artifact removal for in-vivo neural signal

Upload: kafiiut

Post on 02-Jun-2018

234 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    1/46

    Artifact Detection and Removal

    for In-Vivo Neural SignalPresented By:

    Md Kafiul IslamA0080155M

    Supervisor: Dr. Zhi Yang

    Department of Electrical and Computer Engineering

    National University of Singapore

    20thFeb, 2013

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    2/46

    Outline

    IntroductionMotivation

    Artifact Characterizing: Its Types, Sources and Properties;

    Dynamic range analysis

    Literature Review

    Proposed Method

    Wavelet Transform Based Artifact Removal

    Simulation Results

    Comparison with Other Methods

    Future Work

    Real-time Implementation of Proposed Algorithm

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    3/46

    Motivation

    Closed Loop Neural System for BCI or neural prosthetic Application

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    4/46

    Motivation

    Presented By Md Kafiul Islam

    Typical In-Vivo Neural Signal Processing Steps

    Remove50/60HzNoise

    LPF @200Hz

    OffsetRemove

    Smoothing AlignmentUnder

    SamplingLFP

    Data

    Notch FilterField Potentials Recording

    Artifact Detect &Remove

    HPF @1Hz

    BPF(300Hz~5kHz)

    SmoothingNormalization& Alignment

    SpikeData

    Spike Detection

    Spike Detection & Recording

    FeatureExtraction

    Classification

    Compression

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    5/46

    MotivationIn-vivo neural recording

    Investigate brain information processing & data storage

    Better Spatio-temporal resolution.

    Better Signal-to-Noise Ratio(SNR).

    The study of both LFP& Spikesalong with their correlation: more insight on

    how brain works.

    Artifacts

    Recordings corrupted by artifacts, especially in less constrained

    environment.

    Cause mistakesin interpretationof neural information.

    Signal Preprocessing: Automatic Detection and Removal of Artifacts

    The challenges for in-vivo artifact identification compare to EEG/MEG-

    artifacts are:

    No prior knowledge about artifacts unlike EEG-artifacts

    The broad frequency band of in-vivo data (0.1 Hz 5 kHz) makes it difficult

    to separate artifacts from signal

    Presented By Md Kafiul Islam

    Single-multi unit

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    6/46

    What is Artifact!?

    In neural recording, artifacts are interfering signals that originatefrom some source other than the brain of interest.

    In Vivo Recording of Spontaneous

    Neural Activity of a freely moving rat

    Offending artifacts may obscure, distort, or completely

    misrepresent the true underlying recorded signal being observed.

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    7/46

    What is Artifact (Cont)

    Local :localized in space, i.e. appear only in a single recording channel.

    Global :across all the channels of an electrode at the same temporal window. Irregular:only once in the whole recording sequence

    Periodic:regular manner possibly due to some periodic motions of the subject.

    Presented By Md Kafiul Islam

    0 2 4 6 8 10-10

    -5

    0

    5

    0 2 4 6 8 10-10

    -5

    0

    5

    0 2 4 6 8 10-10

    -5

    0

    5

    SignalAmplitude,mV

    0 2 4 6 8 10-5

    0

    5

    0 2 4 6 8 10-5

    0

    5

    0 2 4 6 8 10-4

    -2

    0

    2

    0 2 4 6 8 10-4

    -2

    0

    2

    Time, Sec

    0 2 4 6 8 10-2

    0

    2

    ch 1 ch 2

    ch 4

    ch 6

    ch 3

    ch 5

    ch 7 ch 8

    Global Artifacts

    Irregular/Local Artifacts Periodic Artifacts

    Perspective Artifact Category/Class

    Repeatability Irregular/No Periodic/Regular/Yes

    Origin Internal External

    Appearance Local Global

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    8/46

    Possible Artifact Sources

    Artifacts may generate from 3 general factors :

    i) Environmental factors (e.g. power noise, sound/optical interference, EM-coupling

    from earth: 7.82 Hz and harmonics*, etc.)

    ii) Experiment factors (e.g. electrode position altering, connecting wire

    movement, etc. due to mainly subject motion )

    iii) Physiological factors (e.g. EOG, ECG, EMG, BCG,etc.)

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    9/46

    Artifact Types

    Presented By Md Kafiul Islam

    (Identified by Empirical Observations Based on Real Neural Sequence, there could be

    many other types as well)

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    10/46

    Properties of Artifacts

    Usually the artifacts have very large

    magnitude compared to the neural data

    of interest, i.e. spike and local field

    potential.

    The frequency range for artifact may

    vary from very low(motion artifact) to

    high frequency(artifact due to residue

    charge) range.

    Presented By Md Kafiul Islam

    Local Field Potential => 0.1 Hz ~ 200 Hz, 0.1 ~ 1 mVppNeural Spikes => 300 Hz ~ 5 kHz, 40 ~ 500 uVppArtifacts => 0 ~ 10 kHz or even higher, max amplitude as high as

    20 mVpp. (From real data observation)

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    11/46

    Properties of Artifacts(Comparison in Spectral Domain with Neural Signal of Interest)

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    12/46

    Problems with Artifacts

    Can cause electronics saturation [1]

    High dynamic range required (Higher ENOB in ADC) [2]

    Mislead to spike detection (high freq) [3]

    Misinterpretation for LFP recording(low freq) [4]

    Presented By Md Kafiul Islam

    260 265 270 275 280 285 290 295

    -15

    -10

    -5

    0

    5

    x 10-4

    Time, Second

    Voltage,

    V

    olt

    260 265 270 275 280 285 290 295 300-2.5

    -2

    -1.5

    -1

    -0.5

    0

    0.5

    1x 10

    4

    Time, Second

    Voltage,

    Vo

    lt

    Afte r BPF of In Vivo data fro m 300 Hz t o 5 kHz

    False Spike detection

    0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

    x 105

    -2

    -1.5

    -1

    -0.5

    0

    0.5

    1

    1.5

    2x 10

    -3

    Time Sample

    Voltage,

    V

    9.06 9.08 9.1 9.12 9.14 9.16 9.18 9.2

    x 104

    -15

    -10

    -5

    0

    5

    x 10-5

    Time, Second

    Voltage,

    Vo

    lt

    Local Field Potential

    [1]

    [2] [3][4]

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    13/46

    Dynamic Range Study

    Presented By Md Kafiul Islam

    Subject

    (Fs in kHz)

    B.W.

    No of Data

    Sequences

    (Data Length

    in min)

    Amplifier Circuit

    Noise Floor

    (Vrms)

    DR without

    Artifact

    (Mean SD)

    (Full Spectrum Datain dB)

    DR with

    Artifact

    (Mean SD)

    (Full Spectrum Data

    in dB)

    Increase in DR

    (Full Spectrum

    Data in dB)

    DR without

    Artifact

    (Mean SD)

    (Spike Data indB)

    DR with

    Artifact

    (Mean SD)

    (Spike Data in

    dB)

    Increase in

    DR

    (Spike Data

    in dB)

    Rat

    Hippocampus

    (40)

    0.1 Hz 10 kHz

    134

    (15) 1 69.01 2.10 82.44 4.21 13.43 59.21 4.32 78.35 8.26 19.14

    Human Epilepsy

    (32.5)

    0.5 Hz 9 kHz

    64

    (18) 1 34.45 3.42 64.36 3.42 29.90 28.82 4.605 55.75 6.94 26.92

    0 5 10 15

    40

    45

    50

    55

    60

    65

    70

    75

    80

    85

    90

    Artifact Amplitude, mV

    DynamicRange,

    dB

    Full Spectrum Data with T2 art

    Spike Data with T2 art

    Full Spectrum Data with T1 art

    Spike Data with T1 art

    Full Spectrum Data with T3 art

    Spike Data with T3 art

    Full Spectrum DRWithout Artifact

    Spike DRWithout Artifact

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    14/46

    LiteratureReview(No literature particularly on artifacts for in-vivo neural signals)

    EEG artifacts removal:

    ICA, CCA (offline and manual intervention, at best semi-automatic,

    works for global artifacts only)

    Adaptive filtering (Reference channel to record artifact)

    Wavelet-enhanced ICA/CCA (Identification of artifactual

    Component is a tough job, DWT involved)

    HHT, e.g. EMD or EEMD (Computational complexity and storage

    problem)

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    15/46

    LiteratureReview Limitations of Current Methods

    Assumes the sources are independentand at most onesource can be Gaussiandistributed (ICA)

    Assumes the sources are maximally un-correlated(CCA)

    Requires extrareferencechannel to record artifacts(Adaptive filter)

    Assumes signals and artifacts to be stationarylinear

    random process and known spectral char(Wiener filter)

    Filter model to be linear, a-prioriestimation is Gaussian

    and work only for unimodaldistribution (Kalman filter)

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    16/46

    Comparison of Current Artifact Removal Techniques

    for Physiological Signals(EEG, EMG, ECG, ECoG, fNIRS, PPG, Respiration, etc.)

    Presented By Md Kafiul Islam

    Adopted from: Kevin T. Sweeney, Tomas E. Ward, and Sean F. McLoone, Artifact Removal in Physiological SignalsPractices and

    Possibilities, IEEE Transactions on Information Technology In Biomedicine, vol. 16, no. 3, May 2012.

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    17/46

    Comparison of Current Artifact Removal Techniques

    (Computational Time)

    Adopted from: Kevin T. Sweeney, Sean F. McLoone, and Tomas E. Ward The use of Ensemble Empirical Mode Decomposition

    with Canonical Correlation Analysis as a Novel Artifact Removal Technique, IEEE Transactions on Biomed Eng., 2012.

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    18/46

    About Wavelet Transform(A Multi-resolution Analysis)

    Split Up the Signal into a Bunch of Signals

    Representing the Same Signal, but all Corresponding to Different Frequency Bands

    Only Providing What Frequency Bands Exists at What Time Intervals

    Presented By Md Kafiul Islam

    ( ) ( ) ( ) dts

    ttx

    sss

    xx

    == *1

    ,,CWT

    Translation

    (The location of

    the window)

    ScaleMother Wavelet

    From http://www.cerm.unifi.it/EUcourse2001/Gunther_lecturenotes.pdf, p.10

    Wavelet

    Small wave

    Means the window function is of finite length

    Mother Wavelet

    A prototype for generating the other window functions

    All the used windows are its dilated or compressed and shifted

    versions

    Scale S>1: dilate the signal

    S

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    19/46

    Why Wavelet Transform:

    Artifacts:

    Appear as abrupt change in signal amplitude (e.g. motion artifacts)

    Overlaps with neural signal in both temporal & spectral domain

    Different waveform shapes

    Presented By Md Kafiul Islam

    WT:

    Good time-frequency resolution

    Can work with non-stationary signals, e.g. neural signal

    Easy to implement [complexity:DWT-> O(N); FFT -> O(N log2N);N-> length of signal]

    Can work for both single and multi-channel recordings

    Most importantly it can be used for both detection(from decomposed coefficient) and

    removal(thresholding and reconstruction) of artifacts.

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    20/46

    Why Wavelet Transform:

    DWT is applied on raw neural signal (decomposition level, L = 5) which is contaminated by artifacts. The coefficients of decomposed signal components can

    localize the artifact regions.

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    21/46

    Proposed Solution

    Presented By Md Kafiul Islam

    Purpose of Algorithm

    Minimum (or almost no)

    distortionto neural signal

    Remove artifacts as much as

    possible

    Should be automatic

    Robustnessis important

    Able to implement online

    Should work in both single and

    multi-channel analysis

    Should not depend on artifact

    types.

    Traditional Wavelet

    Denoising

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    22/46

    Why SWT (1) ?We prefer to use Stationary Wavelet Transform (SWT) instead of DWTand CWT:

    Usually DWT or SWT is preferred over CWT when signal synthesis is required

    CWT is very slow and generates way too much of data.

    SWT is translation invariant where DWT is not. So better reconstruction result (No loss ofinformation, preserves spike data and doesnt generate any spike-like artifacts).

    Choice of mother wavelets for CWT is limited.

    SWT implementation complexity [O(N L)] is in between DWT [O(N)] and CWT [O(N Llog2N)].

    N= length of signal, L = decomposition level

    Presented By Md Kafiul Islam

    Digital implementation of SWT:

    A 3 level SWT filter bank and SWT filters

    A 2-Level DWT decomposition and the

    reconstruction structures

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    23/46

    Why SWT (2) ?

    Presented By Md Kafiul Islam

    FPR

    TP = # True Positives (Hit)

    FP = # False Positives (False Alarm)

    TN = # True Negatives (Correct Rejection)

    FN = # False Negatives (Misdetection)

    0 100 200 300 400 500 600 700 800 900 1000-10

    -5

    0

    5Spike data comparison after artifact removal

    NormalizedAmplitude

    0 500 1000 1500 2000-15

    -10

    -5

    0

    5

    10

    15

    Time Sample

    Ref

    DWT

    CWT

    SWT

    Original Spike(True Positive)

    False Spike(False Positive)

    False Spike(False Positive)

    Original Spike(True Positive)

    Original Spike(True Positive)

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    24/46

    Effect of Filtering

    Separate spikes from artifacts

    Presented By Md Kafiul Islam

    0 1 2 3 4 5 6 7 8-1000

    -500

    0

    500Real Data from Monkey Front Cortex

    0 1 2 3 4 5 6 7 8-1000

    -500

    0

    500

    Amplitude

    0 1 2 3 4 5 6 7 8-1000

    -500

    0

    500

    Time, Sec

    Original

    Reconstructedby only SWT

    Reconstructed bySWT + Filtering

    0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

    0

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    0.7

    0.8

    0.9

    1

    ROC for Spike Detection

    FPR

    TPR

    SWT + Filtering

    Only

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    25/46

    ThresholdValue

    Universal Threshold:

    Wi= Wavelet coefficients; i = variance of Wi; N = length of signal

    Modified Threshold:

    Presented By Md Kafiul Islam

    k= kAfor approx. coef.

    kDfor detail coef.

    By empirical observation from

    signal histogram

    5 < m < infinite

    2 < n < 3

    D4, D5, D6contain the frequency

    band of spikes.

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    26/46

    Choice of Threshold Function (Garrote) Hard: Discontinuous which may produce large variance (very sensitive to small changes

    in the input data)

    Soft: Continuous but has larger bias in the estimated signal (results in larger errors)

    Garrote: Less sensitiveto input change, lower bias and more importantly continuous.

    Presented By Md Kafiul Islam

    Hard GarroteSoft

    P f E l ti

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    27/46

    Performance Evaluation(Important Definitions)

    Simulation is performed on both real andsynthesized (semi-simulated) signal database

    from different subjects.

    Removal Measurement

    Lamda, : Amount of artifact reduction

    SNR: Improvement in signal to noise (artifact) ratio

    Distortion Measurement

    RMSE: Root mean square error

    Spectral Distortion:

    Presented By Md Kafiul Islam

    x(n) = Reference signal

    x(n) = Reconstructed signal

    y(n) = Artifactual signal

    e1(n) = error between x & y

    e2(n) = error between x & x

    Rref= auto-correlation of reference signal

    Rrec= cross-correlation between

    reference and reconstructed signal

    Rart= cross-correlation between

    reference and artifactual signal

    Artifact SNR:

    Consider artifact as signal andneural signal as noise:

    h i f i l i

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    28/46

    Data Synthesis for Simulation

    Presented By Md Kafiul Islam

    Clean in-vivo

    Data (Reference)

    Raw In-Vivo Data

    With Artifacts

    Extract Artifact

    Templates

    Synthesized

    ArtifactualData

    Random

    AmplitudeRandom

    Location

    Random

    Duration

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    29/46

    Results (Tested on SynthesizedSequence)

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    30/46

    Results (Tested on SynthesizedSequence)

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    31/46

    Results

    (Tested on RealSequence-1)

    Presented By Md Kafiul Islam

    Data Sample 1: Rat Hippocampus

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    32/46

    Results

    (Tested on RealSequence-2)

    Presented By Md Kafiul Islam

    0 0.5 1 1.5 2 2.5 3 3.5 4

    5

    -8

    -6

    -4

    -2

    0

    2

    4Recorded vs Reconstructed (Before & After Artifact Removal)

    Time Sample

    SignalAmplitude,mV

    Reconstructed

    Recorded

    Data Sample 1: Rat Hippocampus

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    33/46

    Results(Tested on RealSequence-3)

    0 0.5 1 1.5 2

    x 105

    -0.25

    -0.2

    -0.15

    -0.1

    -0.05

    0

    0.05

    0.1

    0.15

    0.2Original vs Reconstructed

    Time Sample

    SignalAmplitude

    Reconstructed

    Original

    Presented By Md Kafiul Islam

    Data Sample 3: Cat Spinal Cord (High-pass Filtered @300 Hz)

    5.7 5.75 5.8 5.85 5.9 5.95 6 6.05 6.1 6.15

    x 104

    -6

    -4

    -2

    0

    2

    4

    6

    8

    Spike Data before & after artifact removal

    Time Sample

    NormalizedAmplitude

    Original Spikes

    Reconstructed Spikes

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    34/46

    Results

    (Tested on RealSequence-4)

    0 0.5 1 1.5 2 2.5 3

    x 105

    -800

    -600

    -400

    -200

    0

    200

    400Original vs Reconstructed

    Time Sample

    SignalAmplitude

    Reconstructed

    Original

    Presented By Md Kafiul Islam

    1 1.02 1.04 1.06 1.08 1.1 1.12

    x 105

    -35

    -30

    -25

    -20

    -15

    -10

    -5

    0

    5

    10

    Spike Data before & after artifact removal

    Time Sample

    NormalizedAmplitude

    Original Spike Data

    Recons Spike Data

    Data Sample 4: Monkey Front Cortex

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    35/46

    Quantitative Evaluation-1

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    36/46

    Quantitative Evaluation-2

    Presented By Md Kafiul Islam

    i i h h h d

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    37/46

    Comparison with Other Methods

    Presented By Md Kafiul Islam

    , dB

    dB

    dB

    Artifact

    Artifact

    C i i h O h M h d

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    38/46

    Comparison with Other Methods

    Presented By Md Kafiul Islam

    Artifact

    dB

    Artifact

    dB

    C i ith Oth M th d

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    39/46

    Presented By Md Kafiul Islam

    Comparison with Other Methods

    Date RMS

    Date RMS

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    40/46

    Conclusion

    First time(to best of knowledge) Investigation of

    artifacts for in-vivo neural data Artifact characterization

    Dynamic range study due to artifacts

    Database synthesis for quantitative evaluation Proposal of a detection and removal algorithm

    Threshold improvement

    Robust (Depends only on neural signals spectral char)Automatic

    Real-time implementable

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    41/46

    Applications

    Any closed loop neural system (e.g. BCI orneural prostheses)

    Basic neuroscience study

    Clinical research Removal of stimulationartifacts.

    Both online and offline implementation

    Both single and multi-channel recordings

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    42/46

    Future Plan-1

    Optimize the algorithm further to allow faster

    processing and less storage.

    Perform additional experiments (simulations) in

    order to fine tune the algorithm.

    Proceed to hardware implementation and

    perform real-time experiments to verify the

    actual performance in practice.

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    43/46

    Future Plan-2

    Publish the artifact database to public domain to

    facilitate future research.

    Development of a Software (MATLAB based) tool

    for offline analysis that will be open for all to

    download.

    Presented By Md Kafiul Islam

    Future Plan 3

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    44/46

    Future Plan-3

    Primary Work (with Dr. Amir)

    A Wavelet Transform Based Algorithm for UsableSpeech*

    Segments Extraction from Co-Channel Signals

    Presented By Md Kafiul Islam

    Unvoiced frame and

    its DWTVoiced frame and its

    DWT

    *Can be replaced by in-vivoneural signal

    P bli ti

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    45/46

    Publications

    In Preparation (Journal):

    1. Md Kafiul Islam, Amir Rastegarnia, Nguyen A. Tuan, and Zhi Yang, A Wavelet Based Artifact Detection

    and Removal Algorithm for In-Vivo Neural Recording In preparation for submission to Journal of

    Neural Eng.

    2. Jian Xu, Md Kafiul Islamand Zhi Yang, A 13W14-BitModulator for Wide Dynamic Range Neural

    Recording In preparation for submission to IEEE Trans. On BioMed. Circuit & Systems.

    Submitted (Conference):

    1. Jian Xu, Md. Kafiul Islam, and Zhi Yang, A 13W 87dB Dynamic Range Implantable Modulator for

    Full-Spectrum Neural Recording Submitted to EMBC13

    2. Azam Khalili, Amir Rastegarnia, Md Kafiul Islam, Zhi Yang, A Bio-Inspired Cooperative Algorithm for

    Distributed Source Localization with Mobile Nodes Submitted to EMBC13

    Accepted (Conference):

    1. Md. Kafiul Islam, N Tuan, Y. Zhou, and Z. Yang, Analysis and Processing of In-Vivo Neural Signal for

    Artifact Detection and Removal - Accepted in the International Conference on BioMedical Engineering

    and Informatics (BMEI), October 2012, Chongqing, China.

    Presented By Md Kafiul Islam

  • 8/11/2019 Artifact Removal for in-vivo Neural Signal

    46/46

    The End

    Q& A

    Thank You