Upload
sheshadri-iyengar
View
66
Download
1
Embed Size (px)
Citation preview
Automated Analysis of Auditory Brainstem Evoked Response using Wavelets And Neural Networks
Dr. V. UDAYASHANKARA PROF. DEPT OF IT,SJCE
MYSORE
BY
JYOTHI.BASST.PROFF, DEPT.OF E&C,
NIE-IT, MYSORE
GUIDEDBY
1
CONTENTS
• Introduction
• Algorithm
• Implementation
• Results
• Conclusion
2
INTRODUCTION
• What is Brainstem Evoked Response Audiometry(BERA)?
• What is the principle?
• What are its applications?
3
• Brainstem response is Electrical signal evoked in human Brainstem due to presentation of sound such as click or tone.
• Brainstem Evoked Response Audiometry is a screening test to monitor the Hearing loss or Deafness in a patient .
4
BLOCK DIAGRAM
5
CLINICAL APPLICATIONS
• Estimation of threshold
• Investigation of hearing loss
6
TYPICAL BERA WAVEFORM
7
ORIGIN OF EACH WAVE:
PEAK ORIGIN
I Cochlear nerve
IIDorsal & Ventral cochlear
nucleus
III Superior olivary complex
IV Nucleus of lateral lemniscus
V Inferior colliculus
VI Medial geniculate body
VII Auditory radiation(cortex8
OBJECTIVE :
The aim of this paper is to develop software
classification model to assist the audiologist with
an automated detection of the ABR waveform and
also detection of peaks for identification of
pathologies.9
ALGORITHM:
• Extraction of BERA from recorded EEG
• Identification of peaks
• Classification
10
FLOW CHART
Get the EEG Data
Is a BERA response present
Identification of peaks
Peak detection
Abnormal Normal
no
Classification
11
IMPLEMENTATION:
Data Acquisition: EEG data is collected in JSS hospital and data
base is created which consists of 30 normal (neonates)and 25
abnormal patients..
12
EXTRACTION:
• Band pass filtering with cut-off freq. 30-3000 Hz.
• Signal adaptive filtering using complex wavelets
• This algorithm uses Dual tree complex wavelet
transform.13
PEAK IDENTIFICATION
• Gaussian Derivative estimation filter used
Steps :
• Find Zero Crossings and
• Peak Detection using Local Max And Min Value.
• Finally Peak Labeling.
14
FLOW CHART:
15
PARAMETERS USED FOR LABELING PEAKS
16
17
CLASSIFICATION :
• Artificial Neural Network(ANN) is used .
• Diagnosis of hearing disorder is one of the applications of
neural network.
18
• ANN is defined by 3 types of parameters:
1. Inter connection pattern layers of neurons.- nonlinear
2. Learning process for updating weights
3. The activation function
19
FEED-FORWARD ARCHITECTURE
20
BLOCK DIAGRAM
21
LEARNING ALGORITHM :
22
PATIENT INFORMATION
• Intensity• Ear• peak I• Amplitude of Peak I• peak III• Amplitude of peak III• peak V• Amplitude of peak V• Inter peak (I – III)• Inter peak (III – V)• Inter peak (I – V)
23
RESULT:
1. If the output node value is >0.7467 --- Normal
2. Else if output node value is <= 0.7467 --- Abnormal
24
TESTS AND RESULTS:
0 100 200 300 400 500 600 700 800-5000
0
5000
10000
0 100 200 300 400 500 600-1
-0.5
0
0.5
1x 10
4 Complex 1-D wavelet
t
(t
)
Fig: Extracted BERA signal25
Fig: Gaussian First Derivative:26
Fig: First deriv. of the Input signal
-15 -10 -5 0 5 10 15-4
-3
-2
-1
0
1
2
3First Derivative
Time
Am
plitu
de
27
Fig: Zero crossing & peak identification28
Fig: Peak Labeled29
GUI in MATLAB :
30
31
GUI in C:
32
33
ANN RESULTS:
Normal Abnormal
Training Number 30 25
Test Number 25 20
Correct Classification
24 18
Rate of classification
96% 90%
34
Accuracy of ANN classification
35
CONCLUSION
• This paper demonstrates the feasibility of an algorithm for extraction of clean BAER waveforms and subsequent automatic peak identification in order to perform functional assessment of the brainstem.
• Automated method gives 96% for normal 90% Accuracy for Abnormal patients.
36
REFERENCES:
• [1]A. Jacquin, E. Causevic, E. R. John, and L. S. Prichep, “Optimal denoising of brainstem auditory evoked response (BAER) for automatic peak identi- fication and brainstem assessment,” in Proc. 28th IEEE
• [2]A. Jacquin, E. Causevic, E.R. John, J. Kovacevic “Adaptive complex wavelet-based filtering of EEG for extraction of evoked potential responses,” ICASSP’04, 2004.
• [3]Automated Analysis of the Auditory Brainstem Response Using Derivative Estimation Wavelets Andrew P. Bradleya, Wayne J. Wilsonb
• [4]Hall III, J., Handbook of auditory evoked responses,Needham Heights, Massachusetts: Allyn and Bacon,1992.
• [5] Mallat, S.G., A Wavelet tour of signal processing,2nd Ed., San Diego: Academic Press, 1999.
37
THANK U
38
39