logo

SCIENTIA SINICA Informationis, Volume 49, Issue 11: 1517-1527(2019) https://doi.org/10.1360/N112018-00180

Time-delay-based neural decoding using LFP signals in the primary motor cortex of mice

More info
  • ReceivedJul 23, 2018
  • AcceptedApr 15, 2019
  • PublishedNov 12, 2019

Abstract

In the research of brain-computer interfaces, the neural electrophysiological signals are commonly used for neural information decoding. However, due to the difficulty in recording the neural electrophysiological signals of mice, especially spike, a brain-computer interface is seldom used. In this study, the local field potential (LFP) signals of the primary motor cortex of mice are recorded, and the power spectral density of the LFP is calculated as the input features. The SVM classification algorithm is used to decode the neural information of mice during the lever-pressing movement. The obtained results illustrate that the predicted movement has an early time between the real movement. To this end, a time-delay SVM model is built to decode the binary motion signal for the only four-channel LFP signal of the mice. The result shows that the time-delay SVM decoding achieved high-accuracy performance.


Funded by

国家十三五重点研发计划项目(2017YFC1308501,2017YFGH001560)

国家自然科学基金重大仪器项目(31627802)

国家自然科学基金面上项目(31371001)


Acknowledgment

感谢浙江大学生物医学工程与仪器科学学院的沈义民老师及凌隽成同学在动物实验和论文写作中给予的指导和帮助.


References

[1] Lebedev M A, Nicolelis M A L. Brain-machine interfaces: past, present and future.. Trends Neurosciences, 2006, 29: 536-546 CrossRef PubMed Google Scholar

[2] Serruya M D, Hatsopoulos N G, Paninski L. Instant neural control of a movement signal. Nature, 2002, 416: 141-142 CrossRef PubMed ADS Google Scholar

[3] Hochberg L R, Serruya M D, Friehs G M. Neuronal ensemble control of prosthetic devices by a human with tetraplegia. Nature, 2006, 442: 164-171 CrossRef PubMed ADS Google Scholar

[4] Leeb R, Friedman D, Müller-Putz G R. Self-Paced (Asynchronous) BCI Control of a Wheelchair in Virtual Environments: A Case Study with a Tetraplegic. Comput Intelligence Neuroscience, 2007, 2007(2): 1-8 CrossRef Google Scholar

[5] Aggarwal V, Mollazadeh M, Davidson A G. State-based decoding of hand and finger kinematics using neuronal ensemble and LFP activity during dexterous reach-to-grasp movements.. J NeuroPhysiol, 2013, 109: 3067-3081 CrossRef PubMed Google Scholar

[6] Chapin J K, Moxon K A, Markowitz R S. Real-time control of a robot arm using simultaneously recorded neurons in the motor cortex.. Nat Neurosci, 1999, 2: 664-670 CrossRef PubMed Google Scholar

[7] Velliste M, Perel S, Spalding M C. Cortical control of a prosthetic arm for self-feeding. Nature, 2008, 453: 1098-1101 CrossRef PubMed ADS Google Scholar

[8] Jun Zhuang , Truccolo W, Vargas-Irwin C. Decoding 3-D reach and grasp kinematics from high-frequency local field potentials in primate primary motor cortex.. IEEE Trans Biomed Eng, 2010, 57: 1774-1784 CrossRef PubMed Google Scholar

[9] Khorasani A, Heydari Beni N, Shalchyan V. Continuous Force Decoding from Local Field Potentials of the Primary Motor Cortex in Freely Moving Rats. Sci Rep, 2016, 6: 35238 CrossRef PubMed ADS Google Scholar

[10] Jiang B. Neural signal analysis and decoding of primary motor cortex in rats. Dissertation for Master Degree. Hangzhou: Zhejiang University, 2012. Google Scholar

[11] Scheid M R, Flint R D, Wright Z A, et al. Long-term, stable behavior of local field potentials during brain machine interface use. In: Proceedings of the 35th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2013. Google Scholar

[12] Buzsáki G, Stark E, Berényi A. Tools for probing local circuits: high-density silicon probes combined with optogenetics.. Neuron, 2015, 86: 92-105 CrossRef PubMed Google Scholar

[13] Yang W, Yuste R. In vivo imaging of neural activity.. Nat Methods, 2017, 14: 349-359 CrossRef PubMed Google Scholar

[14] Sheintuch L, Rubin A, Brande-Eilat N. Tracking the same neurons across multiple days in Ca$^{2+}$ imaging data. Cell Rep, 2017, 21: 1102-1115 CrossRef PubMed Google Scholar

[15] Cortes C, Vapnik V. Support-vector networks. Mach Learn, 1995, 20: 273-297 CrossRef Google Scholar

[16] Paxinos G, Franklin K B J. Paxinos and Franklin's the Mouse Brain in Stereotaxic Coordinates. Amsterdam: Elsevier/AP, 2013. Google Scholar

[17] Ludwig K A, Miriani R M, Langhals N B. Using a common average reference to improve cortical neuron recordings from microelectrode arrays.. J NeuroPhysiol, 2009, 101: 1679-1689 CrossRef PubMed Google Scholar

  • Figure 1

    Integral flow chart of mouse lever pressing behavior. (a) The metal lever; (b) the pump; (c) the pressure sensor, an infrared tube; (d) the motor, drive the stretching of the lever; (e) the micro-controller, as a significant part of the lever-reward system, controlled the transmission of data and order; (f) the neural signal amplifier is connected to the headstage on the head of mouse, to amplify the neural signal; (g) Cerebus multichannel data acquisition system; (h) PC

  • Figure 2

    Neural and motor signals for the two lever-pressing. (a) The original LFP signal; (b) the LFP signal after CAR filter and least squares smoothing; (c) the sound cue for the beginning of each lever-pressing; (d) the motor signal of the lever-pressing. When the sound cue starts, the mice begin to press the lever, and when the lever is pressed down, the motor signal level increases

  • Figure 3

    (Color online) Power spectrum analysis of four channels neural signal of mouse NO. 3. (a)–(d) represent four channels respectively. The black line represents the timestamp of the onset of the lever-pressing. The power spectrum between 0–300 Hz frequency bands superimpose the neural signal of 78 successful lever-pressing trials in a session

  • Figure 4

    The result of four-channel multi-band neural signals decoding by using the normal SVM model. There are 11 lever-pressing in total, the dotted line presents the real movement, the solid line presents the predicted voltage. There are four kinds of decoding results: (1) decoding right, (2) decoding time difference, (3) decoding result jitter, (4) decoding error

  • Figure 5

    The performance of decoding under different delay time

  • Figure 6

    The decoding model was evaluated by comparing (a) the average recall rate, (b) the average accuracy rate and (c) the average F1Score of the 7 training periods

  • Figure 7

    The result of four-channel multi-band neural signals decoding by using time-delay SVM model. There are 11 lever-pressing in total, the dotted line represents the real movement, and the solid line represents the predicted voltage after the time-delay SVM decoding. The decoding results are basically consistent with the actual movement, but there is still a small part of decoding jitter

Copyright 2020 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有

京ICP备18024590号-1