primarily on the right). Using the slope of the linear regression predicting values of the H per subject from the actual age of the subject in Fig 2, we noted that the right temporal region in component 3 has slightly faster rate of thinning compared to the left hemisphere with similar y-intercept.
In fig 2, we noted that the components 4, 5, and to a lesser extent 9, and 10, primarily located frontal superior regions have particularly high rate of thinning given the steep slope less than -3.8. In contrast, regions 6 (Gyral temporal inf), 7 (Occipital Pole), 11 (G rectus), 15 (G insular short), 18 (temporal pole), and 20 (orbital gyri) have less steep slope of decline compared to the average. It is worth noting that these slow thinning regions occur in common at temporal and occipital regions, and particularly near the medial side.
Visualizing individual 20 components of W in Fig 1, the components group vertices that covary and decrease in thickness in a similar rate within the region. Notably, while most components exhibit a degree of symmetry between the two hemispheres, the components emphasizing the temporal region diverge into two distinct components (component 2 predominantly on the left and component 3 primarily on the right). Using the slope of the linear regression predicting values of the H per subject from the actual age of the subject in Fig 2, we noted that the right temporal region in component 3 has slightly faster rate of thinning compared to the left hemisphere with similar y-intercept.
In fig 2, we noted that the components 4, 5, and to a lesser extent 9, and 10, primarily located frontal superior regions have particularly high rate of thinning given the steep slope less than -3.8. In contrast, regions 6 (Gyral temporal inf), 7 (Occipital Pole), 11 (G rectus), 15 (G insular short), 18 (temporal pole), and 20 (orbital gyri) have less steep slope of decline compared to the average. It is worth noting that these slow thinning regions occur in common at temporal and occipital regions, and particularly near the medial side.
·Visualization of the left and right hemispheres of each component from m-opNMF.
·Linear regression predicting each rank of normalized coefficient H matrix from actual age
Conclusions:
The results demonstrate that the patterns of cortical thinning do not align with gyral boundaries of traditional ROI. Using NMF, a data-driven approach, we illuminated thinning patterns that align with functional specialization; frontal regions relating to higher cognitive function thin faster while regions associated with vision are slower.
Lifespan Development:
Aging 1
Modeling and Analysis Methods:
Methods Development 2
Segmentation and Parcellation
Other Methods
Keywords:
Aging
Cortex
Data analysis
Segmentation
STRUCTURAL MRI
Other - Nonnegative Matrix Factorization
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.
Ha, S. M. (2024, June 26). Multiresolution orthonormal projective nonnegative matrix factorization for large surface data [Conference Poster]. Organization for Human Brain Mapping 2024 Annual Meeting, Seoul, South Korea.
Loukas, A. (2019). Graph Reduction with Spectral and Cut Guarantees. Journal of Machine Learning Research, 20(116), 1–42.
Sotiras, A. (2015). Finding imaging patterns of structural covariance via Non-Negative Matrix Factorization. NeuroImage, 108, 1–16. https://doi.org/10.1016/j.neuroimage.2014.11.045
Sotiras, A. (2017). Patterns of coordinated cortical remodeling during adolescence and their associations with functional specialization and evolutionary expansion. Proceedings of the National Academy of Sciences, 114(13), 3527–3532. https://doi.org/10.1073/pnas.1620928114
Yang, Z. (2010). Linear and Nonlinear Projective Nonnegative Matrix Factorization. IEEE Trans. on Neural Networks, 21(5), 734–749. https://doi.org/10.1109/TNN.2010.2041361