Presentation 2007-03-14
Reconstruction of visual images by combining multi-resolution local image decoders
Hajime UCHIDA, Yoichi MIYAWAKI, Okito YAMASHITA, Masa-aki SATO, Hiroki C. TANABE, Norihiro SADATO, Yukiyasu KAMITANI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Recent studies have shown that human brain activity measured by functional magnetic resonance imaging (fMRI) can be decoded to predict visual perceptual parameters such as orientation and motion direction. In this study, we present methods to reconstruct arbitrary visual images from fMRI signals. The image decoder was constructed by combining local image decoders that were trained to predict the mean contrast of local image segments of multiple scales from fMRI activity patterns. We examined three methods for combining local image decoders 1) pixel representation, 2) multi-scale representation, 3) bayes estimation based on generative model for fMRI signals. These methods showed high reconstruction accuracy for arbitrary visual images. Our approach for reconstructing visual images provides a unique tool to study detailed representaion and processing in the visual cortex.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Image reconstruction / Decoding / Visual cortex / fMRI
Paper # NC2006-131
Date of Issue

Conference Information
Committee NC
Conference Date 2007/3/7(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Reconstruction of visual images by combining multi-resolution local image decoders
Sub Title (in English)
Keyword(1) Image reconstruction
Keyword(2) Decoding
Keyword(3) Visual cortex
Keyword(4) fMRI
1st Author's Name Hajime UCHIDA
1st Author's Affiliation Nara Institute of Science and Technology:ATR Computational Neuroscience Laboratories()
2nd Author's Name Yoichi MIYAWAKI
2nd Author's Affiliation ATR Computational Neuroscience Laboratories:ATR Computational Neuroscience Laboratories
3rd Author's Name Okito YAMASHITA
3rd Author's Affiliation ATR Computational Neuroscience Laboratories
4th Author's Name Masa-aki SATO
4th Author's Affiliation ATR Computational Neuroscience Laboratories
5th Author's Name Hiroki C. TANABE
5th Author's Affiliation National Institute of Information and Communications Technology
6th Author's Name Norihiro SADATO
6th Author's Affiliation National Institute of Information and Communications Technology
7th Author's Name Yukiyasu KAMITANI
7th Author's Affiliation ATR Computational Neuroscience Laboratories:Nara Institute of Science and Technology
Date 2007-03-14
Paper # NC2006-131
Volume (vol) vol.106
Number (no) 588
Page pp.pp.-
#Pages 6
Date of Issue