Study of color quality space at subvoxel fMRI resolution using the repetition-suppression approach
Poster No:
949
Submission Type:
Abstract Submission
Authors:
Ali Moharramipour1, Hakwan Lau1
Institutions:
1Laboratory for Consciousness, CBS, RIKEN, Wako, Saitama
First Author:
Co-Author:
Introduction:
The subjective quality of an experience is defined by its comparison to other experiences. It has been suggested that our brain contains neural representations of the so-called mental quality space in which distances between experiences in this space define how subjectively similar they are [1,2,3]. In the present study, we aimed to find the neural representations of the color quality space that shape our subjective color perceptions. We hypothesize finding the color space in not only the visual areas but also higher brain areas like the lateral prefrontal cortex (LPFC) since it is not only the physical attribute of the colors that form our color quality space but also their conscious perceptual contents.
To this end, we recorded fMRI signals during a color viewing task designed to take advantage of the repetition-suppression (RS). With the right experimental design, RS can become a powerful tool to reveal the representational space of the neural populations within a single fMRI voxel [4], a finer resolution than the conventional fMRI approaches, like multivoxel pattern analysis [5]. It is noteworthy that the RS-fMRI approach has been employed before [6,7], but here, we used a block design to maximize its power. Moreover, we ran a simulation, a forward model from the neuronal to the fMRI level, to verify our RS method, address its limitations, and offer solutions.
To this end, we recorded fMRI signals during a color viewing task designed to take advantage of the repetition-suppression (RS). With the right experimental design, RS can become a powerful tool to reveal the representational space of the neural populations within a single fMRI voxel [4], a finer resolution than the conventional fMRI approaches, like multivoxel pattern analysis [5]. It is noteworthy that the RS-fMRI approach has been employed before [6,7], but here, we used a block design to maximize its power. Moreover, we ran a simulation, a forward model from the neuronal to the fMRI level, to verify our RS method, address its limitations, and offer solutions.
Methods:
12 iso-luminant colors (Fig. 1B) were selected and used as 4 sets of 5 colors in the experiment. Four participants were recruited and scanned extensively (~5 hours) to achieve accurate results at individual levels.
The color space for each participant was derived from a behavioral two-alternative forced choice color similarity judgment task (e.g., Fig. 2A). Fig. 1A shows our RS-fMRI task with stimulus blocks of flashing a single color or two alternating colors. If the representational space of an fMRI voxel matches with the color space, it is expected to observe higher suppression during alternating similar colors than dissimilar colors due to their higher overlapping neural representation. Taking this into account, we set up a general linear model analysis consisting of color-related regressors, a regressor picking up general suppressions, and another suppression-related regressor modulated by the behaviorally derived individual color space. The latter regressor, called dissimilarity modulation regressor, would identify the areas containing the color space.
The color space for each participant was derived from a behavioral two-alternative forced choice color similarity judgment task (e.g., Fig. 2A). Fig. 1A shows our RS-fMRI task with stimulus blocks of flashing a single color or two alternating colors. If the representational space of an fMRI voxel matches with the color space, it is expected to observe higher suppression during alternating similar colors than dissimilar colors due to their higher overlapping neural representation. Taking this into account, we set up a general linear model analysis consisting of color-related regressors, a regressor picking up general suppressions, and another suppression-related regressor modulated by the behaviorally derived individual color space. The latter regressor, called dissimilarity modulation regressor, would identify the areas containing the color space.
Results:
Fig. 2B shows the areas with significant dissimilarity modulations (neural representations with the color space) in one participant. The color space areas were consistently observed in the visual areas of V4 to V1, the posterior part of the parieto-occipital sulcus (POS), the parietal association cortex, and the LPFC. To our surprise, the majority of the color space vertices were located in the POS area (covering some of the peripheral visions and area V6) and then the area V4, which is known to hold color representations [8]. It is noteworthy that we rarely found any vertices with an inverted representation of the color space. Furthermore, the above-mentioned areas better represented the spaces of the sets with close colors (sets 2 and 4) than the sets with far colors (sets 1 and 3), except LPFC, which represented both equally well. This may suggest a higher multi-selectivity in the LPFC vertices than in the other areas. Lastly, our simulation verified our RS method and uncovered possible neuronal formation within a voxel with the color space representation.
Conclusions:
We found the neural representation of the color quality space not only in the visual areas but also in higher brain areas using a novel RS-fMRI approach. This may indicate that conscious perception involves broad brain areas, including the LPFC, rather than being confined to a localized visual region. Moreover, the POS area emerged as the main hub with the color space representation despite often being overlooked in the existing literature on color perception.
Higher Cognitive Functions:
Higher Cognitive Functions Other 1
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI)
Methods Development 2
Perception, Attention and Motor Behavior:
Perception: Visual
Keywords:
Cognition
Consciousness
Design and Analysis
FUNCTIONAL MRI
Perception
Other - Subjective Color Perception, Color Quality space, Repetition-Suppression
1|2Indicates the priority used for review
Provide references using author date format
[2] Rosenthal, D. (2010). How to think about mental qualities. Philosophical Issues, 20, 368-393.
[3] Lau, H. (2022). The mnemonic basis of subjective experience. Nature Reviews Psychology, 1(8), 479-488.
[4] Rigotti, M (2016). Estimating the dimensionality of neural responses with fMRI Repetition Suppression. arXiv preprint arXiv:1605.03952.
[5] Lewis-Peacock, J. A. (2014). Multi-voxel pattern analysis of fMRI data. The cognitive neurosciences, 512, 911-920.
[6] Barron, H. C. (2016). Repetition suppression: a means to index neural representations using BOLD?. Philosophical Transactions of the Royal Society B: Biological Sciences, 371(1705), 20150355.
[7] Bhandari, A. (2019). Measuring prefrontal representational geometry: fMRI adaptation vs pattern analysis. In 2019 Conference on Cognitive Computational Neuroscience (pp. 2019-1162). CCN.
[8] Brouwer, G. J. (2009). Decoding and reconstructing color from responses in human visual cortex. Journal of Neuroscience, 29(44), 13992-14003.