wearable respiratory monitoring: algorithms and system …

Wearable Respiratory Monitoring:

Algorithms and System Design

An IEEE Distinguished Lecture, 2011

A/Prof. Wee Ser [email protected]

School of EEE

Nanyang Technological University (NTU)

Singapore



An IEEE Distinguished Lecture, 2011

A/Prof. Wee Ser [email protected]

School of EEE


Singapore

Acknowledgement

IEEE Circuits and Systems Chapter, Vancouver

(Chapter Chair: Professor Ljiljana Trajkovic)

for inviting me to give this IEEE DL Talk

2

The IEEE

1-6

8

10

9

IEEE-USA

IEEE Canada

7

324 sections worldwide

3

IEEE Circuits and Systems

Society (CASS) www.ieee-cas.org

CASS is a technical society of IEEE

IEEE CASS

Membership (38 Societies in IEEE) > 400,000 8,513

Number of Chapters 1,784 152

CASS is the 9th largest IEEE Society (after CS, ComSoc,

MTTS, PES, SPS, SSCS, EDS, and IA)

4

• 1955: Established as the 1st Chinese-

medium university outside China

• 1981: Nanyang Technological Institute (NTI)

set up (English) with 3 Engineering Schools

• 1991: NTU was inaugurated, incorporating

National Institute of Education

• Now (2011): 3rd NTI/NTU President

–College of Engineering

–Nanyang Business School

–College of Science

–College of Humanities, Arts and Social Sciences

–School of Medicine

–Student Population > 30,000


5

Outline of Presentation

I. Introduction – A Clinical Perspective

II. Motivations and Problem Formulation

III. Wearable Wheeze Monitoring System @ NTU

a) Overall System Design

b) Wearable Signal Processing Algorithms

c) Wearable Sensors and Platform Suite

IV. Lab Tests, Pre-Clinical Trials and Results

V. Conclusions and Future Work

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



7

Nanyang Technological University, Singapore

A/Prof. W Ser, Dr. J Zhang, J Yu, Dr. T Zhang

National University Hospital, Singapore

A/Prof. Dr. Daniel YT Goh and Team

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



Research Team

8

Introduction -

A Clinical

Perspective

I

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



9

Respiratory Disorders

Chronic respiratory diseases include disorders that affect

any part of the respiratory system, not only the lung but also

the upper airway (nose, mouth, pharynx, larynx, and

trachea), chest wall and diaphragm, as well as the

neuromuscular system that provides the power for breathing

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



10

Major features of lungs:

bronchi, bronchioles and

alveoli

OSA disorder is characterized

by repetitive collapse of the

upper airway

Our Lungs and Airways

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



11

Basic Types of Respiratory Sounds

Normal Breath Sounds

Normal vesicular sounds (100 – 1000 Hz)

Normal tracheal sounds (from <100 to >3000 Hz) 18

Adventitious Sounds

Continuous sounds (e.g. Wheezes)

Discontinuous sounds (e.g. Crackles)

Upper Airway Sounds

Snore

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



13

Motivations

&

Problem

Formulation

II

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



18

Common and important cause of illness and death

Can affect people of all ages, both genders

300 million people suffer from Asthma and 210 million people with COPD (2007), world-wide

Responsible for over 10% of hospitalizations & over 16% of deaths in Canada (including lung cancer)

1 in 7 in UK affected by chronic lung disease

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



19

Why Monitor Respiratory Disorders ?

Market / Commercialization Potential

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



U$ (billions)

Source: GlobalData Report

20

All ages

Prevalence of Childhood Asthma:

* 10% - 20% (varies with countries)

Adult Asthma:

* 12%

Adult COPD:

* 5%

USA

23 million Asthmatics

Over 15 million cases of

COPD

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



21

Wheezing is a common noise in respiratory disorders

Wheeze Monitoring

Values

Accurate detection and assessment of wheezes

prompt diagnosis & appropriate treatment of Asthma,

COPD, etc,

Improved symptom (wheeze) control reduced

hospitalization healthcare savings

How (Current Clinical Methods)

Auscultation using a stethoscope

Based on history - patient recall or asthma diary or audio

or audio-visual recording by patients

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



22

Problem of Wheeze Monitoring

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



Auscultation with stethoscope is

subjective, not for long-duration

monitoring Next!

History of occurrence is important

but patients’ descriptions are often

subjective with limited accuracy

Wheezes often occur at night;

Intermittent wheezes are not

easily monitored at night

23

The System Needed

A screening system that records and analyzes wheezes

over an extended duration of time

A simple and robust system suitable for both clinical and

home use

A system that allows the patient to use at anywhere and

anytime - hence no skin contact, no external wirings that

restrict movements

Non-Skin Contact Wearable System

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



24

Wearable

Wheeze

Monitoring

System @ NTU

III

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



Overall

System

Design

26

Wearable wheeze detection and recording system with

Signal Analysis Tools for physicians

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

udeOverview of our Solutions

PI: A/Prof. W Ser (NTU) and A/Prof. Daniel YT Goh (NUH)

Funded by A*STAR & MOE

Wearable

System

Signal

Analysis

Tools

Signal

Analysis

Tools

28

Wearable System Design

Respiratory sounds (lung, tracheal) are recorded by sound

sensors (stethoscopes or air-conductance microphones)

Sensor outputs are band-passed, amplified, sampled, and

segmented.

A low-complexity algorithm is applied to detect presence

of wheezes

Segments with wheezes are recorded & their severity

estimated; respiratory & heart rates are estimated too

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



29

Signal Analysis Tools (SAT)

Data (wheezing sounds) recorded in wearable suite are

further processed by the SAT

The SAT has several functions

Signal manipulation

General signal processing

Respiratory signal analysis

The SAT can also be used for training purposes

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



Signal Manipulation

• Signal copy, cut, paste

• Signal amplification

• Labeling

• Filtering

• Segmentation

General Signal Processing

• Signal Energy Analysis

• Power Spectral Analysis

• Pitch Analysis

• Format Analysis

• Zero-Crossing Analysis

• Classification

Respiratory Signal Analysis

Respiratory parameters

• Amplitude & periodicity of wheeze, Wheeze

length in a breathing cycle, etc.

Physiological parameters

• Types of wheeze (monophonic or polyphonic,

inspiration or expiration, etc.), changes in

severity after treatment, Wheeze location, etc.

30

Problems and Challenges I

Sound based design sensitive to external audio noise

How sensors are placed on wearable suite important

Need filter and noise reduction algorithms

Wearable power and packaging constraint

Need to operate at low sampling rates

Need low complexity processing algorithms

Need low-power platform (processor)

Only small dry batteries allowed

No external wiring

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



31

Problems and Challenges II

Unsupervised wearing and monitoring

Need to design for ease of wearing

Need to use multiple sensors

Cannot have air-gap between sensors and body

Different sizes of wearers

Need flexible length of suite fastening

Need a few sizes

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



32

http://www.ericsfunnypictures.com/funnypics1.html

Wearable

Wheeze

Monitoring

System @ NTU

III

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



Wearable

Signal

Processing

Algorithms

33

Types of Wheeze Detection Algorithms

Time-domain methods (waveform analysis,

Kurtosis, zero-crossing rate …)

Frequency-domain spectral analysis methods

Time-frequency analysis (peak detection)

Pattern recognition approach

etc.

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



34

Spectral Analysis Methods

T. R. Fenton, H. Pasterkamp, A. Tal and V. Chernick, “Automated Spectral Characterization

of Wheezing in Asthmatic Children”, IEEE Trans. on Biomedical Engineering, vol. BME-32,

No. 1, pp50-55, 1985.

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



• f1 to f2: Wheeze frequencies

• Pave: Average power in f1 to f2

• Af: Scaling factor (= 15)

• Wheeze Detection Criteria:

• > f1 and > Af x Pave

• Results: (5 Asthmatic and 2

normal teenagers): < 2% for

false +ve and false -ve

35

Pattern Recognition Methods

Typically 3 stages: Feature extraction Dimensionality

reduction Pattern classification.

Commonly used Features:

MFCC, LPC, AR model, Wavelet coefficients, etc.

Dimensionality Reduction Algorithms:

SVD, PCA, etc.

Classification Methods:

k-NN Classifier, DTW, VQ, NN, HMM, etc.

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



37

Pattern Recognition Methods (Example 1)

Use Wavelet Packet Decomposition for feature extraction

Use Learning Vector Quantization for classification

L. Pesu, P. Helistö, E. Ademovic, J.-C. Pesquet, A. Saarinen and A.R.A. Sovijärvi,

“Classification of respiratory sounds based on wavelet packet decomposition and learning

vector quantization,” Technology and Health Care, vol. 6, pp. 65–74, Jun1998.

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



Results { Sensitivity = TP / (TP + FN); Positive predictivity = TP / (TP + FP) }

38


MFCC as Features: x1, x2, …, xD, D = 39

GMM based Classification:

• D-Variate Gaussian PDF:

• Weighted Sum of M Gaussian Densities:

• Decision:

C. Chien, H.-D. Wu, F.-C. Chong, and C.-I. Li, “Wheeze detection using cepstral analysis

and Gaussian mixture models,” Proc. of the 29nd IEEE EMBS, pp. 3168-3171, Lyon,

France, Aug. 2007.

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



39

1

( ) ( )M

i i

i

p x w p x

11

wheeze is present if 0arg max ( );

no wheeze if 0

T

t kk N

t

cc p x

c

1

12 2

( ) ( )1 ( ) exp{ }

2(2 )

T

i i ii D

i

x xp x

; ( ) ( )TE x E x x


Results: More than 90% accuracy in Wheeze detection

C. Chien, H.-D. Wu, F.-C. Chong, and C.-I. Li, “Wheeze detection using cepstral analysis

and Gaussian mixture models,” Proc. of the 29nd IEEE EMBS, pp. 3168-3171, Lyon,

France, Aug. 2007.

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



40


4 Time and Frequency Domain Features, x1 to x4

FDA (Fisher Discriminant Analysis) Classification

• Mean vector:

• Find w that maximizes

• Projection vector:

• Neyman Pearson Hypothesis: Compare yi with

S. Aydore, I. Sen, Y. P. Kahya and M. K. Mihcak, “Classification of respiratory signals by

linear analysis,” Proc. of the 31nd IEEE EMBS, pp. 2617-2620, Minnesota, USA, Sept.

2009.

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



41

T

i iy w x

1; 1,2j ix j

i

m x jn

1 2

Tw m m


Results: 93.5% success rate

S. Aydore, I. Sen, Y. P. Kahya and M. K. Mihcak, “Classification of respiratory signals by

linear analysis,” Proc. of the 31nd IEEE EMBS, pp. 2617-2620, Minnesota, USA, Sept.

2009.

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



42

Problems of Existing Algorithms

Not accurate enough or computationally

too expensive for wearable solutions

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



Our Approach

Review wheeze signal characteristics and

attempt to derive simpler but effective

methods that over-detect/record

43

Re-visiting Wheeze Signals

Energy concentrates at well below 1000 Hz

Contrasting energy distribution across frequency

Typically, wheeze occurs only in part of a breathing cycle

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



1 2 3 4 5 6 7 8

-0.5

0

0.5

Time (s)

Am

pli

tud

e

1 2 3 4 5 6 7 8

-0.5

0

0.5

Time (s)

am

plit

ude

Time (s)

Fre

qu

ency

(H

z)

1 2 3 4 5 6 7 8

0

500

1000

Time (s)

frequency (

Hz)

1 2 3 4 5 6 7 8

0

500

1000

Normal Breath Wheeze

Am

pli

tud

e F

req

uen

cy (

Hz)

44

Our Wheeze Detection Algorithm

Design goal: 1 or 2 parameters, low complexity, robust,

and suitable for wearable systems

Received signals:

• w(t): Wheeze signal

• n(t): Non-wheeze signals (e.g. breath sound, noise)

( ) ( ) ( )s t w t n t

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



45

Our Wheeze Detection Algorithm (Cont’d)

Short-Time Fourier Transform

• h(t) is the window function, τ is the time shift, and “*” is the

complex conjugate operator.

Peak Detection by Masking: For each STFT, determine moving

thresholds and use them to select “peaks” (dominant values)

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



* 2( , ) ( ) ( ) dj ftS f s t h t e t

L

n

a fnSL

fnS1

),(1

),(

0),(),(0

0),(),(),(),(),(

fnSfnSif

fnSfnSiffnSfnSfnS

a

aa

p

46


Entropy based Feature Extraction: Assuming there are N

dominant components, C1, C2, …, CN, in Sp(n, f), calculate

Entropy can be calculated as

Smoothed Entropy (by averaging filtering)

Features can be extracted as, for example,

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



1

, 1,2, ,nn N

nn

cp n N

c

1

log ( )N

n b n

n

E p p

M

t

ta EM

E1

1

)min()max(

)min(/)max(

aadiff

aaratio

EEE

EEE

47


Wheeze detection:

Set Tratio and Tdiff as the thresholds for Eratio and Ediff

Presence of wheeze can be decided by (for example)

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



ratioratio TE diffdiff TE and if Wheeze is present

ratioratio TE diffdiff TE and if No Wheeze

Otherwise Unable to determine

48

Our Wheeze Severity Estimation Algorithm

Eratio or Ediff are computed once every 3 seconds

Severity Score = Number of positive detections in 1 minute

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



49

Our Respiratory Rate Estimation Algorithm

Rate estimated from second peak of autocorrelation

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



50

Wearable

Wheeze

Monitoring

System @ NTU

III

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



Wearable

Sensors &

Platform

Suite

51

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude

Our Inward- & Outward- Hearing

Sensing Designs

Wearable

Suite

+

Outward-hearing Microphone Array

Inward-hearing Distributed Sensors

Medical Knowledge

Sound Sensors

52

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude

Wearable Wheeze Detection System

Version #1: Microphone Array, Sampling (1K, 12-bit)

DSP based System Speaker

Notebook for

Playing Sounds

TMS320C6713 DSP board

4-channels Microphone Array

(amplifier and filter included)

Experimental Set Up LED Indicators

53

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

udeWearable Wheeze Detection System

(Version #2: PDA based, 2-microphone design)

PDA based System

Microphones (head phone)

PDA (HP iPAQ hx2700 series)

Signal conditioning

NI CF6004 DAQ

15-pin Compact

Flash connector

Waveform display

Wheeze and

Snore Indicators

54

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

udeWearable Wheeze Acquisition System

(Version #3: Embedded Processor based System)

Notebook for experimental control Font View of Wearable Suite

Signal Conditioning

NI Daq

Signal conditioning

4-channel modified

stethoscopes or microphones

55

Sensors and Placements of Sensors

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



56

Inward-looking system

Stethoscope is modified by replacing long

acoustic tube with a microphone and wire

Microphone based system uses special acoustic

coupler design

4 distributed sensors: tracheal, left lung, heart,

and right lung

Outward-looking system

4-channel microphone array design

End-fire array configuration

Bandpass Filter, ADC, Platform

The bandpass filter is designed to have a bandwidth of up

to 1 KHz (without DC)

The ADC has a sampling rate of 2 KHz and a resolution of

10 bits

The platform uses one MSP430 chip for each channel (for

4 channels)

The selected sounds are stored in a micro SD card

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



57

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

udeData Processing & Recording

Used for Processing

1 Recording Cycle

(Wheeze Detection)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec) ADC 1

Data Block

(3 sec)

Used for Processing

(Wheeze Detection)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec) ADC 2

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec) ADC 3

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec)

Data Block

(3 sec) ADC 4

Data Block

(3 sec)

58

Lab Tests,

Pre-Clinical

Trials, Results

IV

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



59

Results of Wheeze Detection (Downloaded Data)

At SNR = 2dB, 80% of Sensitivity & Specificity have been

achieved (9 wheezy samples and 6 normal samples)

• Sensitivity = TP/(TP+FN) x 100%; Specificity = TN/(TN+FP) x 100%

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



-10 -8 -6 -4 -2 0 2 4 60

10

20

30

40

50

60

70

80

90

100

SNR (dB)

Dete

ction A

ccura

cy (

%)

true Positive

false Negative

true Negative

false Positive

1 1.5 2 2.5 3 3.50.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

2.2

2.4

Entr

opy D

iffe

rence

Entropy Ratio

normal

wheeze

Sensitivity

Specificity

60

Results of Wheeze Detection (Pre-Clinical Trial)

Performance: Sensitivity > 85%; Specificity > 70%

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



Data Collection at

National University

Hospital, Singapore

More than 80 samples

have been collected

61

Results of Severity Estimation Algorithm (Pre-Trial)

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



Samples Severity Scores Doctor’s Assessment

1 0 No Wheeze

2 10 Low

3 35 Yes

4 15 Low

5 10 Low

6 20 Low

7 5 Low

8 95 Yes

62

Results of Respiratory Rate Estimation (Pre-Trial)

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



63

10 15 20 25 30 35 40 45 5010

15

20

25

30

35

40

45

50

55

Actual RR (bpm)

Estim

ate

d R

R (

bpm

)

Conclusions

and

Future Work

V

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



64

Conclusions & Future Work

Wearable sound based sensing (+ PC signal analysis)

can be effective for automatic wheeze detection

Sensitivity higher than 85% and specificity above 70%

have been achieved using our algorithm/system for data

collected from some real patients (more data are needed

to validate the results)

Future work includes extensive clinical trials and studies

on other types of respiratory disorders

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



65

Some of our Research Publications

W Ser, JM Zhang, JF Yu, and TT Zhang, “A Wearable Multiple Audio-Sensor based

Design for Respiration Monitoring,” patent filed 2010.

W Ser, ZL Yu, JM Zhang, and JF Yu, “A Wearable System Design with Wheeze

Signal Detection,” 5th International Workshop on Wearable and Implantable Body

Sensor Networks, 2008.

W Ser, TT Zhang, JF Yu, and JM Zhang, “Detection of Wheeze by Distributed

Microphone Array Analysis of Breathe and Lung Sounds,” The 6th International

Workshop on Wearable and Implantable Body Sensor Networks, 2009.

JM Zhang, W Ser, JF Yu, and TT Zhang, “A Novel Wheeze Detection Method for

Wearable Monitoring Systems,” The International Symposium on Intelligent

Ubiquitous Computing and Education, 2009.

TT Zhang, W Ser, Daniel YT Goh, JM Zhang, JF Yu, C Chua, IM Louis, “Sound

Based Heart Rate Monitoring for Wearable Systems,” The 7th International Workshop

on Wearable and Implantable Body Sensor Networks, 2010.

Daniel YT Goh, W Ser, JM Zhang, JF Yu, C Chua, IM Louis, TT Zhang,

“Development of a Wearable System to Record and Analyse Respiratory Sounds:

Preliminary Data from Wheeze Analysis,” The 26th International Paediatric

Association Congress, 2010.

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



66

Some other References

T. R. Fenton, H. Pasterkamp, A. Tal and V. Chernick, “Automated Spectral

Characterization of Wheezing in Asthmatic Children”, IEEE Trans. on Biomedical

Engineering, vol. BME-32, No. 1, pp50-55, 1985.

Y. Shabtai-Musih, J.B. Grotberg and N. Gavriely, “Spectral content of forced

expiratory wheezes during air, He, and SF6 breathing in normal humans,” J. Appl.

Physiol., vol. 72, pp. 629–635, 1992.

R. Jané, D. Salvatella, J. A. Fiz, and J. Morera, “Spectral analysis of respiratory

sounds to assess bronchodilator effect in asthmatic patients,” Proc. of the 20th IEEE

EMBS, Hong Kong, 1998.

M. Waris1, P. Helistö1, S. Haltsonen1, A. Saarinen2, A.R.A. Sovijärvi, “A new method

for automatic wheeze detection,” Technology and Health Care, vol. 6, pp. 33-40, Jun

1998.

L. Pesu, P. Helistö, E. Ademovic, J.-C. Pesquet, A. Saarinen and A.R.A. Sovijärvi,

“Classification of respiratory sounds based on wavelet packet decomposition and

learning vector quantization,” Technology and Health Care, vol. 6, pp. 65–74,

Jun1998.

A. Homs-Corbera, R. Jane etc., “Algorithm for time–frequency detection and analysis

of wheezes,” IEEE EMBS, pp. 2977-2980, New York, USA, Sept. 2000.

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



67

Some other References (Cont’d)

S. Lehrer, “Understanding Lung Sounds,” third edition., W.B. Saunders Company,

Pennsylvania, USA, 2002.

M. Bahoura and C. Pelletier, “New parameters for respiratory sound classification,”

Proc. IEEE CCECE, pp. 1457-1460, Montreal, Canada, May 2003.

Y. P. Kahya, E. Bayatli ect. , “Comparison of different feature sets for respiratory

sound classifiers,” Proc. of the 25th IEEE EMBS, pp. 2853-2856, Cancun, Mexico,

Sept. 2003.

A. Kandaswamy, C. Sathish Kumar, Rin. PL Ramanathan, S. Jayaraman, and N.

Malmurugan, “Neural classification of lung sounds using wavelet coefficients,”

Computers in Biology and Medicine, vol. 34, pp. 523-537, 2004.

Patents and product information of KarmelSonic Products

LVQ-PAK Software, http://www.cis.hut.fi/research/som_lvq_pak.shtm, Department of

Information and Computer Science, Helsinki University of Technology, Finland.

School of Medicine, University of California at Davis, USA, Review of Lung Sounds.

[Online] http://medocs.ucdavis.edu/IMD/420C/sounds/lngsound.htm

0 2 4 6 8 10 12 14 16 18

x 104

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

samples

sig

nal am

plit

ude



68

http://www.cis.hut.fi/research/som_lvq_pak.shtm











http://medocs.ucdavis.edu/IMD/420C/sounds/lngsound.htm











Questions

or

Comments?

Thank You

69

wearable respiratory monitoring: algorithms and system …

Documents