real-time electrocardiogram may 2 1...

Real-time Electrocardiogram Monitoring

By: Nicholas Clark, Edward Sandor, and Calvin WaldenAdvised By: Drs. Yufeng Lu and In Soo Ahn

May 2nd, 2017

1

Presentation Outline

1. Motivation2. Previous Work3. Objectives and Constraints4. System Design and Implementation5. Results and Discussion6. Project Management7. Conclusion and Future Work8. References

2

1 in 4deaths in the US stem from a form of heart

disease[1].

3

Motivation 4

25% of the USWill experience an arrhythmia over the age of 50[2]

Figure 1 - Visual representation of Americans affected by arrhythmia.

Motivation 5

Figure 2 - Percentage of annual diagnosis of the major arrhythmias[3].

● Arrhythmias are common heart conditions where the heart does not properly pump blood[4].

● An electrocardiogram (ECG) is used to identify arrhythmias by detecting the electrical signals of the heart[5].

● Holter monitor is the current standard of care, which is a portable ECG monitor that records patient heart data for later review by a physician.

6

Figure 3 - Example of a Holter monitor[6]

Motivation

● Holter monitors only record patient data, requiring a physician to review it for arrhythmias after it has been collected.

● There is currently no standard of care that notifies a doctor of arrhythmia in real-time.

● These monitors also cannot be controlled or configured remotely.

7Motivation

Figure 4 - Sample ECG with PVC[7]

PVC



8

Previous Work

● Overview● Implementation● Results● Recommendations

9

Overview

Real-time Heart Monitoring and ECG Signal Processing[8]

● Completed by Bamarouf, Crandell, and Tsuyuki in Spring 2016.

● Looked into the identification of arrhythmias, specifically premature ventricular contraction (PVC).

● Aimed to improve upon the Holter monitor, a current standard of care.

● Implemented the Pan-Tompkins and Template Matching algorithms.

10

Implementation

● Designed software optimized for the TI CC3200 MCU.

● Transmitted Short Message Service (SMS) packets over WiFi.

11

Figure 5 - TI CC3200 IoT enabled MCU used in this project[9]

Results

● This project could process benchmark ECG data from the MIT-BIH database and identify instances of PVC in real-time.

● The system successfully transmitted plots of detected PVC via SMS.

12

Results (cont’d) 13

Table 1 - Performance of algorithms with sample data

Figure 6 - Plot of PVC transmitted via SMS.

Recommendations

● Interface the system with ECG electrodes to acquire real patient data.

● Improve the signal processing algorithms used in the system.

14



15

Objectives and Constraints

● Develop a portable, mobile device with ECG sensors.

● Implement an embedded system with ECG algorithms to monitor ECG signals in real-time.

● Wirelessly notify a wearer’s doctor of PVC.

16

Constraints: portable, ergonomic design; limited device memory; WiFi connectivity;

battery life.



17

System Design and Implementation

● System Design● Digital Signal Processor● Arrhythmia Detection Algorithms● ECG Electrodes and Pre-filter Board● Embedded Computer● User Interface● Printed Circuit Board

18

System Design 19

Figure 7 - System Block Diagram

System Design 20


System Design 21


System Design 22


System Design 23


System Design

● Digital Signal Processor○ Acquire ECG signal○ Perform arrhythmia detection algorithm

● System Controller○ Manage user interface○ Log data○ Distribute power

● Electrodes and ECG Prefilter● User Interface● Printed Circuit Board

○ Interface subsystems

24

Figure 8 - Real-Time ECG System

System Design 25

Figure 9 - Real-Time ECG System

LCD and Push Buttons ECG Pre-filter Board ECG Electrodes

DSP

System ControllerInterfacing

PCB

Power Supply

Digital Signal Processor

● TI TMS320C5515 eZdsp○ TI low-power DSP family○ 120 MHz fixed-point DSP○ 64 KB of DARAM○ 256 KB of SARAM○ 128 KB of ROM

● Perform Arrhythmia Detection Algorithm● Acquire ECG Signal

○ 10-bit Analog Digital Converter (ADC)○ Sampling rate: 360 sample/s

● Communicate with System Controller○ Via Universal Asynchronous

Receiver/Transmitter (UART) with baud rate of 57600

26

Figure 10 - TMS320C5515 eZdsp

Digital Signal Processor 27

Figure 11 - Flowchart of DSP main loop


Figure 12 - Flowchart of acquisition timer interrupt


Figure 13 - Flowchart of ADC conversion complete interrupt


Figure 14 - Flowchart of UART data received interrupt

Arrhythmia Detection Algorithm

● Two-Stage Algorithm○ Pan-Tompkins algorithm

■ Identify individual heart beats○ Template-matching algorithm

■ Correlate sample heartbeat with healthy heartbeat template to identify arrhythmia

● Integration of Algorithm Code○ Originally written by previous group○ Extracted algorithm to be standalone○ Modified to be platform independent

■ Code can be compiled for DSP or for MATLAB file for validation○ Reduce extraneous buffers to improve memory usage

31

Pan-Tompkins Algorithm 32

Figure 15 - Flowchart of Pan-Tompkins algorithm

Band Pass Filter: 5-11 Hz

Template Matching Algorithm 33

Figure 16 - Flowchart of template matching algorithm

Electrodes and ECG Pre-filter Board

● 3-Lead ECG● AD8232-based breakout

board● High-pass and low-pass

filters with adjustable cutoffs

● Instrument amplifier

34

Figure 17 - Pre-filter board and ECG electrodes

Electrodes and ECG Pre-filter Board

Electrode Placement

35

Figure 18 - Reference diagram for electrode placement[10]

Figure 19 - ECG setup used for system development, with 3 leads

System Controller

Raspberry Pi 3 Model B

● Linux-based Raspbian 8● CPU: 1.2GHz Quad-Core ARM Cortex● RAM: 1GB LPDDR2● SD Card Slot● 27 digital GPIO pins, including UART and I2C● 10/100 BaseT Ethernet● 802.11 b/g/n WLAN

36

System Controller

Acquisition Daemon

● Receives processed or raw ECG data over serial port● Logs data into files on SD card● Signals Communication Daemon if arrhythmia was detected by DSP

37

System Controller

Acquisition Daemon

38

Figure 20 - Flowchart of acquisition daemon

System Controller

Acquisition Daemon

39


System Controller

Acquisition Daemon

40


System Controller

Acquisition Daemon

41


System Controller

Acquisition Daemon

42


System Controller

Communication Daemon

● Uses GMail account to send messages● Startup notification message with system information● Optional ECG report messages sent periodically● Urgent ECG report message sent when signalled by Acquisition

Daemon

43

System Controller


44

Figure 21 - Flowchart of communication daemon

System Controller


45


System Controller


46


System Controller


47


System Controller

Software Design

● Software written in C● Developed packet-based serial communication protocol● libconfig - common configuration file parsing● wiringPi - direct control of GPIO pins● gnuplot - generate plots of ECG data for email

48

User Interface

● Adafruit 16x2 LCD with Keypad○ 16x2 character LCD○ HD44780 LCD Controller○ 5 pushbutton inputs○ MCP23017 I2C I/O Expander

● Lighttpd Web Server○ Light-weight web server on system controller○ Serves PHP5 web pages for user control,

configuration, and information

49

Figure 22 - 16x2 LCD with keypad.

User Interface 50

Figure 23 - Flowchart for LCD process

User Interface

Software Design

● LCD software written in C● Lighttpd web interface written in HTML, PHP5, and CSS● Server and LCD process use named pipes to communicate LCD

configuration options.● LCD process communicates with the LCD controller via I2C through

the GPIO pins.

51

Printed Circuit Board

● Interface between prefilter board, DSP, and Raspberry Pi 3● Raspberry Pi "Hat" formfactor● Additional debug and test points for future work

52

Figure 24 - Screenshot of PCB design Figure 25 - Schematic of interface circuit

Printed Circuit Board 53

Figure 26 - Left: Top of PCB. Right: Bottom of PCB.



54

Results and Discussion

Digital Signal Processor

55

Figure 27 - DSP activity viewed on oscilloscope

Serial Communication120ms

Algorithm Execution21ms

ECG Signal from Pre-filter Board


Digital Signal Processor Acquisition

56

Figure 28 - ECG signal acquired by DSP



57

Figure 29 - Band-pass (5-11 Hz) filtering of ECG signal performed by DSP

LPF: y(n) = 2y(n−1)−y(n−2)+x(n)−2x(n−6)+x(n−12)HPF: y(n) = y(n-1)−x(n)/32+x(n−16)−x(n−17)+x(n−32)/32



58

Figure 30 - QRS peak detection by Pan-Tompkins algorithm



59

Figure 31 - Template of QRS peak used by template matching algorithm



60

Figure 32 - RR interval template used by template matching algorithm


Arrhythmia Detection Algorithm Performance

61

Table 2 - Performance of algorithms with sample data


System Controller

62

Figure 33 - ECG report message sample


System Controller

● Load average: 0.30 to 0.50○ Quad-core system: Means running at 7.5% to 12.5% capacity○ Acquisition Daemon CPU usage: 5% to 9%○ Other software components negligible

● Memory usage: ~200MB used of 923 MB, 20% used

63


User Interface

● Displays menu items including:○ Project Title○ Patient Name○ Web Configuration Address○ LCD Color○ Power Options

● Buttons allow user to cycle through menu

64

Figure 34 - LCD display showing menu item

Results and Discussion 65

Figure 35 - Color menu option Figure 36 - Power menu option

Results and Discussion 66

Figure 37 - Web configuration page served using Lighttpd.


Power Consumption Testing

● Based on Discharge Time of Battery with 5V Regulated Output● Measured Battery Capacity Approximately 14.6 Wh

○ Timed discharge through 50 ohm power resistor for reference

● Battery Runs System for 8.7 hours ● Approximate Power Consumption is 1.7 W● Raspberry Pi 3 reported to use approximately 1.3W[11]

67



68

Division of Labor

● DSP and algorithm evaluation (Ed)

● System Controller acquisition and communication software (Calvin)

● Printed circuit board design (Calvin)

● PCB review (Calvin, Ed, Nick)

● System Controller user interface software (Nick)

● Poster, presentation, and reports (Calvin, Ed, Nick)

● Documentation (Calvin, Ed)

● ECG electrode testing (Calvin, Ed)

69

*Bolded name denotes primary contributor.

Schedule 70

Week of Tasks

1/16/17 ● Implement initial Pan-Tompkins algorithm on choice platform● Begin interfacing board for embedded computer● Interface embedded computer with DSP

1/23/17 ● Refine Pan-Tompkins algorithm, add Template Matching algorithm● Trial wireless communication and SMS

1/30/17 ● Trial ECG data from MIT-BIH database● Refine wireless communication and SMS

2/6/17 ● Refine Pan-Tompkins and Template Matching algorithms● Trial ECG data from MIT-BIH database● Begin embedded computer serial communication

2/13/17 ● Begin real-time ECG testing● Add LCD and pushbutton interface to interfacing board● Begin UI development

2/20/17 ● Refine Pan-Tompkins and Template Matching algorithms● Continue real-time ECG testing● Continue UI development

2/27/17 ● Complete real-time ECG testing● Complete UI development

3/6/17 ● Progress Evaluation● Continue system testing and tuning

Spring Break ● Continue system testing and tuning

Week of Tasks

3/20/17 ● Complete all lab work● Continue written deliverables● Continue system testing and tuning

3/27/17 ● Continue written deliverables● Update Student Scholarship Expo poster● Continue system testing and tuning

4/3/17 ● Finalize final report draft● Continue system testing and tuning

4/10/17 ● Final Report Draft (4/10)● Finalize draft of presentation slides● Finalize Student Scholarship Expo poster (4/10)● Event: Student Scholarship Expo (4/11)● Oral Presentation Preparation (4/13)● Continue system testing and tuning

4/17/17 ● Complete all written deliverables● Oral Presentation Preparation (4/18)● Continue system testing and tuning

4/24/17 ● Event: Project demonstration● Event: Poster Presentation (4/28)● Complete all lab work

5/1/17 ● Event: Presentation of project (5/2)● Project Website Verification (5/2)● Complete all written deliverables



71

Conclusions

● Project's goals were achieved:○ Develop a portable, mobile device with ECG sensors.○ Implement an embedded system with ECG algorithms to monitor ECG signals in

real-time.

○ Wirelessly notify a wearer’s doctor of PVC.

● Runs arrhythmia detection algorithm on data acquired from sensors in real-time

● Logs data and sends it to configured email addresses

72

Future Work

● Integrate more components onto PCB

● Reduce power consumption

● Implement the design utilizing only the Raspberry Pi 3 or comparable custom board

● Add new algorithm to detect additional types of arrhythmias

● Make transmission and storage of patient data secure

73



74

References

[1] “Heart Disease Facts & Statistics | cdc.gov.” [Online]. Available: https://www.cdc.gov/heartdisease/facts.htm.[2] “Who Is at Risk for an Arrhythmia? - NHLBI, NIH.” [Online]. Available: https://www.nhlbi.nih.gov/health/health-topics/topics/arr/atrisk.[3] “About Arrhythmia.” [Online]. Available: http://www.heart.org/HEARTORG/Conditions/Arrhythmia/AboutArrhythmia/About-Arrhythmia_UCM_002010_Article.jsp.[4] M. AlGhatrif and J. Lindsay, “A brief review: history to understand fundamentals of electrocardiography,” J Community Hosp Intern Med Perspect, vol. 2, no. 1, Apr. 2012.[5] J. A. Z. Justo, R. A. G. Calleja, and A. M. Diosdado, “Acquisition software development for monitor Holter prototype signals and its use for pre-diagnosis of cardiac damage based on nonlinear dynamic techniques,” in AIP Conference Proceedings, 2016, vol. 1747, p. 90001.[6] “acardio20140402v0005.jpg.” [Online]. Available: https://api.kramesstaywell.com/Content/ebd5aa86-5c85-4a95-a92a-a524015ce556/medical-illustrations/Images/acardio20140402v0005.jpg.[7]“pvc_21349464440513.jpg (1425×537).” [Online]. Available: https://classconnection.s3.amazonaws.com/833/flashcards/1119833/jpg/pvc_21349464440513.jpg.[8] F. Bamarouf, C. Crandell, and S. Tsuyuki, “Real-time heart monitoring and ECG signal processing,” Bradley University, May 2016.[9] “ee0e66c4-ff50-42e2-85b5-cdf9f64d6208.jpg.” [Online]. Available: http://www.multivu.com/players/English/70647514-add-wi-fi-to-anything-with-ti-s-internet-on-a-chip-new-simplelink/gallery/image/ee0e66c4-ff50-42e2-85b5-cdf9f64d6208.jpg.[10] “tmp14285_thumb1.jpg (372×480).” [Online]. Available: http://what-when-how.com/wp-content/uploads/2012/04/tmp14285_thumb1.jpg.[11] "Power Consumption," Raspberry Pi Dramble, [Online]. Available: https://www.pidramble.com/wiki/benchmarks/power-consumption.

75

Real-time Electrocardiogram Monitoring

Questions?

76

Discussion Points 77

Section Subsection Pages

Motivation

Statistics 3, 4, 5

Background 6, 7

Previous Work

Overview 10

Implementation 11

Results 12, 13

Recommendations 14

Objectives and Constraints - 16


System Design 19, 24, 25


Arrhythmia Algorithm 31

Electrodes and Prefilter 34

Section Subsection Pages


System Controller 36

User Interface 42

Printed Circuit Board 45


Arrhythmia Detection 48

System Controller 55

User Interface 57

Project Management

Division of Labor 61

Schedule 62

Conclusions and Future Work Conclusions 64

Future Work 65

real-time electrocardiogram may 2 1...

Documents