Detection of Brain Abnormalities using Hilbert-Huang Transform

Download Full-Text PDF Cite this Publication

Text Only Version

Detection of Brain Abnormalities using Hilbert-Huang Transform

Munaza Peerjade1

1M.Tech., Students, Rajarambapu Institute of Technology, India.

Dr. Mahesh Kumbhar2

2Assistnat Professor, Rajarambapu Institute of Technology, India.

Abstract:- This paper presents the analysis of classifying between normal and abnormal patients using the featured based on Hilbert-Huang Transform of EEG Signal. By analyzing the EEG Signal from accessible database discrimination is achieved. The intrinsic function information which is in EEG Signal can be extracted with the help of s Hilbert-Huang Transform this information helps to get the local amplitude of the signal and frequency. Based on the information weighted frequency is calculated and the comparison between the intrinsic functions of normal and abnormal determinant is performed.

Keywords: Weighted Frequency, Intrinsic Frequency, EEG, Hilbert-Huang

MATLAB Software is used for implementing and testing of the Classification Algorithm which is proposed. So, using the features based on HHT i.e. Hilbert-Huang Transform the Classification of normal and abnormal activities can be done in this paper.

  1. INTRODUCTION

    EEG i.e. Electroencephalography it is a method providing information necessary for the classification, of normal and abnormal patients by doing proper diagnosis. The information can be obtained with the help of energy content

    Objectives:

    Fig .1. Wakefulness of person

    and the frequency which is divided into delta theta alpha and beta. Normal and Abnormal patients diagnosis can be done with the help of information obtained through the computerized Analysis. The methods efficient in classification of abnormal and normal patient are amplitude, frequency, and phase which are oscillatory information. These feature helps in comparison between normal and abnormal brain activities. Hilbert Transform not only gives spectral information but also gives analytical signal representation. The Aim and Objective of this paper is Coupling of EMD i.e. Empirical Mode Decomposition which is also called Huang Transform with Hilbert Transform to get the results of classification

  2. RELATED WORK

    The abnormality can be seen in the persons brain EEG Signal due to the disturbance or changes in the electrochemical activity of the neurons which leading to abnormal and synchronous discharges. The most difficult process is analysis of brain signals i.e. EEG Signals for brain abnormalities detection. So it develops a need of automatic

    system which is PC based for brain abnormalities detection. Types of brain waves is shown in Fig.1 that are recorded on an EEG test it shows different state of wakefulness in which abnormalities are recorded .In this presented work discrimination is achieved by EEG signals database which is obtained from doctors or any freely accessible sources.

    • Obtain Normal and Abnormal EEG Signal from Data Base.

    • Pre-Processing of EEG Signal.

    • Develop Code for extraction of Features/Parameters like Standard Deviation, Mean, Band Powers,

      Energy etc.

    • Development of code for Hilbert-Huang Transform.

    • Classification of Normal and Abnormal Signal

  3. METHOD

    1. Hilbert Transform

      With the help of EEG Signal Features the classification of the normal and abnormal activities of brain function is achieved using HHT i.e. Hilbert-Huang Transform. The analysis of different oscillatory modes such as energy and the frequency content of every brain wave is done. This can be achieved by tracking amplitude and frequency content of every signal whether normal or abnormal. By calculating the weighted frequency using the Hilbert transform discrimination between normal and abnormal patients is achieved. Huang Transform is used by coupling it with Hilbert Transform for getting the narrow banded signals.

    2. Huang Transform

      Huang transform is also known as EMD i.e. Empirical Mode Decomposition which is an signal processing technique this technique is used to extract the oscillatory modes which are

      embedded in a brain signal. The main advantage is that it does not require any linear or Stationary Data. This EMD mode is Data Driven Mode i.e. it does not require any resolution or harmonics. The amplitude and instantaneous frequency is defined by using Intrinsic Mode Functions. The Hilbert Transform can easily be applied to each and every single intrinsic mode. For the extraction of intrinsic mode the shifting process is done.

      The proposed work contain following Steps,

      1. The Signal Processing Technique named as EMD is used for extraction of the oscillatory mode without the requirement of linear or stationary data. As compared to wavelet mode why EMD is used here is it has no resolution and harmonics present. This decomposition technique is applied to EEG signal with the main objective of extracting intrinsic mode in EEG Signal

      2. Tracking of the amplitude and instantaneous frequency is done by Hilbert Transform.

      3. After Applying the HHT Algorithm based on the features extraction the classification is done between the normal and abnormal patients

    Block Diagram

    Fig.2. Block Diagram

    In this work it is proposed to carry out the classification using Hilbert- Huang Transform of EEG Signals to Detect Brain Abnormality. The block diagram shown in Fig.2 represents the steps which are carried out i.e. The collected EEG Data Base is read first then Signal Processing is done after Comparison and Analysis the results are considered and decision making is done on normal and abnormal person.

        1. Pre-processing

          EEG Matlab Software

          Fig4. Importing Data

          The figure shows way to load data can be any Matlab,Binary.ACSII etc.

          Load Dataset

          Fig 5. Loading Dataset

          The figure shows that the dataset can also be loaded from existing data which is present in .set format

          Data Information Scrolling Data

          Fig 6. 32 Channel Data information

          The figure shows 32 channel data information in the form of signals with time and value with respect X-axis and Y-axis.

          Editing and Preprocessing

          1. Edit/Select Data

            0

            Fig 3. EEG Matlab Software

            The figure shows the first screen after running the EEGtoolbox when there is no data loaded

            Importing Data Import/Load Data

            Fig 7. Editing Data.

            The Figure shows that the data can be edited, event values, event fields and even sampling rate can be edited.

          2. Pre-processing

            Fig 8. Processing Data

            The figure shows various data processing techniques in tool

          3. Rejecting Artifacts in Continuous Data

            Fig. 11 Channel Statistics

            Figure shows channel statistics with respect to Mean, Guassian and triimed.

            2) Feature Selection

            Feature extraction: The dimensionality of the features is reduced by using this feature extraction method. The Characteristics of Original signal without much redundancy is presented..

            Time domain features: Statistical calculations are included in Time Domain Analysis. Mean, Median, Mode, Standard deviation, Maximum and Minimum are the time domain features.

            Fig 9.plot of power vs frequency

            The figure shows channel spectra plot with respect to frequency. Every different color refers to different channel in total 32 channels 32 colors with pop spectopo() as command

          4. Plot Data Spectrum and Maps

            Fig 10. Rejection of Data

            The figure shows we can reject particular data fro signal and new data set can be created with marked regions being removed. There are two event types square and rt. Reaction time is given by rt we can see different reaction at different places. The time and value are represented at x axis and y axis it changes with the position on the signal.

          5. Channel Statistics

  4. RESULT

    1. Waveforms for EEG Normal Signal

      Fig. 12. Original Signal For Normal Signal 1

      Fig 13. Different Ranges of Brain Waves in TD

      Fig 14. Different Ranges of Brain Waves in FD

      Fig 15. Time Frequency Plot for Normal EEG

      Fig 16. Amplitude Vs Time for Different Ranges

      Fig 17. Amplitude Vs Time with IMF for

      Fig 18. IMF Amplitude vs Time for Normal Signal

      Table 1- Time Domain Features for Alpha Waves

      Table 2- Time Domain Features for Beta Wave

      EEG Waves

      Min

      Max

      Entropy

      Std.Deviation

      Normal 1

      -0.1349

      0.1217

      3.0148

      0.0370

      Normal 2

      -0.1347

      0.1257

      3.0342

      0.0375

      Normal 3

      -0.1349

      0.1217

      3.0153

      0.0371

      Normal 4

      -0.1349

      0.1217

      3.0148

      0.0370

      Normal 5

      -0.1246

      0.1008

      2.9762

      0.0348

      Normal 6

      -0.1349

      0.1217

      3.0143

      0.0371

      Normal 7

      -0.1401

      0.1427

      3.0735

      0.0400

      Normal 8

      -0.1254

      0.1382

      3.0691

      0.0394

      Normal 9

      -0.1361

      0.1417

      3.1478

      0.0437

      Normal 10

      -0.1615

      0.1468

      3.1916

      0.0458

      Table 3- Time Domain Features for Delta Waves

      EEG Waves

      Min

      Max

      Entropy

      Std.Deviation

      Normal 1

      -0.0998

      0.1032

      2.8523

      0.0381

      Normal 2

      -0.0978

      0.1033

      2.9222

      0.0381

      Normal 3

      -0.1002

      0.1032

      2.8537

      0.0381

      Normal 4

      -0.0998

      0.1032

      2.8523

      0.0381

      Normal 5

      -0.0958

      0.1027

      2.7933

      0.0392

      Normal 6

      -0.0946

      0.1032

      2.8451

      0.0380

      Normal 7

      -0.1004

      0.1032

      2.8529

      0.0375

      Normal 8

      -0.1665

      0.0981

      2.8592

      0.0677

      Normal 9

      -0.1827

      0.1212

      1.8698

      0.0767

      Normal 10

      -0.1881

      0.1226

      2.8210

      0.0867

      Table 4-Time Domain Features For Gamma Waves

      EEG Waves

      Min

      Max

      Entropy

      Std.Deviation

      Normal 1

      -0.1265

      0.1362

      3.1330

      0.0411

      Normal 2

      -0.1265

      0.1362

      3.1352

      0.0413

      Normal 3

      -0.1265

      0.1362

      3.1330

      0.0411

      Normal 4

      -0.1265

      0.1362

      3.1330

      0.0411

      Normal 5

      -0.1265

      0.1362

      3.1570

      0.0418

      Normal 6

      -0.1265

      0.1362

      3.1340

      0.0411

      Normal 7

      -0.1265

      0.1362

      3.1519

      0.0424

      Normal 8

      -0.1265

      0.1362

      3.1504

      0.0419

      Normal 9

      -0.1257

      0.1362

      3.1597

      0.0425

      Normal 10

      -0.1262

      0.1521

      3.2026

      0.0446

      Table 5-Time Domain Features For ThetaWaves

      EEGWaves

      Min

      Max

      Entropy

      Std.Deviation

      Normal 1

      -0.1862

      0.1560

      3.2323

      0.0570

      Normal 2

      -0.1868

      0.1559

      3.2164

      0.0563

      Normal 3

      -0.1862

      0.1560

      3.2321

      0.0570

      Normal 4

      -0.1862

      0.1560

      3.2323

      0.0570

      Normal 5

      -0.1849

      0.1545

      3.2363

      0.0556

      Normal 6

      -0.1862

      0.1560

      3.2314

      0.0569

      Normal 7

      -0.1862

      0.1512

      3.2172

      0.0552

      Normal 8

      -0.1278

      0.1416

      3.0841

      0.0420

      Normal 9

      -0.1488

      0.1223

      3.2789

      0.0529

      Normal 10

      -0.1410

      0.1239

      3.3617

      0.0519

      EEGWaves

      Min

      Max

      Entropy

      Std.Deviation

      Normal 1

      -0.1677

      0.1507

      3.410

      0.0619

      Normal 2

      -0.1715

      0.1511

      3.4115

      0.0617

      Normal 3

      -0.1677

      0.1507

      3.4102

      0.0619

      Normal 4

      -0.1677

      0.1507

      3.4101

      0.0619

      Normal 5

      -0.1737

      0.1528

      3.3757

      0.0606

      Normal 6

      -0.1677

      0.1507

      3.4102

      0.0619

      Normal 7

      -0.1677

      0.1519

      3.3846

      0.0595

      Normal 8

      -0.1912

      0.1612

      3.4156

      0.0652

      Normal 9

      -0.1777

      0.1656

      3.5006

      0.0724

      Normal10

      -0.1725

      0.1594

      3.4755

      0.0719

      0.0619

      EEGWaves

      Min

      Max

      Entropy

      Std.Deviation

      Normal 1

      -0.1677

      0.1507

      3.410

      Normal 2

      -0.1715

      0.1511

      3.4115

      0.0617

      Normal 3

      -0.1677

      0.1507

      3.4102

      0.0619

      Normal 4

      -0.1677

      0.1507

      3.4101

      0.0619

      Normal 5

      -0.1737

      0.1528

      3.3757

      0.0606

      Normal 6

      -0.1677

      0.1507

      3.4102

      0.0619

      Normal 7

      -0.1677

      0.1519

      3.3846

      0.0595

      Normal 8

      -0.1912

      0.1612

      3.4156

      0.0652

      Normal 9

      -0.1777

      0.1656

      3.5006

      0.0724

      Normal10

      -0.1725

      0.1594

      3.4755

      0.0719

      1. Waveforms for Abnormal Signal

    Fig 19. Original Signal for Abnormal Signal

    Fig 20. Different Ranges of Brain Waves in TD

    Fig 21. Different Ranges of Brain Waves in TD Abnormal Signal

    Fig 22. Time-Frequency Plot for Abnormal Signal 1

    Fig 23. Amplitude Vs Time for Different Ranges in Abnormal EEG Signal1

    Fig 24. Amplitude Vs Time with IMF for Different Ranges in Abnormal EEG Signal 1

    Fig 25. IMF Amplitude vs Time for Abnormal Signal 1

    Table 6-Time Domain Features For Abnormal Alpha Waves

    EEG Waves

    Min

    Max

    Entropy

    Std. Deviation

    Abnormal 11

    -0.1840

    0.2068

    3.5141

    0.0725

    Abnormal 12

    -0.1839

    0.2072

    3.5264

    0.0731

    Abnormal 13

    -0.1809

    0.2050

    3.5729

    0.0775

    Abnormal 14

    -0.1827

    0.2064

    3.5524

    0.0767

    Abnormal 15

    -0.1822

    0.1771

    3.4553

    0.0710

    Abnormal 16

    -0.2033

    0.1839

    3.5451

    0.0758

    Abnormal 17

    -0.1764

    0.1753

    3.4793

    0.0718

    Table 7-Time Domain Features For Abnormal Beta Waves

    EEGWaves

    Min

    Max

    Entropy

    Std.Deviation

    Abnormal11

    -0.2414

    0.2214

    3.3813

    0.0570

    Abnormal12

    -0.2509

    0.1972

    3.3617

    0.0557

    Abnormal13

    -0.2257

    0.2041

    3.2918

    0.0510

    Abnormal14

    -0.2260

    0.2041

    3.2947

    0.0514

    Abnormal15

    -0.2234

    0.1795

    3.2890

    0.0517

    Abnormal16

    -0.1549

    0.1712

    3.2375

    0.0471

    Abnormal17

    -0.2237

    0.1831

    3.2789

    0.0517

    Table 8-Time domain feature extraction for delta waves

    EEGWaves

    Min

    Max

    Entropy

    Std. Deviation

    Abnormal11

    -0.2134

    0.1180

    2.5338

    0.0777

    Abnormal12

    -0.2127

    0.1179

    2.6311

    0.0795

    Abnormal13

    -0.1882

    0.1242

    2.7136

    0.0795

    Abnormal14

    -0.1857

    0.1226

    2.6454

    0.0805

    Abnormal15

    -0.1912

    0.1227

    2.5008

    0.0850

    Abnormal16

    -0.1936

    0.1523

    3.1195

    0.0988

    Abnormal17

    -0.1870

    0.1227

    2.6425

    0.0807

    Table 9-Time domain feature extraction for gamma waves

    EEG Waves

    Min

    Max

    Entropy

    Std.Deviation

    Abnorma111

    -0.2108

    0.1769

    3.2973

    0.0532

    Abnormal 12

    -0.2108

    0.1769

    3.3042

    0.0521

    Abnormal13

    -0.2051

    0.1769

    3.2452

    0.0477

    Abnormal 14

    -0.2051

    0.1679

    3.2328

    0.0472

    Abnormal 15

    -0.1445

    0.1638

    3.1698

    0.0430

    Abnormal 16

    -0.3178

    0.2827

    3.3062

    0.0537

    Abnormal 17

    -0.2189

    0.1816

    3.2254

    0.0464

    Table 10-Time domain feature extraction for theta waves

    EEGWaves

    Min

    Max

    Entropy

    Std.Deviation

    Abnormal 1

    -0.1840

    0.2068

    3.5141

    0.0725

    Abnormal 2

    -0.1839

    0.2072

    3.5264

    0.0731

    Abnormal 3

    -0.1809

    0.2050

    3.5729

    0.0775

    Abnormal 4

    -0.1827

    0.2064

    3.5524

    0.0767

    Abnormal 5

    -0.1822

    0.1771

    3.4553

    0.0710

    Abnormal 6

    -0.2033

    0.1839

    3.5451

    0.0758

    Abnormal 7

    -0.1764

    0.1753

    3.4793

    0.0718

  5. DISCUSSION

    After HHT Coding in MATLAB Software different results are obtained such as min and max of the respective signal, the entropy and Standard deviation , for the classification.

  6. CONCLUSION

For the diagnosis of epilepsy EEG Signal is very much important. For the detection of Brain abnormalities and brain disease and keeping the record of patient in long term and diagnosing huge amount of EEG data is needed. For detection of Epilepsy EEG Signal is must and the process is performed by experts in EEG labs of respective hospital this data is very private and secure. This HHT algorithm has the potential in classification of normal and abnormal activities using feature extraction method.

REFERENCES

  1. Enas W Abdulhay and Rami J Oweis (2011)Seizure classification in EEG signals utilizing Hilbert-Huang transform Oweis and Abdulhay BioMedical Engineering

  2. Rehman N, Xia Y, Mandic DP (2010) Application of Multivariate empirical mode decomposition for seizure detection in EEG signals .

  3. Xinyang Li, Cuntai Guan, Haihong Zhang, Kai Keng Ang (2017) Discriminative Ocular Artifact Correction for Feature Learning in EEG AnalysisIEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 64, NO. 8, AUGUST

  4. Saleha Khatun , Ruhi Mahajan , Bashir I. Morshed (2016) Comparative Study of Wavelet-Based Unsupervised Ocular Artificial Removal Techniques for Single-Channel EEG Data (2016) 2168-2372 2016 IEEE. Translations and content mining are permitted for academic research only.

  5. Robert Keight; Dhiya Al-Jumeily; Conor Mallucci (2017)Towards the discrimination of primary and secondary headache: An intelligent systems approach International Joint Conference on Neural Networks (IJCNN)

Leave a Reply

Your email address will not be published. Required fields are marked *