Feedforward representation of face information in the fusiform face area revealed by VASO 7T layer fMRI

Recognizing faces is a critical cognitive process in social interaction. Face recognition has been extensively studied with functional magnetic resonance imaging (fMRI), ranging from large coverage studies comprehensively examining the entire face pathway to high-resolution laminar-specific (layer) fMRI studies probing feedforward and feedback signals in early visual and face-selective areas during particular face processing. However, it is still unclear which brain region structures face information in the face pathway. To further elucidate this mechanism, we investigated fusiform face area (FFA) during passive face image presentation using a whole-brain vascular space occupancy (VASO) 7T layer fMRI open dataset combined with multi-voxel pattern analysis (MVPA). We analysed lateral occipital cortex (LOC) as a control region and BOLD data in the dataset as a control contrast. We trained multiple binary classification decoders with various image categories (e.g., face vs. others, house vs. others) and three cortical layer groups independently to assess information representation across cortical depths. Our results revealed that decoding accuracies in VASO data peaked in the middle layer of FFA, suggesting a feedforward signature specifically during face processing. This effect was not observed in LOC or in decoders for other image categories. These findings indicate that face information is structured prior to processing in the FFA, consistent with previous reports suggesting that face-related features are partially extracted before reaching higher-level face areas. Overall, this study highlights the potential of combining high-specificity VASO layer-fMRI with high-sensitivity MVPA to dissect cortical information flow.