pubmed.ncbi.nlm.nih.gov

Surround suppression supports second-order feature encoding by macaque V1 and V2 neurons - PubMed

Review

Surround suppression supports second-order feature encoding by macaque V1 and V2 neurons

Luke E Hallum et al. Vision Res. 2014 Nov.

Abstract

Single neurons in areas V1 and V2 of macaque visual cortex respond selectively to luminance-modulated stimuli. These responses are often influenced by context, for example when stimuli extend outside the classical receptive field (CRF). These contextual phenomena, observed in many sensory areas, reflect a fundamental cortical computation and may inform perception by signaling second-order visual features which are defined by spatial relationships of contrast, orientation and spatial frequency. In the anesthetized, paralyzed macaque, we measured single-unit responses to a drifting preferred sinusoidal grating; low spatial frequency sinusoidal contrast modulations were applied to the grating, creating contrast-modulated, second-order forms. Most neurons responded selectively to the orientation of the contrast modulation of the preferred grating and were therefore second-order orientation-selective. Second-order selectivity was created by the asymmetric spatial organization of the excitatory CRF and suppressive extraclassical surround. We modeled these receptive field subregions using spatial Gaussians, sensitive to the modulation of contrast (not luminance) of the preferred carrier grating, that summed linearly and were capable of recovering asymmetrical receptive field organizations. Our modeling suggests that second-order selectivity arises both from elongated excitatory CRFs, asymmetrically organized extraclassical surround suppression, or both. We validated the model by successfully testing its predictions against conventional surround suppression measurements and spike-triggered analysis of second-order form responses. Psychophysical adaptation measurements on human observers revealed a pattern of second-order form selectivity consistent with neural response patterns. We therefore propose that cortical cells in primates do double duty, providing signals about both first- and second-order forms.

Keywords: Filter-rectify-filter; Primary visual cortex; Receptive field; Second-order; Surround suppression; V2.

Copyright © 2014 Elsevier B.V. All rights reserved.

PubMed Disclaimer

Figures

Figure 1
Figure 1

Stimulus construction. We modulated the contrast of a large, circular patch of sinusoidal carrier grating (A) using a relatively low spatial frequency, raised sinusoidal modulator (B) forming the second-order stimulus (C). We optimized the carrier grating’s orientation and spatial frequency for each neuron, and typically set its drift rate to approximately 5 Hz. The modulator always drifted at 0.75 Hz. The contrasts of the carrier grating and the modulator were 75% and 100%, respectively. In each experiment, from trial to trial, we varied the modulator’s drift direction (0, 45, 90, … 315 deg relative to the carrier grating’s drift direction) and the modulator’s spatial frequency (0, 0.125, 0.25, … 0.75× the carrier grating’s spatial frequency). Shown here the modulator drifts at 45 deg relative to the carrier with spatial frequency 0.25×.

Figure 2
Figure 2

First- and second-order orientation-selectivity. (A) Polar plot of “first-order” orientation-tuned responses of neuron m628r24(V1) to a small, circular patch of high-contrast, drifting sinusoidal grating (1.58 c/deg) presented to the classical receptive field (CRF) approximated during hand-mapping. This neuron preferred a grating drifting at 157 deg; here, and in all subsequent plots, we have rotated the data so that the preferred grating is represented as vertical and drifting to the right. (B) Polar plot of “second-order” orientation-tuned responses of the same neuron to sinusoidal contrast modulations (0.79 c/deg) of a large patch (5.1 deg) of carrier grating. For each data point, angle of elevation and radial distance indicate the modulator’s drift direction and the neuron’s modulated response amplitude, respectively. The cycle histograms beneath each stimulus icon show how the spike rate synchronized to the passage of the modulator, which drifted at temporal frequency 0.75 Hz (period = 1.33 s), over the CRF and extraclassical surround. As shown, fundamental (0.75 Hz) response amplitudes reliably signaled the orientation of the contrast modulation. Here, the second-order orientation selectivity index, OSI = 0.15 (p < 0.05). (C) First-order orientation-tuned responses of V2 neuron m628r48(V2). The drift direction and spatial frequency of the carrier grating were 202 deg and 1.58 c/deg, respectively. (D) Second-order orientation-tuned responses of m628r48(V2) for modulator spatial frequency = 0.79 c/deg. Here, second-order OSI = 0.09 (p > 0.05). Error bars show standard error of the mean across trials.

Figure 3
Figure 3

Second-order orientation selectivity. (A) Distribution of second-order orientation selectivity index (OSI) computed at the moderate modulator spatial frequency (0.5× carrier grating’s spatial frequency). Using a permutation test, we identified neurons with significant second-order OSI (p < 0.05, black). For V1, the mean and standard deviation of the OSI were 0.18 and 0.14, respectively. For V2, the mean and standard deviation were 0.17 and 0.12, respectively. (B) Distribution of relative preferred second-order orientation (in deg relative to the carrier grating’s orientation) of the four second-order orientations tested. Across the population of neurons, all second-order orientations were represented.

Figure 4
Figure 4

Model of the spatial organization of the excitatory classical receptive field (CRF) and suppressive extraclassical surround. (A) For neuron m628r24(V1), we computed the response to each modulator spatial frequency and drift direction, and pooled responses to identically oriented modulators that drifted in opposite directions (for this neuron, one spatial frequency and eight directions are shown in Fig. 2B). Different colors pertain to different modulator orientations (0, 45, 90, 135 deg relative to the carrier grating’s orientation: orange, gray, purple, green). Shaded regions show standard error of the mean response amplitude across trials. The solid lines show the fitted difference-of-two-spatial-Gaussians model in the frequency domain. These curves capture the organization of the CRF and extraclassical surround. The magnitude of surround suppression is given by the relative response at the optimal modulator spatial frequency (here, approximately 0.4× the carrier grating’s spatial frequency) and zero. The vertical, dashed line indicates the modulator spatial frequency (0.79 c/deg) pertaining to the plot in Fig. 2B. At that modulator spatial frequency, the neuron was second-order orientation selective (orientation selectivity index, OSI = 0.15). (B) We inverse Fourier transformed the model fit shown in (A), revealing the spatial organization of the CRF (red) and the extraclassical surround (blue). This center-surround RF describes an envelope of sensitivity to contrast modulations of the carrier grating, and should not be mistaken for a classical simple cell’s luminance-responsive receptive field profile. The dashed circle, here and in all other figures, indicates the extent of the stimulus, which comprised eight cycles of the carrier grating. The asymmetric arrangement of the field imparts response selectivity for the orientation of the modulator. To aid visualization, the number at the bottom left (0.61) indicates the overall strength of suppression: the spatially integrated surround divided by the spatially integrated CRF. (C) We superimpose the contours shown in (B) on one of the second-order stimuli (Fig. 1) used to stimulate this neuron. This stimulus cyclically evoked a strong spiking response from the neuron (Fig. 2B) because, periodically, the carrier grating engaged the CRF but was completely withdrawn from the most contrast-sensitive region of the suppressive surround, as shown.

Figure 5
Figure 5

Modeled spatial organization of the excitatory classical receptive field (red) and suppressive extraclassical surround (blue) of 25 neurons. Graphical conventions are as in Fig. 4B, including the inset number indicating the strength of the suppressive surround if detected. We rotated these receptive fields (RFs), neuron by neuron, so that the (preferred) carrier grating is shown as vertical and drifting to the right. For each RF, the dashed circle indicates the extent of the stimulus, which comprised eight cycles of the carrier grating. The lightly and heavily dashed circles indicate neurons encountered in V1 and V2, respectively. Here, we have ordered RFs left-to-right, top-to-bottom, by second-order orientation selectivity index (OSI) computed at the moderate modulator spatial frequency (0.5× carrier grating’s spatial frequency). All RFs in the top row have OSI < 0.10. RFs in the bottom row have OSI > 0.4. The coefficient of validation for all RFs shown was > 0.75.

Figure 6
Figure 6

Validation of the modeled center-surround receptive fields (RFs). (A) For neuron m616r17(V1), we used responses to sinusoidal contrast modulations of the carrier grating (Fig. 1) to model the RF, shown inset using the graphical conventions of Fig. 4B. Then, we stimulated this model RF using circular patches of carrier grating and scaled its response (solid line). The shaded area shows the measured response amplitudes to circular patches (standard error of the mean across trials) which we compared to the modeled response. For this neuron, the coefficient of validation was R2 = 0.90. (B) In neuron m616r30(V1), modeling revealed a weak suppressive surround. The coefficient of validation was R2 = 0.93. Graphical conventions are as in A. (C) Distributions of the coefficient of validation for V1 (upper panel) and V2 (lower panel) neurons that were sufficiently responsive and amenable to modeling over the range of modulator spatial frequencies tested. As in (A) and (B), we validated the modeled RFs using responses, scaled and offset as necessary, to circular patches of carrier grating.

Figure 7
Figure 7

Center-surround receptive field (RF) estimation using spike-triggered modulator averaging (STMA). (A) We constructed a stochastic stimulus (first row) by multiplying a carrier grating and a modulator (e.g., Fig. 1). In the second row we show the modulator alone. The carrier grating drifted and the modulator was static but its spatial frequency and orientation were randomized at 6 Hz; spatial frequency was drawn from a uniform distribution on [0, 0.75] cycles/degree relative to the carrier, and orientation was drawn from a uniform distribution on [0, 180) degrees. For each spike, we incorporated a modulator into the average using the delay (orange arrow) that maximized STMA power. (B) For neuron m637r42(V1), we used responses to sinusoidal contrast modulations of the 4 c/deg carrier grating to model the RF. The model revealed how neuron achieved second-order orientation selectivity, responding vigorously to vertical but not horizontal contrast modulations: Via strong surround suppression (blue) on the right flank of the excitatory classical receptive field (CRF) shown in red. Graphical conventions are as in Fig. 6. (C) We estimated m637r42(V1)’s center-surround RF using STMA (upper panel). Note the agreement with the model RF in (B): In the STMA, the surround was most suppressive on the right flank of the CRF. We used an implausible delay (-100 ms) to generate a control STMA (lower panel) which showed no RF structure.

Figure 8
Figure 8

Second-order psychophysical detection and discrimination. Two observers’ second-order thresholds were affected by adaptation in a way that was consistent with a filter-rectify-filter model that is selective for second-order orientation. (A) Subject 1’s threshold for the detection of a 0.5 c/deg, vertical contrast modulation of an oblique, 2 c/deg grating (circles) was elevated by a vertical, second-order adapter, but not by a horizontal, second-order adapter. Nor was this threshold elevated by 0.5 c/deg vertical or horizontal luminance gratings. Thresholds for the detection of a horizontal contrast modulation (squares) showed the corresponding pattern. The gray bands show the 95% confidence interval of the detection threshold for a horizontal, second-order target when the adapter was a uniform, mean-luminance field. The right panel shows that discrimination thresholds (second-order vertical vs. second-order horizontal) were increased by second-order adapters, but remained relatively low for 1st-order adapters. The gray band again shows the 95% confidence interval of the discrimination threshold when the adapter was a uniform, mean-luminance field. (B) Thresholds for subject 2. Graphical conventions are as in (A). Error bars show 95% confidence intervals.

Figure 9
Figure 9

Scatter plot of the maximum radius (2σ) of model excitatory classical receptive field (CRF) and suppressive extraclassical surround, showing the 20 V1 neurons (●, ▼) and 11 V2 neurons (○, ▽) that survived both modeling and validation (validation score > 0.75). The model revealed surround suppression in 25 neurons (circles) and excitation only (i.e., no suppression) in six others (triangles). To convert visual angle (deg) to cortical radius (mm), we assumed a cortical magnification factor of 3 mm/deg, appropriate to our parafoveal eccentricities. Arrows show excitatory and suppressive means, and the gray square indicates neuron m628r24(V1) (see text and Figure 4). On average, the suppressive extraclassical surround was 1.65× wider than the excitatory mechanism largely responsible for the CRF.

Similar articles

Cited by

References

    1. Angelucci A, Levitt JB, Walton EJ, Hupe JM, Bullier J, Lund JS. Circuits for local and global signal integration in primary visual cortex. The Journal of Neuroscience. 2002a;22(19):8633–8646. - PMC - PubMed
    1. Angelucci A, Levitt JB, Lund JS. Anatomical origins of the classical receptive field and modulatory surround field of single neurons in macaque visual cortical area V1. Progress in Brain Research. 2002b;136:373–388. - PubMed
    1. Bair W, Cavanaugh JR, Movshon JA. Time course and time-distance relationships for surround suppression in macaque V1 neurons. Journal of Neuroscience. 2003;23(20):7690–7701. - PMC - PubMed
    1. Braitenberg V, Schüz A. Anatomy of the cortex: Statistics and geometry. Berlin: Springer-Verlag; 1991.
    1. Carandini M, Heeger DJ. Normalization as a canonical neural computation. Nature Reviews Neuroscience. 2013;14(2):51–62. - PMC - PubMed

Publication types

MeSH terms