Neural Coding of Multiple Stimulus Features in Auditory Cortex

Size: px

Start display at page:

Download "Neural Coding of Multiple Stimulus Features in Auditory Cortex"

Gregory Chase
5 years ago
Views:

1 Neural Coding of Multiple Stimulus Features in Auditory Cortex Jonathan Z. Simon Neuroscience and Cognitive Sciences Biology / Electrical & Computer Engineering University of Maryland, College Park

Computational Sensorimotor Systems Laboratory Students Juanjuan Xiang Maria Chait

Moss Huan Luo Ling Ma Claudia Bonin Murat Aytekin Raul Esteban Alain de Cheveigné

Shantanu Supported by NIH (NIDCD/NIBIB/NIA) 1R03DC004382, 1R01EB004750,

2 Computational Sensorimotor Systems Laboratory Students Juanjuan Xiang Maria Chait Nayef Ahmar Minsuk Park Victor Grau-Serrat Faculty David Poeppel Shihab Shamma Cindy Moss Huan Luo Ling Ma Claudia Bonin Murat Aytekin Raul Esteban Alain de Cheveigné Catherine Carr Postdocs Yadong Wang Mounya Elhilali Lab Staff Jeff Walker Ray Shantanu Supported by NIH (NIDCD/NIBIB/NIA) 1R03DC004382, 1R01EB004750, 1R01AG027573, 1R01DC007657, 1F31NS Luo, H., Y. Wang, D. Poeppel & J. Z. Simon (2006), J. Neurophysiology

3 Outline Magnetoencephalography (MEG) as a tool of Non-Invasive Auditory Physiology MEG in the Frequency Domain Neural Encoding of Modulations

4 Magnetoencephalography (MEG) Non-invasive, Passive, Silent Neural Recordings Simultaneous Whole-Head Recording (~200 sensors) Sensitivity high: ~100 ft (10 13 Tesla) low: ~10 4 ~10 6 neurons Temporal Resolution: ~1 ms Spatial Resolution coarse: ~1 cm ambiguous

Functional Imaging Hemodynamic techniques

(Functional brain imaging) Electromagnetic

MEG Excellent spatial resolution ( ~ 1-2 mm) Poor

across-subject averaging fmri and MEG can capture

5 Functional Imaging Hemodynamic techniques Non-invasive recording from human brain (Functional brain imaging) Electromagnetic techniques Functional magnetic resonance imaging fmri Positron emission tomography PET Electroencephalography EEG Magnetoencephalography MEG Excellent spatial resolution ( ~ 1-2 mm) Poor temporal resolution ( ~ 1 s) PET, EEG require across-subject averaging fmri and MEG can capture effects in single subjects Poor spatial resolution ( ~ 1 cm) Excellent temporal resolution ( ~ 1 ms)

6 Primary Neural Current Photo by Fritz Goro

7 MEG Measures Neural Currents SQUID Gradiometer Sink Source Magnetic Dipolar Field (Projection) Direct electrophysiological measurement not hemodynamic real-time No unique solution for distributed source

8 MEG Response Flattened Isofield Contour Map Sink Source t = 98 ms 40 ft/step Instantaneous Magnetic Field

9 MEG Response 3-D Isofield Contour Map Chait, Poeppel and Simon, Cerebral Cortex 2006

10 MEG Response Spatial Map of Time Series 500 (ft) 500 (ms)

11 MEG Response Butterfly Plot 400 t = 98 ms Time (ms)

12 Time Course of MEG Responses Evoked Responses MEG Events Time-Locked to Stimulus Event Pure Tone Broadband Noise

13 Spatial Auditory MEG Responses Auditory Responses Robust Strongly Lateralized M50 Change Onset SSR ICA

14 MEG as Auditory Physiology Tool Advantages of humans over animals Subjects can be rented (by the hour) Subjects can be trained in minutes Better grasp of subjects perceptual space (?) Access to Speech & Language processing (?) Advantage of Whole Head Recording Disadvantage of Neural Source Localization Coarseness/Ambiguity in Source Location Blindness to Many Kinds of Coding Neutral Aspects Neural Source is Dendritic Current (not Spikes) Humans not typical mammals (?) New Technique/Immature Analysis Tools

15 Outline Magnetoencephalography (MEG) as a tool of Non-Invasive Auditory Physiology MEG in the Frequency Domain Neural Encoding of Modulations

16 An Alternative to Time: Frequency Use Stimuli localized in Frequency rather than time Examine Response at Same Frequency Steady State Response (SSR) Frequency Response/Transfer Function

17 Frequency Response 32 Hz Modulation 400 Hz tone carrier s (concatenated) Single MEG Channel Precise Phase-Locking: 0.01 Hz Little trial-to-trial jitter Amplitude + Phase...

18 Whole Head Steady State Response 32 Hz

19 Complex Magnetic Field

V = V + j V Re Im V(θ) = V cos(θ) + V sin(θ) Re Im V Re V Max θ Max 90 90 V Im V Min

20 V = V + j V Re Im V(θ) = V cos(θ) + V sin(θ) Re Im V Re V Max θ Max V Im V Min Intensity: V Max Phase: θ Max Sharpness: η = V Min /V Max Simon and Wang, J. Neurosci. Methods 2005

21 Outline Magnetoencephalography (MEG) as a tool of Non-Invasive Auditory Physiology MEG in the Frequency Domain Neural Encoding of Modulations

22 Modulation Encoding Simple Modulations Simple Cortical Encoding Simple Amplitude Modulation coding often used for slower modulations Rate coding (invisible to MEG) often used for faster modulations Applies to general modulations: AM, FM, other Simple Amplitude Modulation coding is easily detectable in Fourier/Spectral domain (SSR) = Spectral Peak at Modulation Frequency Coding for multiple modulations of different kinds?

23 Sample Dual Modulation Stimuli f FM = 3.1 Hz f AM = 37 Hz f FM = 8 Hz f AM = 37 Hz time (s) time (s)

24 SSR Carrier Dependence SSR Amplitude SSR Phase Ross et al. (2000) SSR Phase Follows Carrier Phase Modulation Encoding Patel & Balaban (2004) 0.08 Hz

25 Neural Modulation Models Stimulus Carrier Frequency (FM) Stimulus Envelope Amplitude (AM) Amplitude Modulation (AM) Coding Neural Modulation Coding Phase Modulation (PM) coding time Neural Averaged Response Lower Sideband SSR Frequency Upper Sideband Neural Response Spectrum c.f. Patel & Balaban (2004) α AM = 0 or 2π α = (ϕ upper ϕ SSR ) (ϕ SSR ϕ lower ) Neural Response Phase Encoding Parameter α PM = π

26 Spectral Sideband Responses 37 Hz 37.3 Hz upper sidebands 37.5 Hz 37.8 Hz 0.3 Hz 0.5 Hz 0.8 Hz Frequency (Hz) Frequency (Hz) Frequency (Hz) 37 Hz 38.0 Hz 38.7 Hz 1.0 Hz 39.1 Hz 1.7 Hz 2.1 Hz Frequency (Hz) Frequency (Hz) Frequency (Hz) 40.0 Hz 42.0 Hz 45.0 Hz 3.0 Hz 5.0 Hz 8.0 Hz 37 Hz Frequency (Hz) Frequency (Hz) Frequency (Hz)

27 Sideband Responses Normalized Upper Sideband Responses Chance Response Level Only Upper Sideband contributes Lower Sideband Responses Chance Response Level f FM (Hz) Modulation Encoding, with coding transition at f FM ~ 5 Hz

28 Modulation Encoding Type 2π π 0 f FM (Hz) Phase Modulation Encoding below f FM ~ 5 Hz

29 Neural Population Model S( t) = ( 1 + mcos( 2 f t + )) cos(2 f t + cos( 2 f t )) 8 + GWN

Model Results Experimental Results Model Results θ θ 2π π adj Upperdiff

4 0.8 0 0.4 0.8 α π π π 1 2 π AI 0 π -1 0.3 0.5 0.8 1.0 1.7 2.1 3.1 5.1 8.

1 Stimulus condition f (Hz) AM Modulation index m AM modulation index m

30 Model Results Experimental Results Model Results θ θ 2π π adj Upperdiff adj Lowerdiff 2π π adj Upperdiff adj Lowerdiff θ 2 π π θ 2π 2π 2 π α π π π 1 2 π AI 0 π Stimulus condition f (Hz) AM Modulation index m AM modulation index m S(t)= (1 + mcos(2π f t +θ)) cos(2π f t + π FM AM 8 cos(2π f FM t )) + GWN Amplitude Modulation AM in quadrature with PM Why? Phase Modulation

31 Summary Combined AM/FM modulations are encoded in Auditory Cortex Phase Modulation seen at lowest FM rates Modulation Encoding changes at higher rates Single Sideband Modulation unexpected Speculate: Single Modulation Encoding type? Or: Two populations of AM and PM encoding neurons whose phase happens to cancel in lower sideband? Magnetoencephalography (MEG) Directly generated by neural currents Excellent time/frequency resolution Spatial Localizability an open question

32 Thank You

Magnetoencephalography and Auditory Neural Representations

Magnetoencephalography and Auditory Neural Representations Jonathan Z. Simon Nai Ding Electrical & Computer Engineering, University of Maryland, College Park SBEC 2010 Non-invasive, Passive, Silent Neural