Technical Program

Paper Detail

Paper: PS-1B.3
Session: Poster Session 1B
Location: H Fl├Ąche 1.OG
Session Time: Saturday, September 14, 16:30 - 19:30
Presentation Time:Saturday, September 14, 16:30 - 19:30
Presentation: Poster
Publication: 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany
Paper Title: Attention biases neural representations of hierarchical visual features
Manuscript:  Click here to view manuscript
License: Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
Authors: Tomoyasu Horikawa, ATR, Japan; Yukiyasu Kamitani, Kyoto University, Japan
Abstract: Humans can voluntarily regulate how we perceive the external world by top-down mental processes. Visual attention highlights specific features of visual targets and enhances brain activity associated with the attended features. Previous studies demonstrated that the attentional modulation of brain activity allows decoding of attended features from the brain. However, it remains unclear whether and how brain activity associated with multiple levels of visual features can be affected by selective attention. Here, we quantified how hierarchical neural representations are modulated by object-based selective attention using fMRI and the brain decoding technique assisted by deep neural networks (DNNs). Using statistical models that decode fMRI activity into hierarchical DNN features, we decoded fMRI activity measured while subjects attended to one image of a superposition of two images. The decoded features were found to be biased to attended images over unattended images with greater effects for lower-/higher-level DNN features in lower-/higher-level visual areas. Furthermore, image reconstructions generated from the decoded features resembled attended images, demonstrating faithful reconstructions of mental images. Our analyses showed fine-grained attentional modulation for hierarchical visual features. The results also indicate that the attention-guided mental image reconstruction may provide a substrate for developing systems of neurofeedback and brain-machine interfaces.