Stimulus smoothness influences pRF parameters
Poster No:
1350
Submission Type:
Abstract Submission
Authors:
David Linhardt1, Johanna Bogner1, Michael Woletz1, D. Samuel Schwarzkopf2, Christian Windischberger1
Institutions:
1High Field MR Center, Medical University of Vienna, Austria, 2University of Auckland, New Zealand
First Author:
Co-Author(s):
Introduction:
Population receptive field (pRF) modeling with functional MRI (fMRI) is the gold-standard method for revealing how a subject's visual field maps onto the visual cortex. Subjects maintain focus on a central point while a bar aperture moves across their visual field. Traditionally, this bar jumps to a new position with each fMRI volume captured, tying the model's temporal resolution to the scan's repetition time (TR). This method, however, also impacts spatial resolution, as it fails to distinguish small pRFs within the range of a single bar jump. Further, the jump width of the stimulus also imposes a constraint on the pRF size that cannot be effectively analyzed when too small. To assess the impact of the stimulus smoothness on pRF mapping, here we tested stimuli varying in their movement patterns, from classical jumps to smooth transitions, and explicitly incorporated stimulus temporal resolution in the pRF model.
Methods:
We designed 3 stimulus conditions based on a classical pRF mapping paradigm. A bar (width=1.6°) moved through the field of view in eight different directions, revealing a reversing checkerboard pattern. Subjects were instructed to fixate on the central dot. The stimulus reached a total diameter of 12° visual angle:
Standard: the bar jumped once per TR for ⅓ of the bar width
Small jumps: the bar jumped twice per TR
smooth: the bar jumped 8 times per TR, resulting in smooth movement
A total of 6 healthy subjects (5 female, age: 24.5±2.6) were measured on a 7T SIEMENS MAGNETOM scanner. For every stimulus condition, we acquired two functional runs (TE/TR=25.6/1000ms, 1.2mm isotropic) in one session, resulting in 6 runs per subject. Together with an additional structural T1-weighted image (0.75mm isotropic) the data were minimally pre-processed using containerized solution fMRIPrep v23.1.4 (fmriprep.org). For the pRF mapping, the analysis tool SamSrf v9.7 (github.com/samsrf/samsrf)[1] was used, which allows for the creation of the pRF model at a higher temporal resolution. Using this option, the analysis created models in the native temporal resolution of the stimulus. These high resolution models were then subsampled to the timings of the acquired data (at every TR). This procedure was used to estimate the pRF position (eccentricity, polar angle) and size (sigma). Only data in V1 and exceeding a variance explained threshold of 10% were included in the analysis.
Coverage plots were created as the maximum surface of all pRFs using prfresult v0.1.1 (github.com/dlinhardt/prfresult). Reproducibility was assessed using Spearman's correlation, between the two runs of each condition. For every pRF parameter, the correlation was calculated for every subject independently and then averaged across subjects.
Standard: the bar jumped once per TR for ⅓ of the bar width
Small jumps: the bar jumped twice per TR
smooth: the bar jumped 8 times per TR, resulting in smooth movement
A total of 6 healthy subjects (5 female, age: 24.5±2.6) were measured on a 7T SIEMENS MAGNETOM scanner. For every stimulus condition, we acquired two functional runs (TE/TR=25.6/1000ms, 1.2mm isotropic) in one session, resulting in 6 runs per subject. Together with an additional structural T1-weighted image (0.75mm isotropic) the data were minimally pre-processed using containerized solution fMRIPrep v23.1.4 (fmriprep.org). For the pRF mapping, the analysis tool SamSrf v9.7 (github.com/samsrf/samsrf)[1] was used, which allows for the creation of the pRF model at a higher temporal resolution. Using this option, the analysis created models in the native temporal resolution of the stimulus. These high resolution models were then subsampled to the timings of the acquired data (at every TR). This procedure was used to estimate the pRF position (eccentricity, polar angle) and size (sigma). Only data in V1 and exceeding a variance explained threshold of 10% were included in the analysis.
Coverage plots were created as the maximum surface of all pRFs using prfresult v0.1.1 (github.com/dlinhardt/prfresult). Reproducibility was assessed using Spearman's correlation, between the two runs of each condition. For every pRF parameter, the correlation was calculated for every subject independently and then averaged across subjects.
Results:
Differences in coverage plots are shown for two subjects in Figure 1. The columns represent results obtained from the standard, small jump and smooth stimulus respectively. The standard analysis revealed pronounced clustering artifacts in both subjects, with the clustered pRFs exhibiting sizes notably smaller than typical. While diminished, these artifacts persisted in the small jump stimulus. For both conditions, the distance between clusters roughly corresponds to the jump width of the bar. Only results for the smoothly moving bar in the third column are free of these artifacts.
Comparative analysis of reproducibility across standard, small jump, and smooth stimuli yields negligible differences, with values for standard (eccentricity: 0.91; polar angle: 0.97; sigma: 0.26), small jump (0.92; 0.98; 0.27), and smooth (0.91; 0.96; 0.29) stimulation. Variance explained averages across subjects show a marginal increase for small (26%) and smooth (25%) compared to standard (24%) stimulation.
Comparative analysis of reproducibility across standard, small jump, and smooth stimuli yields negligible differences, with values for standard (eccentricity: 0.91; polar angle: 0.97; sigma: 0.26), small jump (0.92; 0.98; 0.27), and smooth (0.91; 0.96; 0.29) stimulation. Variance explained averages across subjects show a marginal increase for small (26%) and smooth (25%) compared to standard (24%) stimulation.
Conclusions:
Though we could not find an improved performance of the smooth stimuli in terms of reproducibility or goodness of fit, artifacts driven by tiny pRFs were heavily reduced and we found an overall more uniform coverage using a smoother stimulus design.
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 1
Other Methods
Novel Imaging Acquisition Methods:
BOLD fMRI
Perception, Attention and Motor Behavior:
Perception: Visual 2
Keywords:
FUNCTIONAL MRI
MRI
Vision
Other - pRF mapping, retinotopy
1|2Indicates the priority used for review