Prewhitening in time series extraction affects subsequent processing in PPI analysis
Poster No:
1400
Submission Type:
Abstract Submission
Authors:
Vicky He1,2, Bahman Tahayori1,2, David Vaughan1,2, Heath Pardoe1,2, Graeme Jackson1,2, Chris Tailby1,2, David Abbott1,2, for the Australian Epilepsy Project Investigators1
Institutions:
1The Florey Institute of Neuroscience and Mental Health, Melbourne, Australia, 2University of Melbourne, Melbourne, Australia
First Author:
Vicky He
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
Co-Author(s):
Bahman Tahayori
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
David Vaughan
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
Heath Pardoe, PhD
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
Graeme Jackson
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
Chris Tailby
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
David Abbott, PhD
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
The Florey Institute of Neuroscience and Mental Health|University of Melbourne
Melbourne, Australia|Melbourne, Australia
for the Australian Epilepsy Project Investigators
The Florey Institute of Neuroscience and Mental Health
Melbourne, Australia
The Florey Institute of Neuroscience and Mental Health
Melbourne, Australia
Introduction:
Psychophysiological interaction (PPI) analysis is a widely used regression method that tracks how connections between a seed region and others change in a task-dependent manner (Friston et al., 1997). The regression includes the main task effect, the seed's time course main effect, and the interaction (PPI effect) between the two. To construct the interaction term, it is recommended to multiply the mean-centred task regressor (Di et al., 2017) by the deconvolved seed time course (Gitelman, 2003), followed by convolving their product with the haemodynamic response function (HRF).
The first step in a PPI analysis is usually to extract a seed region's time series. SPM's (https://www.fil.ion.ucl.ac.uk/spm/) extraction command includes high-pass filtering, prewhitening, and confound regression (Fig. 1a). We argue that this affects two downstream processes: 1) prewhitening could alter the structure of HRF, making deconvolution of the prewhitened data potentially suboptimal (Fig. 1b); and 2) prewhitening is applied again during PPI model fitting, which means that the seed regressor is whitened twice. We propose multiplying the extracted seed time series by the inverse of the whitening matrix as a potential solution.
The first step in a PPI analysis is usually to extract a seed region's time series. SPM's (https://www.fil.ion.ucl.ac.uk/spm/) extraction command includes high-pass filtering, prewhitening, and confound regression (Fig. 1a). We argue that this affects two downstream processes: 1) prewhitening could alter the structure of HRF, making deconvolution of the prewhitened data potentially suboptimal (Fig. 1b); and 2) prewhitening is applied again during PPI model fitting, which means that the seed regressor is whitened twice. We propose multiplying the extracted seed time series by the inverse of the whitening matrix as a potential solution.
Methods:
We previously confirmed that mean-centring of regressors is required when deconvolution is undertaken as part of a PPI analysis (He et al., 2024). We have now repeated this analysis to examine the independent impact of mean-centring and whitening inversion. This resulted in four analyses: 1) without whitening inversion and without mean-centring, which is the default approach used in the gPPI toolbox (McLaren et al., 2012), a widely used toolbox for implementing PPI analysis; 2) with whitening inversion but without mean-centring; 3) without whitening inversion and with mean-centring, which is the recommended approach by Di et al. (2017); and 4) with whitening inversion and with mean-centring (our proposed approach).
A total of 201 participants from the Australian Epilepsy Project completed a block design language fMRI task contrasting rhyming against pattern matching. We performed a PPI analysis seeding from the left fusiform gyrus (FusG) to test for upregulation of information flow between FusG and major language nodes during rhyming (positive PPI effects).
A total of 201 participants from the Australian Epilepsy Project completed a block design language fMRI task contrasting rhyming against pattern matching. We performed a PPI analysis seeding from the left fusiform gyrus (FusG) to test for upregulation of information flow between FusG and major language nodes during rhyming (positive PPI effects).
Results:
Without mean-centring or whitening inversion, we observed implausible PPI effects covering essentially the whole brain, with a peak at the seed location (Fig. 2a). Since a main effect seed regressor was included in the PPI model, this finding is erroneous, as most variance at the seed location should have been accounted for by the seed regressor. Whitening inversion successfully removed the spurious effects at the seed (Fig. 2b). The remaining widespread effects support the argument made by Di et al. (2017) that mean-centring is necessary; otherwise, the PPI term carries unwanted correlation with the seed due to imperfect deconvolution. As confirmed in Fig. 2c, mean-centring largely resolved this issue. With both mean-centring and whitening inversion, the key results are reinforced, and expected clusters appear larger with higher t-scores (Fig. 2d).
Conclusions:
We propose including both a whitening inversion step and a mean-centring step in PPI analysis. An alternative is to skip prewhitening in time series extraction, as beta estimates in confound regression remain asymptotically unbiased despite autocorrelation. However, this would require significant code changes, as SPM's confound regression relies on prewhitened first level beta maps. In conclusion, mean-centring prevents main effects from contaminating the PPI term, while whitening inversion further refines PPI estimates. The proposed changes make PPI a more effective method for studying how the brain actively adapts to changing task demands.
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
fMRI Connectivity and Network Modeling 1
Methods Development 2
Univariate Modeling
Keywords:
Data analysis
FUNCTIONAL MRI
Modeling
Statistical Methods
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.
Friston, K. J., Buechel, C., Fink, G. R., Morris, J., Rolls, E., & Dolan, R. J. (1997). Psychophysiological and modulatory interactions in neuroimaging. NeuroImage, 6(3), 218–229.
Gitelman, D. R., Penny, W. D., Ashburner, J., & Friston, K. J. (2003). Modeling regional and psychophysiologic interactions in fMRI: The importance of hemodynamic deconvolution. NeuroImage, 19(1), 200–207.
He, V., Tahayori, B., Vaughan, D.N., Jackson, G.D., Abbott, D.F., Tailby, C., & for the Australian Epilepsy Project Investigators. (2024, June). Dynamic modulation of information flow from occipitotemporal cortex according to cognitive demands. OHBM 2024 – 30th Annual Meeting of the Organization for Human Brain Mapping, Seoul, Korea.
McLaren, D. G., Ries, M. L., Xu, G., & Johnson, S. C. (2012). A generalized form of context-dependent psychophysiological interactions (gPPI): A comparison to standard approaches. NeuroImage, 61(4), 1277–1286.