Optimizing the timing of fMRI experiments for estimating BOLD non-linearities
Poster No:
1583
Submission Type:
Late-Breaking Abstract Submission
Authors:
Peter Zeidman1, Shokoufeh Golshani1, Martina Callaghan1
Institutions:
1University College London, London, England
First Author:
Co-Author(s):
Late Breaking Reviewer(s):
Rosanna Olsen
Rotman Research Institute, Baycrest Academy for Research and Education
Toronto, Ontario
Rotman Research Institute, Baycrest Academy for Research and Education
Toronto, Ontario
Introduction:
Timing is critical in the design of fMRI experiments. This is because the BOLD response acts as a filter, providing a window into neural dynamics within a limited range of temporal frequencies. Previously, experimental designs have been optimized for efficiently estimating the amplitude of BOLD responses. Here, we introduce an approach for optimizing designs using a richer characterisation of the BOLD response, which accounts for both their amplitude (first order effects) and their change as a function of time elapsed since the last trial (second order effects). This is particularly relevant for clinical fMRI studies aiming to characterise neurovascular dysfunction, and cognitive studies where non-linear effects are of interest, e.g. Repetition Suppression designs.
Methods:
We simulated a simple parametric experimental design with three levels of stimulus amplitude. We sought to optimise the stimulus duration and the minimum and maximum jitter between trials.
To model the design, we revisited the Volterra series - a generic mathematical device for representing a system. A second-order Volterra series has previously been used to characterise the BOLD response as linear combination of two "kernels": a first order-kernel (the modelled BOLD response to a brief stimulus) and a second-order kernel (the change in response as a function of time since the last stimulus). These kernels can be estimated using the standard GLM framework (Friston et al., 1998).
For each candidate design, we specified a design matrix containing three regressors needed to estimate the first-order kernel and six regressors needed to estimate the second-order kernel. Using the efficiency statistic introduced in Friston et al. (1999), we quantified the efficiency of the design for detecting the amplitude of the BOLD response and non-linearities, under the assumption that noise variance was independent of the design (Mechelli et al., 2003). For each set of design parameters (stimulus duration, minimum and maximum jitter), we generated 40 designs with randomized trial order and recorded the optimal efficiency for each.
To model the design, we revisited the Volterra series - a generic mathematical device for representing a system. A second-order Volterra series has previously been used to characterise the BOLD response as linear combination of two "kernels": a first order-kernel (the modelled BOLD response to a brief stimulus) and a second-order kernel (the change in response as a function of time since the last stimulus). These kernels can be estimated using the standard GLM framework (Friston et al., 1998).
For each candidate design, we specified a design matrix containing three regressors needed to estimate the first-order kernel and six regressors needed to estimate the second-order kernel. Using the efficiency statistic introduced in Friston et al. (1999), we quantified the efficiency of the design for detecting the amplitude of the BOLD response and non-linearities, under the assumption that noise variance was independent of the design (Mechelli et al., 2003). For each set of design parameters (stimulus duration, minimum and maximum jitter), we generated 40 designs with randomized trial order and recorded the optimal efficiency for each.
Results:
The optimal experimental design for detecting first-order effects (the amplitude of the BOLD response) was 5s stimulus duration with a varied inter-trial interval between 0-12s. The optimal stimulus duration for detecting non-linearities was also 5s, but with 0-8s ITI. Thus, when non-linear effects are of interest, a jittered event-related design is desirable.
For comparison, we also first fitted a more typical GLM model used in fMRI analysis, with just one regressor per condition. The optimal design for contrasting each condition to the unmodelled baseline was 16s stimulus duration, with a fixed inter-trial interval (ITI) of 12s. The optimal design for contrasting the three stimulus conditions against one another was 20s stimulus duration and no ITI (fixed at 0s). Thus, when non-linear effects are not modelled, a slow block design is optimal.
Finally, we compared the efficiency of our best non-linear design against the gold standard design used in many studies of a common vascular disorder, Cerebral Amyloid Angiopathy (CAA), which use 20s blocks with 28s ITI (Dumas et al., 2012). Our optimized design was more efficient, as illustrated in Figure 2.
For comparison, we also first fitted a more typical GLM model used in fMRI analysis, with just one regressor per condition. The optimal design for contrasting each condition to the unmodelled baseline was 16s stimulus duration, with a fixed inter-trial interval (ITI) of 12s. The optimal design for contrasting the three stimulus conditions against one another was 20s stimulus duration and no ITI (fixed at 0s). Thus, when non-linear effects are not modelled, a slow block design is optimal.
Finally, we compared the efficiency of our best non-linear design against the gold standard design used in many studies of a common vascular disorder, Cerebral Amyloid Angiopathy (CAA), which use 20s blocks with 28s ITI (Dumas et al., 2012). Our optimized design was more efficient, as illustrated in Figure 2.
·Comparison of efficiency of different experimental designs. The axes of each plot identify the minimum and maximum duration of gaps (inter-trial intervals, ITIs) between trials.
·Comparison of (unitless) efficiency scores, between our optimized experimental designs and the gold standard typically used in studies of a vascular disorder, CAA (20s on, 28s off blocks).
Conclusions:
It is well-established that "event-related" experimental designs, which use brief stimuli and interleave different experimental conditions, are more efficient for estimating the shape of the BOLD response than slower "block" designs. Despite this, many studies avoid event-related designs, over concerns that non-linear effects will confound standard (Linear Time Invariant, LTI) analyses. Our results show that by properly accounting for these non-linearities using a Volterra series, together with optimising the timing, a design can be found that offers greater statistical efficiency than currently popular block designs.
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI)
Methods Development 1
Univariate Modeling
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Cerebral Metabolism and Hemodynamics 2
Keywords:
Blood
Data analysis
Design and Analysis
FUNCTIONAL MRI
MRI
Other - BOLD;volterra;GLM
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. (1998). Nonlinear event‐related responses in fMRI. Magnetic resonance in medicine, 39(1), 41-52.
Friston, K. J. (1999). Stochastic designs in event-related fMRI. Neuroimage, 10(5), 607-619.
Mechelli, A. (2003). Estimating efficiency a priori: a comparison of blocked and randomized designs. Neuroimage, 18(3), 798-805.