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



Sensorimotor Control

開催日 2014/9/12
時間 11:00 - 12:00
会場 Poster / Exhibition(Event Hall B)

Sparse Linear Regressionによる下行性神経活動からの昆虫歩行運動の推定
Reconstruction of insect locomotion from descending neural activities using sparse linear regression

  • P2-117
  • 小川 宏人 / Hiroto Ogawa:1 首藤 智宏 / Tomohiro Shudo:2 染谷 真琴 / Makoto Someya:2 春野 雅彦 / Masahiko Haruno:3 
  • 1:北海道大院・理 ・生物科学 / Dept Bio Sci, Fac Sci, Hokkaido Univ, Hokkaido, Japan 2:北海道大院・生命 ・生命システム / Biosystem Sci, Grad Sch Life Sci, Hokkaido Univ, Hokkaido, Japan 3:NICT脳情報通信融合研究センター / Center for Information and Neural Network, NICT,Osaka, Japan 

Descending neural signals is essential for motor control of speed and direction of locomotion in behaving animals. Crickets perform oriented walking behavior in response to air-current stimuli. Recently, we found that descending signals from cephalic ganglia are required for stimulus angle-dependent control of walking direction and turn angle in the wind-elicited walking behavior (Oe & Ogawa, 2013). However, it has been unknown how and which descending neurons contribute to encode information for the locomotion control. To reveal the relationship between population activity of the descending neurons and locomotion parameters such as direction and speed, we simultaneously recorded descending spikes from the cricket walking on a spherical treadmill and its walking velocity and turn angle. Further, we tried to decode the locomotion parameters from the ensemble activity of descending neurons using sparse linear regression (SLiR) algorithm. Comparing in the decoding performance between voluntary and wind-elicited walking behaviors, the SLiR model constructed from spike activity during voluntary walking predicted the walking velocity in both voluntary and wind-elicited walking with high accuracy. However, the model constructed from wind-elicited walking data set provided high accuracy for prediction of walking velocity in the wind-elicited walking, but did not predict that in the voluntary walking. This fact means that the SLiR model constructed from voluntary walking data set has higher generalization than the model for the wind-elicited walking. That is, the voluntary walking contains a variety of locomotion patterns while the wind-elicited walking is a stereotypical locomotion. The model for predicting voluntary walking velocity selected most of the descending units with a light weight. In contrast, the model for predicting wind-elicited walking velocity heavily weighted fewer units. From these results, it is supposed that larger number of descending neurons are involved in encoding walking velocity during voluntary walking comparing to wind-elicited walking.

Copyright © Neuroscience2014. All Right Reserved.