automatic qrs complex detection algorithm designed for a novel electrocardiogram recording device
DESCRIPTION
Automatic QRS Complex Detection Algorithm Designed for a Novel Electrocardiogram Recording Device. Dorthe Bodholt Nielsen, Ph.D. student, DELTA / Technical University of Denmark Contact: [email protected]. Co-authors Kenneth Egstrup, OUH Svendborg Hospital Jens Branebjerg, DELTA - PowerPoint PPT PresentationTRANSCRIPT
Automatic QRS Complex Detection Algorithm Designed for a Novel Electrocardiogram Recording Device
Co-authorsKenneth Egstrup, OUH Svendborg HospitalJens Branebjerg, DELTAGunnar Bjarne Andersen, DELTAHelge B. D. Sørensen, Technical University of Denmark
Dorthe Bodholt Nielsen, Ph.D. student, DELTA / Technical University of DenmarkContact: [email protected]
Agenda
Application Example: Atrial Fibrillation
Advantages of our wireless ePatch technology
Algorithm: Automatic QRS complex detection
Detection Results
Conclusions and Future Work
The Heart and ECG Signals
Reference: http://elf.cs.pub.ro/pm/wiki/eestec/3
Atrial Fibrillation (AF)
Definition:Irregular and very fast activation of the atria
Irregular and fast pulse (ventricular contractions)
Prevalence:1 – 2 % of the general population
The prevalence increases with age:
5 – 15 % at the age of 80 years
Progression of disease:Paroxysmal → persistent → permanent
SymptomsPalpitations (“hjertebanken”)
Dyspnoea
No symptoms
Atrial Fibrillation
Adverse clinical eventsheart failure
Death rate is doubled
Risk of stroke is 5-fold compared to general population
Treatment of AFStroke prophylaxis with anticoagulation therapy
Importance of early detection of AFIt is very important to diagnose patients with AF early to start anticoagulation treatment and decrease stroke risk.
Asymptomatic patients: Screening for AF in the general population or high risk groups.
Paroxysmal AF: Very long term monitoring might be needed to find an episode of AF and diagnose the patient.
Advantages of the ePatch Heart Monitor
The ePatch heart monitor Traditional HOLTER monitor
http://flightphysical.com/Exam-Guide/CV/Holter-Monitor.htm
Automatic AF Detection
Embedded implementation of automatic signal processing algorithms for detection of cardiac arrhythmias, like atrial fibrillation.
Hardware implementation of automatic ECG arrhythmia
detection algorithms
Atrial Fibrillation in ECG Signals
Definition of AF in ECG signalsSurface ECG shows irregular RR intervals
Surface ECG shows no distinct P waves
The interval between two atrial activations is usually variable and <200ms
Example of AF recorded with the ePatch heart monitor:
Example of normal ECG recorded with the ePatch heart monitor:
Step I: Detection of Heart Beats
Automatic AF detection algorithms in the literature have three different approaches for automatic AF detection:
Detection based on the irregular RR intervals
Detection based on the absence of P-waves
Detection based on both irregular RR intervals and absence of P-waves
In order to apply either of these, it is necessary to design an automatic QRS complex detection algorithm.
Automatic QRS Complex Detection
Schematic illustration of the designed automatic QRS complex detection algorithm:
Automatic QRS Complex DetectionRaw ECG, Lead I
Feature I, Lead I
Adaptive thresholding, Feature I, Lead I
Binary feature signal, Feature I, Lead I
Final QRS position
Databases
The ePatch database:30 minute records from 11 different patients
Manual annotation of more than 22,000 heart beats
The MIT-BIH Arrhythmia Database (standard database)30 minute records from 48 different patients
Manual annotation of more than 91,000 heart beats
QRS Detection Results – ePatch database
Performance measures:Sensitivity = TP/(TP + FN)
Positive predictivity = TP/(TP + FP)
QRS detection performance:
All abnormal beats were correctly detected by the algorithm
# of patients Sensitivity Positive predictivity11 99.57 % 99.57 %
9 99.95 % 99.92 %
QRS Detection Results – Standard Database
Detection results compared to other studies using a 2 channel approach to automatic QRS complex detection:
Study Sensitivity Positive predictivityThis work 99.63 % 99.63 %
Ghaffari et al. 99.94 % 99.91 %
Boqiang et al. 99.91% 99.93 %
Chiarugi et al. 99.76 % 99.81 %
Conclusions and Future Work
Promising performance:The algorithm should, of course, be evaluated on a larger ePatch database
This algorithm might be applied to initiate different arrhythmia detection algorithms that rely on the detection of heart beats.
Our current work is to design new algorithms for automatic detection of critical heart arrhythmias, like atrial fibrillation.