An Adaptive Fuzzy ECG Classifier Presented by: Lei W ai Kei Dept. - - PowerPoint PPT Presentation

An Adaptive Fuzzy ECG Classifier

Presented by: Lei W ai Kei

• Dept. Electronic & Electrical Eng,

Faculty of Science & Technology, University of Macau

Outlines

Introduction ECG Signal and Patterns A Fuzzy ECG Classifier An Adaptive Fuzzy ECG Classifier Demonstrations Conclusions

Introduction

ECG signals reflect the health status of

heart by completely recording the subtle cardiovascular circulation.

Automatic ECG analysis is essential to

implement ECG monitoring in Home Healthcare.

Here an Adaptive Fuzzy ECG Classifier

(AFC-ECG) is proposed to integrate the knowledge from medical expertise and by statistical analysis as well.

ECG Cardiology

P:

Contraction of Atrium

QRS: Contraction of Ventricle T:

Relaxation of Heart

P Q R S T

Left Atrium Right Atrium Left Ventricle Right Ventricle

Heart ECG signal

Characteristic ECG Features

Characteristic Features: P Q R S T

ST Segment PR Segment QRS Interval

ECG Features ECG Features P Peak Value R Peak Value T Peak Value Prior Heart Rate Post Heart Rate PR Segment

ECG Classification

ECG Classifier Structure

Extracting ECG Features Intelligent Inference Machine Classification of ECG Beats ECG Signal Medical Knowledge ECG Parameterization Inference Result

Methods for ECG Classifiers

Fuzzy Inference Networks (FIN)

Performs imprecise and interpretable inference. But without self-adaptation mechanism.

Neural Networks (NN)

Capable of self-adaptation inference but blind to medical experts.

Adaptive FIN

Not only performs imprecise, interpretable inference like FIN, but also improves the classifier performance by statistics-based self-adaptation.

Input Result

Membership Function Rule Set

Input Result

Weights

FIN ECG Classifiers

Fuzzification

Membership Grading

Classification

Rule-based fuzzy Inference

Results

Over 10 kinds of heart beats are defined, 4 of them are chosen for testing:

(1) Normal Beat (N) (2) Premature Atrial Contraction (PAC) (3) Left Bundle Block Beat (LBBB) (4) Right Bundle Block Beat (RBBB) ECG Features Fuzzification Knowledge Base Classification Result

Fuzzification

Calculating the Membership Grades of

the input ECG Features.

Obtaining the Linguistic Variables

ECG Features P Peak Value R Peak Value T Peak Value RR0 RR1 PR Segment Fuzzification Linguistic Variable Disappear, Early… Upward, Downward… Upward, Downward… Short, Normal… Short, Normal… Reference Line

Applied Membership Functions for Fuzzification

S-Function Z-Function

Similar to S-Function

Gaussian Function

(RI: means Relationship Index)

( ) ( ) ( ) ( ) ( ) ( )

⎪ ⎪ ⎩ ⎪ ⎪ ⎨ ⎧ ≤ = ≤ < + − − ⋅ − = + ≤ < − − ⋅ = ≤ = x b RI b x b a a b a x RI b a x a a b a x RI a x RI 1 2 / 2 1 2 / 2

2 2 2 2

( )

] 2 exp[

2 2

b a x RI − − =

Type Type a b S, Z Lower Boundary Upper Boundary G Mean Value Standard Deviation

Illustration of Fuzzification

Membership Functions of “Normal Beat”

Linguistic Variable ECG Feature Function Type Parameter a Parameter b P upward P Peak Value S 0.10mV 0.15mV QRS upward R Peak Value S 0.70mV 0.80mV T upward T Peak Value S 0.10mV 0.15mV RR0 normal Prior-HR Gaussian 80bpm 20bpm RR1 normal Post-HR Gaussian 80bpm 20bpm

Rule Template

• f Fuzzy Classification

IF “ “Feature 1 Feature 1” ” is “Linguistic Variable 1” & “ “Feature 2 Feature 2” ” is “Linguistic Variable 2” & … “ “Feature N Feature N” ” is “Linguistic Variable N” THEN “ “Beat Type Beat Type” ” is “Class Name”

Beat Type

• No. of Rules
• No. of Hypotheses

Normal 1 5 RBBB 2 8 LBBB 2 8 PAC 2 8

Illustration of Fuzzy Classification

Rule for Normal Beat (N)

IF “P is Upward” & “QRS is Upward” & “T is Downward”& “RR0 is Normal” & “RR1 is Normal” THEN Type is “Normal”

Problems in FIN ECG Classifiers

Variations in ECG Signals

An illustration of variation occurs in “Normal Beat”

Problems in FIN ECG Classifiers

Hard to define membership functions

Linguistic Variable ECG Feature Function Type Parameter a Parameter b P upward P Peak Value S 0.10mV 0.15mV 0.80mV 0.15mV 20bpm 20bpm QRS upward R Peak Value S 0.70mV T upward T Peak Value S 0.10mV RR0 normal Prior-HR Gaussian 80bpm RR1 normal Post-HR Gaussian 80bpm

Adaptive Fuzzy ECG Classifier

Features: Features:

Population Estimation of incoming signals:

Obtaining the statistical parameters of incoming signals, e.g., mean values and standard deviations

Modifying membership functions:

Base on above statistical parameters, adapting membership functions to current ECG Signals current ECG Signals.

Rationales: Rationales:

Physiological:

Incoming ECG signal rarely numerical match with the medical

• knowledge. The capacity of adaptation is necessary to deal with the

variance variance which occurs in input.

Mathematical:

Statistics is pertaining to the analysis, interpretation and presentation of data. It provides a way to draw inferences draw inferences about the population of incoming signal.

Statistical analysis for Population Estimation

Two parameters are used to estimate the population:

• Mean Value

Mean Value

• Standard Deviation (SD)

Standard Deviation (SD)

( )

N x N x

N i i N i i

/ /

1 2 1

⎟ ⎠ ⎞ ⎜ ⎝ ⎛ − = ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ =

∑ ∑

= =

μ σ μ

% 99 . 99 4 % 73 . 99 3 % 45 . 95 2 % 27 . 68 σ σ σ σ

SD Samples

Self-adaptation of Membership Function

According to statistic analysis of ECG signals, about 95% of the values are within 2 standard deviation.

S, Z-function Gaussian

σ μ σ μ σ μ 2 2 2 = = + = − = b a b a

σ μ 2 − σ μ 2 +

AFC-ECG Structure

Enhance the conventional FIN ECG

classifiers by statistical learning methods.

ECG Features Fuzzification Adjust? Knowledge Base Classification Adaptation Results Yes No

Adaptive Process

The first 10 minutes of ECG signals are taken advantage for self-adaptation:

Population calculation by Statistical Analysis

• Modifying the parameters of membership

function

ECG Features Fuzzification Adjust? Knowledge Base Adaptation Yes Learning set Normal LBBB RBBB PAC

σ μ,

Modification by ( )

Classification Process

The remaining 20 minutes of ECG signals

are used for further classification.

ECG Features Fuzzification Adjust? Knowledge Base Classification Result No Test Beats Beat Type Normal LBBB RBBB PAC

Experiment Results

Record Type Annotation Before Learning After Learning 106

Normal

1507 1208 80.2% 1258 83.5%

PAC

7

• 1
• RBBB

30

• 27
• LBBB
• 1
• Uncertain

262

• 220
• 118

Normal

1

• PAC

96 49 51.0% 72 75.0%

RBBB

2166 1602 74.0% 1831 84.5%

LBBB

• Uncertain

610

• 359
• 207

Normal

19

• 2
• PAC

107 127 81.3% 105 98.1%

RBBB

86 71 82.6% 80 93.0%

LBBB

1457 1357 93.1% 1389 95.3%

Uncertain

76

• 74

Experiment Analysis & Discussion

• Accuracy

Accuracy:

The average accuracy of FIN ECG classifier:

77.0%

The average accuracy of AFC-ECG: 88.2%

Worst case is in the Record “118”

because there are no enough learning

• samples. Even after self-adaptation, its

accuracy is still low to 75.0%.

Significances

The proposed adaptive fuzzy ECG classifier

integrates the advantages of Fuzzy Logics in human knowledge expression and statistical learning method for self-adaptation.

Advantages:

Eliminating the effects of physiological signal variation; Refining the parameters of fuzzy rule by self-adaptation;

Disadvantages:

Dependent on initial fuzzy rules; Invalid to unknown ECG signals.

Future work

Increasing the types of classifiable ECG beats. Extending to Fuzzy Neural Networks