Spatial frequency representation in V2 and V4 of macaque monkey

Curation statements for this article:
  • Curated by eLife

    eLife logo

    eLife assessment

    This manuscript makes a major contribution to the study of early visual representation in primates by showing that intermediate cortical areas V2 and V4, as well as primary cortical area V1 (previously shown), contain orthogonal maps of orientation and spatial frequency, which are recursive across the visual field representation. This is a fundamental principle of functional mapping across the two-dimensional cortical surface that ensures and optimizes the complete representation of all combinations across two coding dimensions.

This article has been Reviewed by the following groups

Read the full article

Abstract

Spatial frequency (SF) is an important attribute in the visual scene and is a defining feature of visual processing channels. However, there remain many unsolved questions about how extrastriate areas in primate visual cortex code this fundamental information. Here, using intrinsic signal optical imaging in visual areas of V2 and V4 of macaque monkeys, we quantify the relationship between SF maps and (1) visual topography and (2) color and orientation maps. We find that in orientation regions, low to high SF is mapped orthogonally to orientation; in color regions, which are reported to contain orthogonal axes of color and lightness, low SFs tend to be represented more frequently than high SFs. This supports a population-based SF fluctuation related to the ‘color/orientation’ organizations. We propose a generalized hypercolumn model across cortical areas, comprised of two orthogonal parameters with additional parameters.

Article activity feed

  1. eLife assessment

    This manuscript makes a major contribution to the study of early visual representation in primates by showing that intermediate cortical areas V2 and V4, as well as primary cortical area V1 (previously shown), contain orthogonal maps of orientation and spatial frequency, which are recursive across the visual field representation. This is a fundamental principle of functional mapping across the two-dimensional cortical surface that ensures and optimizes the complete representation of all combinations across two coding dimensions.

  2. Reviewer #1 (Public Review):

    There are five major claims. First is a replication of lower spatial frequency representation in V4. This is based on two examples shown in Fig. 3. The differences look clear but should be analyzed statistically.

    The second claim is that, on the large scale of the visual field representation in V2 and V4, spatial frequency is mapped from high to low going from fovea to periphery, here estimated as lateral to medial, as in V1. The analysis for this is to plot the geometric centroids of 6 different spatial frequency band responses, for orientation contrasts (Fig. 4A) and color contrasts (Fig. 4B), and show that they progress in position from lateral to medial for high to low frequencies. This seems like an unusual analysis that obscures most of the original data concerning the relationship of spatial frequency response profiles to the two-dimensional imaging area. And, the centroids do not show a continuous map, but rather a bunching of points except for the extremes. The additional data are 4 supplemental examples with three, different frequency ranges. These data are not analyzed statistically.

    The third and most important claim is that spatial frequency and orientation are mapped orthogonally (and recursively) in V2 and V4, as seen in Fig. 5 and the Fig. 5 supplement. Together these figures present two regions in V4 and two regions in V2. If these are the only analyzed regions, the authors need to specify more clearly how they were selected. Presumably, though, other regions were analyzed, and the authors should present results from all analyzable regions, and use statistical analyses to establish significance.

    The fourth claim is that color-sensitive regions in V4 are more associated with low spatial frequencies. The one significant example (the analysis and statistical tests need to be explained), shown in Fig. 6, shows a weak relationship to color for both spatial frequency bands, and the other examples presented in the supplementary are not significant and have even lower absolute relationships. These results, if presented, should be considered inconclusive.

    The fifth claim is for stripe-like periodicity of spatial frequency representation in V2, related to color tuning. This is supported by ostention to binary maps of spatial frequency tuning in Fig. 7 and supplement. Establishing this periodicity would require statistical analysis, and in any case, seems impossible since only a sliver of V2 is visible in these brain surface images, so stripes orthogonal to the V1/V2 boundary (i.e. CO stripes) cannot be distinguished from other patterns of spatial frequency tuning. In fact, Fig. 5E and S5I do not appear to have iso-frequency contours biased toward that orientation.

  3. Reviewer #2 (Public Review):

    This paper reports the results of optical imaging experiments on areas V4, V2, and V1 in anesthetised macaque monkeys. The experiments were designed to reveal details of the representation of spatial frequency (SF), orientation (OR), and color in these areas. Evidence for gradients of SF selectivity across the areas is presented. It is also shown that SF and OR maps in V2 and V4 have iso-parameter contours that intersect at right angles, in agreement with, and extending, observations made in V1 and visual areas in cats and other species. Color domains tend to be located in low-SF domains and avoid the higher SF regions. Relationships between V2 stripes and SF preference are also established.

    These findings are a potentially valuable contribution to understanding the maps that exist in V4, which have received less attention than those in areas V1 and V2. However, I have some serious concerns about the validity of the results.

  4. Reviewer #3 (Public Review):

    This is a potentially fundamental study in which the authors used intrinsic signal optical imaging to characterize orientation, spatial frequency (SFs), and color maps in area V2 and V4 of macaques. They show that foveal regions have higher SF preferences and V1 has a preference for higher SFs compared to V2 and V4. They also show that color regions prefer lower SFs. Strikingly, they show that orientation and SFs are mapped orthogonally in V2 and V4. Finally, they show evidence of periodicity in SF preference in V2.

    The data look convincing, but I would like the authors to clarify/discuss certain aspects of the analyses, as detailed below. Overall, I think this study is well done and adds to our understanding of the architecture of the primate visual cortex.

    Major:

    1. I found the proposed hypercolumn architecture in Figure 1B very difficult to understand. SFs vary in a continuum, so why are only two levels (low SF and high SF) shown in two different colors? Iso-orientation and Iso-SF lines could have been shown in different colors also (say HSV colors for orientation to show the circular mapping and gray colormap for SF going from low to high). Similar to what has been done for iso-hue and iso-brightness lines in the color region. Perhaps it may be worthwhile to show the same proposed architecture in V1 as well, in which orientation maps form pinwheels and colors are in separate blobs. It was unclear to me how this architecture could have pinwheels/blobs as well.

    2. It is not clear to me how the details of the functional maps depend on the choice of stimuli. In single unit studies, typically a large number of orientations and SFs are used to independently map the SF and orientation tuning preferences. In contrast, here only 2 orientations are used in one case to map the color space. Even for mapping the orientation space, only 4 orientations are used. For mapping the color space also, only the hues along the red-green axis are varied (L-M pathway). I understand that some of these choices could be due to the recording modality (imaging), but it would be very useful if the authors could discuss how/if these stimulus choices can affect their results. More details of the stimuli, such as the drift rate of the gratings, and the cie (x,y, Y) coordinates of red & green hues would be useful.

    3. Can you show the iso-contour lines for orientation on the orientation maps also as a supplementary figure to see how well the algorithm works? Figure 5A shows iso-orientation lines on the SF map. The iso-SF contours shown in Figure 5B easily correspond to the colors in the SF map shown in 5A, but I had difficulty mapping the orientation. Also, I was wondering whether the way the comparisons are done to get the maps (for example, in Figure 4, the same 4 stimuli are compared in two different ways to get orientation and color maps) can potentially impose some constraints on those maps. I say this because it is striking to me that almost every red and blue line shown in 5C and 5G appears to intersect orthogonally (as also shown in 5D and 5H).

    4. To me the orthogonality of SF and Orientation contours in Figure 5 was the most striking result. Can you show how this analysis looks for V1? The supplementary figure also shows only V2 and V4.

    5. The claim about periodicity is not well quantified. If the authors wish to make this claim, they need to show the Fourier transform of the activation pattern as a function of space and show clear peaks in the spectrum. Also, the authors can perhaps clarify what is the spatial resolution of the imaging technique itself.