• Top page
  • Timetable
  • Per session
  • Per presentation
  • How to
  • Meeting Planner

演題詳細

Poster

視覚
Visual System

開催日 2014/9/13
時間 14:00 - 15:00
会場 Poster / Exhibition(Event Hall B)

サル下側頭葉皮質から記録した皮質脳波を用いた素材カテゴリーのデコーディング
Decoding of material categories from ECoG signals in the macaque inferior temporal cortex

  • P3-124
  • 寺本 傑 / Takashi Teramoto:1 川嵜 圭祐 / Keisuke Kawasaki:1 岡谷 貴之 / Takayuki Okatani:2 澤畑 博人 / Hirohito Sawahata:3 鈴木 隆文 / Takafumi Suzuki:4 長谷川 功 / Isao Hasegawa:1 
  • 1:新潟大学大学院 医歯学総合研究科 統合生理学分野 / Dept Physiol, Niigata Univ Med Sch , Niigata, Japan 2:東北大学大学院情報科学研究科・システム情報科学専攻 / Tohoku Univ Grad Sch of Info Sci,Miyagi,Japan 3:豊橋技術科学大学電気・電子情報工学系 / Dept of Electrical and Electronic Inf Eng Toyohashi Univ of Tech,Aichi,Japan 4:情報通信研究機構・脳情報通信融合研究センター / NICT Cinet ,Osaka,Japan 

The visual categorization of materials is an important inference for appropriate interactions with objects. Previous studies using fMRI suggested that neural substrates for material perception and categorization were present in the ventral visual areas in both monkeys and humans. To specify the higher spatial-temporal aspects of this process,we conducted electrocorticographic recording using 128 channels with 2.5-mm electrode spacing in the inferior temporal cortex. Thirty images from six material categories (foliage,glass,leather,metal,plastic,and wood) were presented while the monkey performed a visual fixation task. As a comparison, we also presented 30 object images from six object categories (monkey face,human face,animal face,fruits,buildings,and tools). Using whole 128-channel visually evoked potentials during stimulus presentation as a feature vector, we constructed linear support vector machine classifiers to predict the six material categories. The decoding performance for the six material categories was 26%, which was well above the performance expected by chance (17%) but lower than that for the six object categories (55%). By systematically varying the latencies and duration of evoked potentials as feature vectors, we found that the decoding performance for material categories rose in shorter latency (50ms) and reached to the max in shorter duration (140ms) compared with those for object categories (75ms in latency and 180ms in duration). Using signals only from the subdivision of electrodes quartering along the anterior-posterior axis, we found that the signals from the most posterior subdivision were exclusively critical for material categorization, whereas those from the second most anterior portion were most prepotent. The signals from other portions were also contributive to object categorization. These results suggest a fast material categorization process operated in the posterior region of the inferior temporal cortex, and this process may be distinct from object categorization in the middle and anterior inferior temporal cortices.

Copyright © Neuroscience2014. All Right Reserved.