an fpga implementation of real-time qrs detection h.k.chatterjee dept. of ece camellia school of...

13
An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera, M.Mitra Dept. of Applied Physics University Of Calcutta Kolkata India m5151117 Yumiko Kimezawa October 11, 2012 1 RPR

Upload: milo-freeman

Post on 25-Dec-2015

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

An FPGA implementation ofreal-time QRS detection

H.K.Chatterjee Dept. of ECECamellia School of Engineering & Technology

Kolkata India

R.Gupta, J.N.Bera, M.MitraDept. of Applied Physics

University Of Calcutta Kolkata India

m5151117

Yumiko Kimezawa

October 11, 2012 1RPR

Page 2: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Outline

• Introduction• Materials and Methods• Testing and Results• Conclusion

October 11, 2012 RPR 2

Page 3: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Introduction

• QRS detection is one of the important and primary job and very often used for heart rate computation

• In recent years, there has been considerable use of FPGA based system for ECG monitoring, QRS detection and feature extraction

• This paper illustrates a real time QRS detection algorithm using an FPGA based embedded system

October 11, 2012 RPR 3

Page 4: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Materials and Methods

October 11, 2012 RPR 4

Data Port

Status Port

Control Port

ptb-db file

Data Capture & Analysis

Section

Display Section

8 Switches

8 LEDsInterfacing Unit

Parallel port

PC

FPGA Xilinx Spartan 2

Figure 1: Block diagram of the system

Trigger pulsetrain

Start Capture

Page 5: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Materials and Methods

The entire work• Generation of digitized ECG from ptb-db

file

• Development and testing of the algorithm in FPGA platform

October 11, 2012 RPR 5

Page 6: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Materials and Methods• Generation of digitized ECG from ptb-db

file

October 11, 2012 RPR 6

Figure 2: Generation of ECG data by PC

Data send from PC parallel port (D0 – D7)

Data accepted by FPGA (P108 – 111, P113 – 115, P119)

Pulse train generated by FPGA (P121), & accepted by PC parallel port (S7)

“Start Capture” pulse generated by PC parallel port (C0) & accepted by FPGA (P120)

Page 7: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Materials and Methods

• Real time QRS detection from the ECG samples- The training zone: The first 1500 samples- A characterization of QRS polarization is performed

based on 20 point slope- ECG samples are stored in a group of memory cells

which holds the last 42 samples- Computing 20 point average slope by calculation

differences like R20-R19, R19-R18,….., R2-R1

October 11, 2012 RPR 7

Formula:

Page 8: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Materials and Methods

October 11, 2012 RPR 8

R1

R 42

R 20

Current index point

Current index point

Current pointof reference

R 20

Figure 3: Illustration of characterization of QRA complex

Average of both side slope: (|R20-R40| + |R21-R0|)/2

Group I: Left side 20 pt. slope: R42-R21 > 0 & Right side 20 pt. slope: R20-R1 < 0

Group II: Left side 20 pt. slope: R42-R21 < 0 & Right side 20 pt. slope: R20-R1 > 0

Page 9: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Testing and Results

• Test using normal and abnormal data in MIT-PTB database and MIT-BIH arrhythmia database

• Initially performed in the MATLAB platform• 30000 samples (Single lead data)• Resolution: 8-bit• sampling interval: 1 ms

October 11, 2012 RPR 9

Page 10: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Testing and Results

• The evaluation criteria are Sensitivity (Re) and Positive Predictivity (P+), defined as:

October 11, 2012 RPR 10

and

TP (True-Positive): correctly found R peaksFN(False-Negative): missed R peaksFP: the number of misdetection

Page 11: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Testing and Results

• MATLAB simulation- With mit-db, 60,000 samples• Re and P+ of 97.82 %• 98.35 % respectively

- With ptb-db, 120 leads• An average sensitivity of 99.47 %• Predictivity of 95 %

October 11, 2012 RPR 11

Page 12: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Testing and Results

• FPGA implementation- With a total of 100 single lead data, each

containing 7,000 samples• An average Re and P+ of 94.8 %• 98.17 % respectively

October 11, 2012 RPR 12

Page 13: An FPGA implementation of real-time QRS detection H.K.Chatterjee Dept. of ECE Camellia School of Engineering & Technology Kolkata India R.Gupta, J.N.Bera,

Conclusion

• The algorithm is implemented with synthetic ECG data from ptb-db and mit-db

• In the present approach, 20 point average slope eliminates the effect of high frequency noise and to minimize the effect of any momentary spike

• 2 point slope should not exceed the slope threshold criteria

October 11, 2012 RPR 13