Introduction The aim the project is to analyse non real time EEG (Electroencephalogram) signal using different mathematical models in Matlab to predict.

IntroductionIntroductionThe aim the project is to analyse non real time EEG (Electroencephalogram) The aim the project is to analyse non real time EEG (Electroencephalogram) signal using different mathematical models in Matlab to predict abnormal signal using different mathematical models in Matlab to predict abnormal derivation of the signal applying frequency spectral analysis for linear, derivation of the signal applying frequency spectral analysis for linear, continuous or discrete input data signal. This will involve a filtering pre-continuous or discrete input data signal. This will involve a filtering pre-processing stage, Short Time Fourier Transform, DFT, FFT, AR Model, processing stage, Short Time Fourier Transform, DFT, FFT, AR Model, Sonification and Hidden Markov Model (HMM) for more that one signal Sonification and Hidden Markov Model (HMM) for more that one signal with a further application in Bayes Networks Classifier.with a further application in Bayes Networks Classifier.

ObjectivesObjectivesThe project will study new techniques for the analysis of EEG and the The project will study new techniques for the analysis of EEG and the automated diagnostic of the pathologiesautomated diagnostic of the pathologies..

Data will be analysed using AR model because this technique will study Data will be analysed using AR model because this technique will study information extraction from signals that are a- periodic, noisy, intermittent information extraction from signals that are a- periodic, noisy, intermittent or transient from a tiny signal, which contain very small amplitude and or transient from a tiny signal, which contain very small amplitude and period .period .

Sonification of the EEG data is applied to obtain an acoustic representation Sonification of the EEG data is applied to obtain an acoustic representation of the signal in a spectral form. The sonification technique will convert the of the signal in a spectral form. The sonification technique will convert the spectrogram frequencies of the EEG data in audible sound to detect the spectrogram frequencies of the EEG data in audible sound to detect the disease.disease.

Hidden Markov Model (HMM) will process different EEG data as stochastic Hidden Markov Model (HMM) will process different EEG data as stochastic sequences of events.sequences of events.

EEG data will be imported by Matlab and the model is applied in a selected EEG data will be imported by Matlab and the model is applied in a selected normal and abnormal signal as Epilepsy, Arrhythmia or whatever EEG normal and abnormal signal as Epilepsy, Arrhythmia or whatever EEG supplied datasupplied data . .

Scanning and Detection of EEG Scanning and Detection of EEG Diseases Using Medical Signal Diseases Using Medical Signal

ProcessingProcessing1

MethodologyMethodology

The system has analysed two different data sets from the next sources: The 1st The system has analysed two different data sets from the next sources: The 1st data source contains normal EEG data from Colorado State University The 2nd data source contains normal EEG data from Colorado State University The 2nd data source contains Epilepsy EEG data from Bonn University (Germany).data source contains Epilepsy EEG data from Bonn University (Germany).

EEG Data is provided in mat file or txt file.EEG Data is provided in mat file or txt file. Matlab will give the option to create Matlab will give the option to create scripts for the models using the DSP, System Identification, Hidden Markov scripts for the models using the DSP, System Identification, Hidden Markov Model (HMM), Wavelet Transform and Neural Networks toolbox,.Model (HMM), Wavelet Transform and Neural Networks toolbox,.

Feature Extraction: EEG signal will be pre-processed to eliminate the noise using Feature Extraction: EEG signal will be pre-processed to eliminate the noise using the Band Pass filter Butterworth IIR because the 1st data set contains noise as row the Band Pass filter Butterworth IIR because the 1st data set contains noise as row signal. It can affect to the next applied models, but the wrong results affects signal. It can affect to the next applied models, but the wrong results affects mainly to the periodogrammainly to the periodogram..

AR (Autoregressive) Model will study the behaviour of the EEG signal AR (Autoregressive) Model will study the behaviour of the EEG signal coefficients for large or small frequency samples in linear form. ARburg model is coefficients for large or small frequency samples in linear form. ARburg model is applied for small EEG data windows and frequency samples.applied for small EEG data windows and frequency samples.

Sonification model will analyse the spectrum of the signal by differential Sonification model will analyse the spectrum of the signal by differential sonification and Short Time Fourier Transform (STFT) to find the harmonics and sonification and Short Time Fourier Transform (STFT) to find the harmonics and lobe bands. Frequencies generates audible tones (5 to 90 Hz). lobe bands. Frequencies generates audible tones (5 to 90 Hz).

Hidden Markov Model (HMM) analyses data to detect the diseases by observation Hidden Markov Model (HMM) analyses data to detect the diseases by observation of the input classes or sequences. Also HMMof the input classes or sequences. Also HMM classifies it by events in a Gaussian classifies it by events in a Gaussian 2D of each state of the signal. Then the signal will contain a sequence of events 2D of each state of the signal. Then the signal will contain a sequence of events called Markov Chain with Gaussian densities. Bayesian Classification estimates called Markov Chain with Gaussian densities. Bayesian Classification estimates the optimal sequence by Viterbi Algorithm.the optimal sequence by Viterbi Algorithm.

EEG DATA IN FILE FORM

MATLAB ANALYSISTOOLBOX OF ALGORITHMS

INPUT TIME SIGNAL

FEATURE EXTRACTION

HIDDEN MARKOV MODEL

ARMODEL

DECISIONCLASSIFICATION

LEVINSON DURBINRECURSION

WAVELETANALYSIS (optional)

SONIFICATIONS FOR EEG DATA ANALYSIS

DATA SETS

2

AR ModelAR ModelNormal EEG Data, AR 9Normal EEG Data, AR 9thth , 10 , 10thth and 11 and 11thth

window. Blinking Eyeswindow. Blinking EyesEpilepsy EEG Data, AR 7Epilepsy EEG Data, AR 7 thth , 8 , 8thth and 9 and 9thth

windowwindow. .

SonificationSonificationNormal Spectrogram EEG Data 9Normal Spectrogram EEG Data 9 th,th,

10 10thth and 11 and 11thth window window. .

Epilepsy Spectrogram EEG Data 7Epilepsy Spectrogram EEG Data 7 th,th,

8 8thth and 9 and 9thth window window. .

Periodogram EEG C3 (noisy line) channelPeriodogram EEG C3 (noisy line) channel

Periodogram EEG Epilepsy 7Periodogram EEG Epilepsy 7 thth window channel window channel

Hidden Markov Model (HMM)Hidden Markov Model (HMM)

Class A/B

Component 1/2

Gaussianmu, sigma

InputNode 1 Node 2

OutputNode 3

Model Accuracy (%) p-value

EEG Data HMM-1 Must be 100% Must be 1.0

EEG Data HMM-2 Must be 100% Must be 1.0

3

ResultsResultsAR model estimates the arburg coefficients from normal EEG signal and AR model estimates the arburg coefficients from normal EEG signal and Epilepsy signal.. Epilepsy signal..

Normal EEG Data: linear vector. Normal EEG Data: linear vector. Epilepsy EEG Data: logarithmic curve vector with an optimal point to show Epilepsy EEG Data: logarithmic curve vector with an optimal point to show

the critical state in the seizure. the critical state in the seizure.

Sonification: Sonification: 1. The Probability Density Estimation calculates three gaussian kernel 1. The Probability Density Estimation calculates three gaussian kernel

bandwidth approximations (default widths).bandwidth approximations (default widths). Normal EEG Data: Gaussian widths almost matched.Normal EEG Data: Gaussian widths almost matched. Epilepsy EEG Data: Mismatch gaussian widths.Epilepsy EEG Data: Mismatch gaussian widths. 2. The spectrograms show small amplitude values for light colours and high 2. The spectrograms show small amplitude values for light colours and high

amplitude values for dark colours in the Short Time Fourier Transform. The amplitude values for dark colours in the Short Time Fourier Transform. The intensity of the frequency colours give the harmonics of the pattern plotted.intensity of the frequency colours give the harmonics of the pattern plotted.

3. Spectral sonification is audible to human ear (5Hz to 90Hz). The 3. Spectral sonification is audible to human ear (5Hz to 90Hz). The amplidude of the EEG signal changes the tone range.amplidude of the EEG signal changes the tone range.

Hidden Markov Model (HMM): EEG values have to be -5 to 5 to avoid Hidden Markov Model (HMM): EEG values have to be -5 to 5 to avoid mismatches between data and initial random process. EEG signals are low mismatches between data and initial random process. EEG signals are low correlated, except sleep stages. AR coefficients (EEG signal) are trained in correlated, except sleep stages. AR coefficients (EEG signal) are trained in two models with higher (log-) likelihood value. HMM1 and HMM2 models two models with higher (log-) likelihood value. HMM1 and HMM2 models are compared in the table to show the classification accuracies and the are compared in the table to show the classification accuracies and the intervals with the standard deviation.intervals with the standard deviation.

Future WorkFuture WorkImplement Hidden Markov Model (HMM) using the Factorial Markov Model (FMM) and Implement Hidden Markov Model (HMM) using the Factorial Markov Model (FMM) and Boyen-Kollen algorithm for a Bayes Network Classifier.Boyen-Kollen algorithm for a Bayes Network Classifier.

Classification using Neural Network Classifier.Classification using Neural Network Classifier.

EEG analysis using Wavelet Transform and classification of the Wavelet Feature EEG analysis using Wavelet Transform and classification of the Wavelet Feature Extraction.Extraction.

Luis Acevedo – MSc Embedded Systems (2004)Luis Acevedo – MSc Embedded Systems (2004)Supervisor : Dr. Yvan PetillotSupervisor : Dr. Yvan Petillot

4