Senior Capstone Design Project Real-Time Electrocardiogram Spectral Analyzer Ryan Carnathan, Keith Berry, Mark Stadick, & Greg Michaelson Electrical &

BIOMEDICAL SIGNAL PROCESSING LAB

b s p . p d x . e d u

PORTLANDNIVERSITYU

STATE ELECTRICAL & COMPUTERENGINEERING

Senior Capstone Design Project Real-Time Electrocardiogram Spectral Analyzer

Ryan Carnathan, Keith Berry,

Mark Stadick, & Greg MichaelsonElectrical & Computer Engineering



PORTLANDNIVERSITYU


Presentation Overview

• Introduction

• Hardware

• QRS Detector

• Spectral Analysis

• User Interface

• Summary and Conclusion

• Demonstration



PORTLANDNIVERSITYU


Requirements

• Apply knowledge

• Solve Real World Problems

• Work in Multidisciplinary Teams

• To become actively involved in a company or community organization.



PORTLANDNIVERSITYU


Purpose

• Used in conjunction with the HP Merlin bedside monitor

• Real-time spectral analysis of the beat-to-beat variability

• Assess the potential for:– Improved bedside clinical monitoring – Diagnostic and predictive information



PORTLANDNIVERSITYU


Function

• Graphical interface for the display and analysis of:– ECG Waveforms– QRS Complex Detection– Heart Rate Variability (HRV) Spectrogram

• These functions are performed and displayed in real-time.



PORTLANDNIVERSITYU


Implementation

• Hardware

• QRS Detection


• Integration and User Interface



PORTLANDNIVERSITYU


Definition of terms

• Electrocardiogram (ECG)

• Heart Rate Variability (HRV)

• QRS Complex

• RR v.s. NN

• The ECG Waveform

• The QRS Complex



PORTLANDNIVERSITYU



• Introduction

• Hardware

• QRS Detector


• User Interface


• Demonstration



PORTLANDNIVERSITYU


Hardware

• Data Acquisition Card (DAQ)

• Laptop Computer

• Accessories



PORTLANDNIVERSITYU


Data Acquisition Card

• Capabilities– Sample Rate

500 kS/s

– Number of Channels8 Differential16 Single Ended

– Gains0.5, 1, 2, 5, 10, 20, 50,

100

– Voltage Levels±10 V - ±50 mV

• Configuration– Sample Rate

1 kS/s

– Number of Channels1 Differential

– Gain Setting1

– Voltage Levels±5 V



PORTLANDNIVERSITYU


Laptop Computer

• CPU– Intel 600 MHz

Pentium

– 256 k Level 2 Cache

• Storage– 20 GB Hard Drive

– 3 ½ inch Floppy Drive

• Memory– 512 MB of Ram

• Peripherals– Infrared Transfer

– 10/100 Network Card

– 56k Modem

– (2) PCMCIA Slots1 Taken by DAQ

• Operating System– Windows 2000 SP 1



PORTLANDNIVERSITYU


Laptop Configuration

• CPU– 600 MHz

– 500 MHz (Battery)

• Hard Drive– 15.4 GB Free

• Ram Used– 128 MB (System)

– 384 MB Available

• DAQ Input– 1 Differential Channel

– 1 kHz Sample Rate

– BNC Connection



PORTLANDNIVERSITYU


Accessory Box

• Component Box• Connector Block• BNC Connection• Ribbon Cable• Strain Relief Device



PORTLANDNIVERSITYU


Connections

• Merlin To DAQ– 6.35 mm Audio Cable

To BNC

• DAQ to Laptop– 68 Pin Ribbon Cable

(0.2 M)



PORTLANDNIVERSITYU


Software Integration

• Program Functions– Initialize DAQ

– Start Acquisition

– Check Buffer

– Transfer Buffer

– Stop Acquisition

• Software– LabWindows CVI

– MatLab 6.0



PORTLANDNIVERSITYU


Acquisition Sample



PORTLANDNIVERSITYU


Battery Efficiency

• CPU– Voltage = 1.35V (Full Power)

Average Power < 2.0 Watts

– Voltage = 1.1V (Battery) Average Power < 1.0 Watts

• Battery Life– 3:15 per Battery

2 Batteries Possible

• DAQ Requirements– ±5 VDC (±5%)

280 mA typ400 mA max70 mA unused



PORTLANDNIVERSITYU



• Introduction

• Hardware

• QRS Detector


• User Interface


• Demonstration



PORTLANDNIVERSITYU


QRS Detection Overview

• QRS detection algorithm

• Current Status

• Future work

• Summary



PORTLANDNIVERSITYU


QRS Detector Algorithm

• Block Diagram

Remove Offset Filter Clip

QRS Peak Detection

PeakAcceptance/Rejection

Output QRSp-p Interval



PORTLANDNIVERSITYU


QRS Detector Algorithm cont.

• ECG Input Data

• Remove Offset

• Filter

• Clip

• QRS Peak Detection

• Output QRS N-N Interval



PORTLANDNIVERSITYU


QRS Detector Sample Output

0 500 1000 1500 2000

-2

0

2

4

6

8

10Raw ECG

0 500 1000 1500 2000

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

Scaled ECG

0 500 1000 1500 2000

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

Clipped ECG

0 500 1000 1500 2000

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

Detected QRS w/ Scaled ECG



PORTLANDNIVERSITYU


Current Status

• Operational– Remove d.c. offset– Clip below amplitude threshold– Detect QRS complexes (clean signals only)

• Pending – Integration with graphical user interface– Digital Filter– QRS Accept/Reject function



PORTLANDNIVERSITYU


Future Work

• 1st Priority– Integration with graphical user interface– Digital Filter– QRS Accept/Reject function– Final report outline– Testing and debugging



PORTLANDNIVERSITYU


Future Work Continued

• 2nd Priority– Adaptive amplitude threshold function– Final report draft– Speed and performance improvements

• 3rd Priority– Final report– Final presentation



PORTLANDNIVERSITYU



• Introduction

• Hardware

• QRS Detector


• User Interface


• Demonstration



PORTLANDNIVERSITYU


• Heart Rate Variability (HRV)– Inter Beat Intervals (IBI)

variation in the time intervals between beatsassumes irregular beats have been removednormal to normal (NN) beats

– Instantaneous Heart Rate (IHR)variation in consecutive instantaneous heart rates inverse of IBI

Spectral Analysis



PORTLANDNIVERSITYU


• Power Spectral Density (PSD)– Distribution of Variance Over Frequency

• Three Operations– Resample– Filter– Generate Spectrogram

Fast Fourier Transform (FFT)Transforms Time Signal to Frequency

Spectral Analysis



PORTLANDNIVERSITYU


Resample

Filter

FFT

QRSDetector

GUIPlot

Block Diagram



PORTLANDNIVERSITYU


• Elapsed Time vs Number of Heart Beats

QRS Output



PORTLANDNIVERSITYU


• Re-Interpreted Data From QRS Detector– NN Intervals – Indexed by Heart Beats

NN Intervals



PORTLANDNIVERSITYU


• Interpolate– Generate Continuous Curve– Passes Through Existing Points

• Sample New Series– Constant Time Period– Required for Fast Fourier Transform (FFT)– NN Intervals vs Time

Resampling Overview



PORTLANDNIVERSITYU


Resampling

– Input Signal from QRS Detector:

– Resampled Signal:



PORTLANDNIVERSITYU


Filtering

• Spikes in Signal – Missed Beats by QRS Detector– Distort Valid Data

• Filter– Flattens Large Excursions– Modifies Only Large Outlying Data Points



PORTLANDNIVERSITYU


Filtering

– Resampled Signal:

– Filtered Signal:



PORTLANDNIVERSITYU


• Estimating Spectral Power– Remove Mean Value– Perform FFT – Power Calculation - |FFT|2

Spectrogram Algorithm



PORTLANDNIVERSITYU


• PSD vs Time

Spectrogram Example



PORTLANDNIVERSITYU



• Introduction

• Hardware

• QRS Detector


• User Interface


• Demonstration



PORTLANDNIVERSITYU


Features

• Implemented– Plots

ECG

– User ControlsStart/StopWindow SizeScrolling

– ModesHistoricalReal Time

• Planned– Plots

Beat to BeatSpectrogramScaleogram

– User ControlsGraph SelectionNotes



PORTLANDNIVERSITYU


Process Budget

• Limits– Data Acquisition

10-15% of CPULittle Memory

– Beat Detection5-10% of CPULots of Memory

– Spectrogram Calculation30% of CPU



PORTLANDNIVERSITYU



• Introduction

• Hardware

• QRS Detector


• User Interface


• Demonstration



PORTLANDNIVERSITYU


Completed

• Hardware– Specified Needs– Purchased / Acquired– Built– Software Completed– Tested on Merlin

• GUI– 2 Plots– Scrolling– Integration of Hardware– Real-time ECG Display– Historical Mode

• QRS Detector– Literature Review

– Scaling

– Clipping

– Beat Detection

• Spectrogram– Resampling

– Median Filter

– Spectral Estimation



PORTLANDNIVERSITYU


Remaining Tasks

• GUI– Spectrogram

Integration

– QRS Detector Integration

– User Notepad

• QRS Detector– Filtering

– Ectopic Beat Rejection

• Spectral Analysis– Filter Refinement

– Higher Resolution Spectrogram Plot



PORTLANDNIVERSITYU



• Introduction

• Hardware

• QRS Detector


• User Interface


• Demonstration

Senior Capstone Design Project Real-Time Electrocardiogram Spectral Analyzer Ryan Carnathan, Keith Berry, Mark Stadick, & Greg Michaelson Electrical &

Documents

qrs complex slide

user interface slide

acquisition sample slide

conclusion demonstration

ibi spectral analysis

ma unused slide

predictive information

qrs detector nn intervals