Shared and unique fMRI responses during naturalistic movie viewing between humans and monkeys
Poster No:
1961
Submission Type:
Abstract Submission
Authors:
Ma Feilong1, Haiyan Wang2, Guo Jiahui1, Xiaolian Li2, Royoung Kim3, Fan Xiaoxuan1, Jane Han1, Qi Zhu4, M. Ida Gobbini5, WON MOK SHIM3, James Haxby1, Wim Vanduffel2
Institutions:
1Dartmouth College, Hanover, NH, USA, 2KU Leuven Medical School, Leuven, Belgium, 3CNIR/SKKU, Suwon-si, South Korea, 4University Paris-Saclay, Gif/Yvette, France, 5University of Bologna, Bologna, Italy
First Author:
Co-Author(s):
Introduction:
Primates, including humans, share similar sensory representations due to their common ancestry. However, differences exist between species that contribute to distinct cognitive abilities. Investigating variations and similarities sheds light on fundamental questions in systems neuroscience regarding brain function, organization, and evolution. Furthermore, understanding the commonality and differences enables us to transfer knowledge obtained from studying one species to another. To establish correspondence between human and monkey brains, we conducted an fMRI study while monkeys and humans watched the same movie, "Monkey Kingdom" and applied novel comparative analyses.
Methods:
We scanned 24 humans and 2 rhesus monkeys using fMRI while they freely watched 5 clips of the movie "Monkey Kingdom" (900s per clip). For the monkey scans (3T) we used contrast-agent enhanced fMRI with an isotropic voxel resolution of 1.25mm and a TR of 1s. The humans were scanned at 3T with a spatial resolution of 2.5mm and a TR of 1s. We convolved human responses with monkey HRF, and monkey responses with human HRF, to account for hemodynamic differences between species and imaging methods.
To derive the fMRI response components shared between humans and monkeys, and those specific to each species, we developed a new method called multivariate variance partitioning. Each time, we trained the model using data from 23 humans and 1 monkey, withholding 1 human and 1 monkey data set for testing. We performed a group-PCA on the training data for each species separately, to derive the PC time series for each species. In this analysis, human data were concatenated across subjects along the vertex dimension. We then used multivariate ridge regression to predict human and monkey PCs based on the respective other species' components. The brain responses explained by the other species were labeled as shared, and the residual brain responses as species-specific. This analysis yielded two sets of shared responses, one from each regression (human-to-monkey, monkey-to-human), which were concatenated. Subsequently, we performed another PCA on the 3 types of responses separately to derive the final shared, human- and monkey-specific PCs, and retained the first 30 PCs each for further analysis.
To derive the fMRI response components shared between humans and monkeys, and those specific to each species, we developed a new method called multivariate variance partitioning. Each time, we trained the model using data from 23 humans and 1 monkey, withholding 1 human and 1 monkey data set for testing. We performed a group-PCA on the training data for each species separately, to derive the PC time series for each species. In this analysis, human data were concatenated across subjects along the vertex dimension. We then used multivariate ridge regression to predict human and monkey PCs based on the respective other species' components. The brain responses explained by the other species were labeled as shared, and the residual brain responses as species-specific. This analysis yielded two sets of shared responses, one from each regression (human-to-monkey, monkey-to-human), which were concatenated. Subsequently, we performed another PCA on the 3 types of responses separately to derive the final shared, human- and monkey-specific PCs, and retained the first 30 PCs each for further analysis.
Results:
In the first analysis, we examined whether brain responses of the test subjects can be explained by each of the 3 groups of components (shared, human- and monkey-specific), and specifically, in which brain regions responses can be explained by shared or species-specific components. The prediction model was trained using the test subject's responses to the first 3 clips of the movie, and evaluated using the responses to the last 2 clips. The performance was evaluated based on Pearson correlation between measured and predicted time series of each cortical vertex, and the correlation maps were averaged across all subjects of the same species (Figure 1).
In the second analysis, we examined temporal dynamics of the 3 component types. For each component type, we computed the response pattern similarity between every pair of time points, yielding a time-point-by-time-point similarity matrix. Human-specific components demonstrated a longer temporal receptive window compared to monkey-specific components (Figure 2).
In the second analysis, we examined temporal dynamics of the 3 component types. For each component type, we computed the response pattern similarity between every pair of time points, yielding a time-point-by-time-point similarity matrix. Human-specific components demonstrated a longer temporal receptive window compared to monkey-specific components (Figure 2).
·Figure 1. Predicting brain responses to the movie based on shared and species-specific components.
·Figure 2. Brain response pattern similarities across time based on shared and species-specific components.
Conclusions:
We used a novel computational method to derive the brain response components shared between humans and monkeys, and components specific to each species. The shared components account for a substantial amount of responses in test subjects from both species. The 2 sets of species-specific components are associated with different brain parts, and exhibit different temporal dynamics. These findings provide an innovative perspective on similarities and differences between humans and monkeys in perceiving complex visual environments, offering insights into the evolutionary trajectory of brain function and neural mechanisms underlying unique human abilities.
Higher Cognitive Functions:
Higher Cognitive Functions Other
Modeling and Analysis Methods:
Multivariate Approaches 1
Perception, Attention and Motor Behavior:
Perception: Auditory/ Vestibular
Perception: Multisensory and Crossmodal 2
Perception: Visual
Keywords:
Other - fMRI; naturalistic viewing; MVPA; temporal receptive window; evolutionary differences
1|2Indicates the priority used for review