Population receptive field estimates in human visual cortex
Introduction
Bridging the gap between technologies that measure neural signals at different length scales is an important objective in neuroimaging research. Here, we introduce new experimental and computational methods that quantitatively couple fMRI signals, measured at the millimeter scale, with receptive field properties of visual neurons measured at the micron scale.
The new methods build on techniques that were developed for visual field mapping (DeYoe et al., 1996, Dumoulin et al., 2003, Engel et al., 1997, Engel et al., 1994, Sereno et al., 1995). In visual field mapping, the experimenter measures responses to contrast-defined rings and wedges shown at a series of visual field locations. Conventionally, the responses to these stimuli are used to estimate the visual field position that produces the largest fMRI response for each voxel.
While conventional methods estimate only the most effective visual field location, the neuronal population within a voxel in fact responds to a range of visual field locations. The region of visual space that stimulates the recording site is the population receptive field (pRF) (Victor et al., 1994). The pRF method estimates the visual field map, the pRF size, laterality and surround suppression using the temporal responses to multiple stimuli.
Intuitively, the pRF size can be estimated because we understand how pRF size influences the fMRI time course. This pRF influence was observed by Tootell et al. (1997), who noticed different time courses in V1 and V3A in response to expanding ring stimuli commonly used to map the visual field (Engel et al., 1994). They explained this time course difference by suggesting that pRF sizes in V3A exceed those of V1. Smith et al. (2001) extended this analysis by measuring the relative amount of active versus inactive epochs (duty cycle) in the fMRI response to the ring stimulus (see also Larsson and Heeger, 2006, Li et al., 2007). In addition to comparisons between areas, Smith et al. (2001) also found a systematic relationship between the duty cycle and visual eccentricity.
The advance we describe here is a quantitative framework to model pRF properties and fit these models to the fMRI time series. Specifically, we show how to derive visual field maps and pRF sizes by integrating data from multiple types of stimuli, including rings, wedges, and moving bars. We show that this model-based approach reconstructs the cortical visual field map more accurately than conventional mapping methods. We use the pRF method to obtain quantitative estimates of population receptive field sizes in lateral and ventral occipital regions of human visual cortex. Finally, we show that the human pRF size estimates in areas V1–V3 agree well with electrophysiological receptive field measurements in the corresponding areas in monkey and human visual cortex.
Section snippets
Subjects
Measurements were obtained from six subjects (one female; ages 24–36 years). All subjects participated in experiments containing wedges and ring stimuli, three subjects participated in experiments containing the moving bar stimuli. All subjects had normal or corrected-to-normal visual acuity. All studies were performed with the informed written consent of the subjects and were approved by the Stanford Institutional Review Board.
Stimulus presentation
The visual stimuli were generated in the Matlab programming
V1 and LO respond differently to identical traveling wave stimuli
The V1 and lateral occipital (LO) responses to the same stimulus differ qualitatively. The time series in Figs. 3A and B are typical examples of how the responses to a rotating wedge stimulus differ. These voxels respond preferentially to a similar wedge position, conventional analyses would assign them similar phases (Engel et al., 1994, Sereno et al., 1995), and duty-cycles or duty-cycle-related measures (21.5% and 33.5%, respectively) (Larsson and Heeger, 2006, Li et al., 2007, Smith et al.,
Discussion
We introduce a functional MRI method that computes a model of the pRF from responses to a wide range of stimuli. The method estimates both a visual field map estimate as well as other neuronal population properties, including size and laterality. We show that the visual field maps are more accurate than conventional methods. The eccentricity maps are improved by the new method, whereas the polar-angle maps around the fovea are improved by the new methods in combination with new stimuli. We
Conclusion
The pRF method links fMRI measurements at the millimeter scale to response neuronal properties at the micron scale and reduces the gap between functional imaging and electrophysiology.
The estimated human pRF agrees with electrophysiological and conventional fMRI estimates in three ways. First, the pRF-derived visual field maps agree well with those estimated from monkey electrophysiology and conventional human fMRI visual field mapping methods (Fig. 5). Second, pRF size increases systematically
Acknowledgments
This work was supported by a National Eye Institute Grant RO1 EY03164 to BW and a Larry L. Hillblom Foundation fellowship 2005/2BB to SD. We thank Kaoru Amano, Michal Ben-Shachar, Alyssa Brewer, Sing-Huang Cheung, Robert Dougherty, Kalanit Grill-Spector, Yoichiro Masuda, and Rory Sayres for their help and comments on the manuscript. Parts of this work have been published in abstract form (Dumoulin et al., 2006). The software is freely available and is distributed as part of the VISTA software (//white.stanford.edu/software/
References (78)
- et al.
The variability of human, BOLD hemodynamic responses
NeuroImage
(1998) - et al.
Spatial heterogeneity of the nonlinear dynamics in the fMRI BOLD response
NeuroImage
(2001) - et al.
Receptive field size-dependent attention effects in simultaneously presented stimulus displays
NeuroImage
(2006) - et al.
Anatomical and physiological asymmetries related to visual areas V3 and VP in macaque extrastriate cortex
Vision Res.
(1986) - et al.
Temporal dynamics of the BOLD fMRI impulse response
NeuroImage
(2005) - et al.
Automatic volumetric segmentation of human visual retinotopic cortex
NeuroImage
(2003) - et al.
Analysis of fMRI time-series revisited
NeuroImage
(1995) - et al.
Event-related fMRI: characterizing differential responses
NeuroImage
(1998) Deconvolution of impulse response in event-related BOLD fMRI
NeuroImage
(1999)- et al.
The lateral occipital complex and its role in object recognition
Vision Res.
(2001)