Statistical Power Enhancement with SPARC: Spectral Permutation Analysis for Robust Correction
Poster No:
1556
Submission Type:
Abstract Submission
Authors:
Niousha Dehestani1, Sina Mansour L.1, Thomas Yeo2, Juan Helen Zhou1
Institutions:
1National University of Singapore, Singapore, Singapore, 2Centre for Sleep and Cognition, National University of Singapore, Singapore, Singapore
First Author:
Co-Author(s):
Introduction:
Detecting statistically meaningful patterns in brain activity faces challenges like low sensitivity from multiple comparison corrections (Eklund et al., 2016; Noble et al., 2020). While large-scale studies reveal widespread network effects across the neocortex and subcortical regions, most task-fMRI research relies on smaller samples, limiting statistical power and resulting in the detection of localized regions indicating only the highest effect sizes (Noble et al., 2022). This reduced power mainly stems from the high dimensionality of neuroimaging data, demanding multiple testing corrections across hundreds of thousands of voxels/vertices. Brain eigenmodes, derived from structural properties like cortical geometry and axonal connections, offer a solution by reducing high-dimensional brain maps to low-dimensional spaces, capturing most of the signal energy while removing high-frequency noise components (Mansour et al., 2024; Pang et al., 2023). This can be utilized as a means to address power limitations in high-dimensional brain imaging. Here, we introduce Spectral Permutation Analysis for Robust Correction (SPARC), a novel method that enhances the statistical power of multiple comparison corrections via nonparametric permutation analysis through eigenmode-based spectral filtering.
Methods:
SPARC builds on permutation-based nonparametric multiple comparison correction (Winkler et al., 2014) by integrating a graph-spectral filter in the permutation step (Figure 1). In short, using a general linear model (GLM), individual-level statistical maps are created, reflecting voxel- or vertex-wise activity variations between contrasts. These maps are subsequently used to compute group-level statistics. Group-level maps are reconstructed via low-pass graph Fourier filtering utilizing structural connectome brain eigenmodes. This filter retains low-frequency signal components representing structured brain activity while removing high-frequency components of noise. Through permutation testing, null distributions of low-pass approximated statistics are generated, enabling nonparametric voxel-wise FWER correction. SPARC assumes low-pass approximations contain critical signal variance and optimizes the number of eigenmodes (K) to balance accuracy and dimensionality. This study used Human connectome Project (HCP) data to benchmark SPARC's inference performance across different cognitive tasks (including Gambling, Emotion, Social, Working memory, and Relational) and sample sizes (10, 20, 30, 40, 50, 60, 80, 100, and 200) demonstrating its superior sensitivity and robustness.
Results:
We demonstrated the efficacy of SPARC in improving the accuracy and sensitivity of brain activity detection across varying sample sizes and tasks. First, we observed a significant improvement in the correlation between the spectral-filtered group-level maps generated by SPARC in small samples and the group-level maps computed from large samples, with an approximate 20% increase in alignment (Fig.1). This underscores the advantage of SPARC in reconstructing brain activation patterns. Moreover, we compared SPARC with conventional nonparametric methods for voxel-wise activation inference. We find that SPARC yields 20% to 35% improvement in statistical power compared to conventional methods, highlighting its ability to detect brain activity patterns at smaller sample sizes. This improvement was consistent across tasks, suggesting that SPARC is robust and generalizable across diverse experimental paradigms. Lastly, we conducted additional evaluations to ensure that this increased power was not at the cost of increased false discovery. By using the bookmaker Informedness, we found an approximately 20% increase in informedness when using SPARC. This ensures that sensitivity improvements achieved are not at the expense of inflating false positive errors.
Conclusions:
Overall, SPARC is a valuable tool for researchers working with smaller sample sizes or investigating complex, distributed spatial patterns.
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 2
Methods Development 1
Keywords:
ADULTS
Aging
Cognition
Computational Neuroscience
Emotions
fMRI CONTRAST MECHANISMS
FUNCTIONAL MRI
Memory
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.
2. Noble, S., Scheinost, D., & Constable, R. T. (2020). Cluster failure or power failure? Evaluating sensitivity in cluster-level inference. Neuroimage, 209, 116468.
3. Noble, S., Mejia, A. F., Zalesky, A., & Scheinost, D. (2022). Improving power in functional magnetic resonance imaging by moving beyond cluster-level inference. Proceedings of the National Academy of Sciences, 119(32), e2203020119.
4. Mansour L, Sina, Hamid Behjat, Dimitri Van De Ville, Robert E. Smith, BT Thomas Yeo, and Andrew Zalesky. "Eigenmodes of the brain: revisiting connectomics and geometry." bioRxiv (2024): 2024-04.
5. Pang, J. C., Aquino, K. M., Oldehinkel, M., Robinson, P. A., Fulcher, B. D., Breakspear, M., & Fornito, A. (2023). Geometric constraints on human brain function. Nature, 618(7965), 566-574.
6.Winkler, A. M., Ridgway, G. R., Webster, M. A., Smith, S. M., & Nichols, T. E. (2014). Permutation inference for the general linear model. Neuroimage, 92, 381-397.