Advancing HRF Analysis: Thin Plate Spline Regression for Adaptive Spatial Modeling in fMRI
Presented During: Poster Session 3
Friday, June 27, 2025: 01:45 PM - 03:45 PM
Friday, June 27, 2025: 01:45 PM - 03:45 PM
Presented During: Poster Session 4
Saturday, June 28, 2025: 01:45 PM - 03:45 PM
Saturday, June 28, 2025: 01:45 PM - 03:45 PM
Poster No:
1565
Submission Type:
Abstract Submission
Authors:
Jungin Choi1, Martin Lindquist1, Abhirup Datta1
Institutions:
1Johns Hopkins University, Baltimore, MD
First Author:
Co-Author(s):
Introduction:
A majority of fMRI studies focus on analyzing the Blood Oxygenation Level Dependent (BOLD) response, which is typically modeled as the convolution of a stimulus function with a hemodynamic response function (HRF). Thus, the HRF plays a key role in linking neural activity to the observed BOLD signal. However, over 90% of fMRI studies rely on a fixed canonical HRF across all individuals and brain regions. This simplifying assumption overlooks individual differences and spatial variations, potentially leading to inaccurate HRF estimates.
In addition to the limitations of using a fixed HRF, another challenge is the use of Gaussian kernel smoothing for spatial modeling. Although Gaussian smoothing is widely used to reduce measurement noise and improve spatial coherence, it uses a uniform bandwidth across all brain regions, ignoring the heterogeneity in structural and functional properties. This results in over-smoothing in regions with fine-grained structures and under-smoothing in larger, more homogeneous areas. Furthermore, Gaussian kernel methods smooth and alter the BOLD signal data itself for subsequent analysis with significant noise underestimation.
In addition to the limitations of using a fixed HRF, another challenge is the use of Gaussian kernel smoothing for spatial modeling. Although Gaussian smoothing is widely used to reduce measurement noise and improve spatial coherence, it uses a uniform bandwidth across all brain regions, ignoring the heterogeneity in structural and functional properties. This results in over-smoothing in regions with fine-grained structures and under-smoothing in larger, more homogeneous areas. Furthermore, Gaussian kernel methods smooth and alter the BOLD signal data itself for subsequent analysis with significant noise underestimation.
Methods:
Thin Plate Spline (TPS) Regression Model
To address the limitations, we employed the TPS regression for adaptive spatial modeling. This method reformulates the HRF estimation problem by modeling β with splines, introducing a new variable Γ. The number of TPS is optimized using criteria like AIC or BIC. Since the TPS regression captures adaptive spatial structure of HRF, linear regression can fit the observed BOLD signal without changing the original data while using information from neighboring voxels for smoothing.
Advanced HRF Estimation Models
In addition to thin plate spline regression, we developed and integrated advanced HRF methods such as B-splines and Smooth FIR.
Evaluation Framework
We compared these advanced methods across non-spatial, Gaussian kernel smoothing, and TPS regression models using simulated data and two fMRI datasets (HCP Motor Task Data and SpaceTop Pain Task Data). Performance was assessed in terms of HRF estimation accuracy and inferential performances in capturing neural activation.
To address the limitations, we employed the TPS regression for adaptive spatial modeling. This method reformulates the HRF estimation problem by modeling β with splines, introducing a new variable Γ. The number of TPS is optimized using criteria like AIC or BIC. Since the TPS regression captures adaptive spatial structure of HRF, linear regression can fit the observed BOLD signal without changing the original data while using information from neighboring voxels for smoothing.
Advanced HRF Estimation Models
In addition to thin plate spline regression, we developed and integrated advanced HRF methods such as B-splines and Smooth FIR.
Evaluation Framework
We compared these advanced methods across non-spatial, Gaussian kernel smoothing, and TPS regression models using simulated data and two fMRI datasets (HCP Motor Task Data and SpaceTop Pain Task Data). Performance was assessed in terms of HRF estimation accuracy and inferential performances in capturing neural activation.
Results:
Simulated data featured an 8-shaped continuous spatial distribution of the true β. Comparisons among three HRF estimation approaches-the non-spatial, the Gaussian kernel smoothing, and the TPS regression model-showed that the TPS regression model outperformed the others, including the kernel-adaptive Gaussian smoothing model. It showed superior estimation accuracy and inference accuracy, particularly in reducing false positive rates.
When applied to fMRI datasets, including the HCP Motor Task Data and SpaceTop Pain Task Data, the TPS regression model showed notably high performances. It provided greater stability of HRF estimates across brain regions, effectively recovered spatial activation patterns such as the motor homunculus, and achieved significant noise reduction while minimizing bias in HRF amplitude estimates.
When applied to fMRI datasets, including the HCP Motor Task Data and SpaceTop Pain Task Data, the TPS regression model showed notably high performances. It provided greater stability of HRF estimates across brain regions, effectively recovered spatial activation patterns such as the motor homunculus, and achieved significant noise reduction while minimizing bias in HRF amplitude estimates.
Conclusions:
This study clearly demonstrates the advantages of TPS regression over conventional Gaussian kernel smoothing for spatial modeling in HRF analysis. By leveraging adaptive smoothing and preserving the original BOLD signal, TPS regression achieves:
1. Improved HRF estimation accuracy across simulations and real datasets
2. Stable inference and spatial coherence in HRF estimates
3. Enhanced ability to capture fine-grained spatial activation patterns
Furthermore, advanced HRF modeling methods provide strong alternatives to canonical HRF models. These results highlight TPS regression with advanced HRF estimation models as an effective tool for enhancing spatial modeling and inference in fMRI studies.
1. Improved HRF estimation accuracy across simulations and real datasets
2. Stable inference and spatial coherence in HRF estimates
3. Enhanced ability to capture fine-grained spatial activation patterns
Furthermore, advanced HRF modeling methods provide strong alternatives to canonical HRF models. These results highlight TPS regression with advanced HRF estimation models as an effective tool for enhancing spatial modeling and inference in fMRI studies.
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 2
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling
Methods Development 1
Novel Imaging Acquisition Methods:
BOLD fMRI
Keywords:
Design and Analysis
FUNCTIONAL MRI
Modeling
Statistical Methods
Other - Hemodynamic Response Function
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.
2. Lindquist, M. A., & Wager, T. D. (2007). Validity and power in hemodynamic response modeling: A comparison study and a new approach. Human Brain Mapping, 28(8), 764–784. https://doi.org/10.1002/hbm.20310
3. Walther, A., Nili, H., Ejaz, N., Alink, A., Kriegeskorte, N., & Diedrichsen, J. (2016). Reliability of dissimilarity measures for multi-voxel pattern analysis. NeuroImage, 137, 188–200. https://doi.org/10.1016/j.neuroimage.2015.12.012
4. Wood, S. N. (2003). Thin plate regression splines. Journal of the Royal Statistical Society Series B: Statistical Methodology, 65(1), 95–114. https://doi.org/10.1111/1467-9868.00374
5. Degras, D., & Lindquist, M. A. (2014). A hierarchical model for simultaneous detection and estimation in multi-subject fMRI studies. NeuroImage, 98, 61–72. https://doi.org/10.1016/j.neuroimage.2014.04.052
6. Glover, G. H. (2011). Overview of functional magnetic resonance imaging. Neurosurgery Clinics of North America, 22(2), 133–139. https://doi.org/10.1016/j.nec.2010.11.001
7. Jung, H., Amini, M., Hunt, B. J., Murphy, E. I., Sadil, P., Halchenko, Y. O., Petre, B., et al. (2024). A multimodal fMRI dataset unifying naturalistic processes with a rich array of experimental tasks. Neuroscience. https://doi.org/10.1101/2024.06.21.599974
8. Paciorek, C. J. (2010). The importance of scale for spatial-confounding bias and precision of spatial regression estimators. Statistical Science, 25(1). https://doi.org/10.1214/10-STS326
9. Phạm, D. Đ., McDonald, D. J., Ding, L., Nebel, M. B., & Mejia, A. F. (2023). Less is more: Balancing noise reduction and data retention in fMRI with data-driven scrubbing. NeuroImage, 270, 119972. https://doi.org/10.1016/j.neuroimage.2023.119972