Development of Advanced Holter Monitor with
Extended Recording and Episode Detection
Domain: Healthcare
IMPRINT Project Code: 6849
Team
Dr. Goutam Saha, IIT Kharagpur
Dr. Anirban Mukherjee, IIT Kharagpur
Dr. Samit Ari, NIT Rourkela
Dr. Nitish Naik, AIIMS New Delhi
Dr. Harikrishnan S. Pillai, SCTIMST, Trivandrum
The document is meant only for expert reviewers of IMPRINT-India.
It may contain intellectual property matters which is yet to be protected.
This is not to be put in public domain.
Private &
Confidential
Project Timeline
Date of sanction letter: 29/11/2016
Receipt of MHRD grant (1st year): 02/03/2017
Receipt of ICMR grant (1st year): 05/10/2017
Project staff recruitment: Allam Jaya Prakash on 13-07-2017 (NITR)
(completed: 3 round of Advt.#) Supratim Manna on 04-09-2017 (IITKGP)
Sudestna Nahak on 30-01-2018 (IITKGP)
#Not getting project staff in spite of repeated advertisements created some difficulty.
Private &
Problem Addressed Confidential
Issue with:
• Off-line processing
• Missing of some of the
arrhythmia burden
• Computation of heart
rate variability
• Missing important
episodes
• Expensive
• Proprietary
• Incorporation of
advanced signal
processing & pattern
recognition tech.
Holter Monitor • Real-time analysis &
‘Jeff’ Holter of 1962 use of result
currently in use
Private &
Confidential
Objectives
To develop and evaluate
(i) an Advanced Holter Monitor for extended period and
episode detection and
(ii) wireless transmission of ECG signal from the wearable
prototype to a remote device/ monitor.
The aim is to couple the wearable external device to a
mobile device which can serve as a computing and display
unit for episodes of interest or serve as a vehicle to connect
to a server that can take higher computational load.
Private &
Work done Confidential
• From human body the ECG signal
will go to the DSP through sensors.
• After processing the ECG data,
those data will be sent to an
android application via wireless
communication (via bluetooth).
• For data acquisition we are using
TMS320F2812 DSP. It is 150 MHz
processor with an internal 16
channel 12-bit ADC.
• The Serial Peripheral Interface (SPI)
will be used to store data in an
external memory (microSD card).
• A Serial Communication Interface
(SCI) will be used for bluetooth
communication with an android
device.
Private &
Work done Confidential
Interfacing android application for wireless transmission of ECG signal
• We are using HC-05
bluetooth module for the
communication between
the android application
and the TMS320F2812
processor.
• In the android application
interface we can see that
HC-05 bluetooth module is
available for
communication and it is
already paired with the
processor.
Private &
Work done Confidential
Real time plotting of the data obtained from three
• By using “Connect” and
different ADC channel
“Disconnect” button we can
communicate with bluetooth.
• By using ‘Start Stream” button
we can control when we want
to start the plotting.
• By using “Auto Scroll X” button
we can scroll the X-axis.
• By using “Lock X axis” button
we can lock the scrolling of X-
axis.
• By using “F” and “S” button we
can choose different
combinations of ADC channels
for data plotting.
❖ By using “All” button we can
see the plot of all ADC
channels.
Raw data from 12 channels
Private &
Work done Confidential
Computer-aided analysis is
required for studying long ECG
records in short time.
Steps:
• Pre-processing of the
acquired ECG signal
• Determination of R-peak
location (Pan-Tompkins
Algorithm)
• Extraction of the feature
set from the ECG signal
• Detection of arrhythmias
from ECG signal
Private &
Confidential
Work done
Normal (N)
Pre-processing and Feature ECG arrhythmia
ECG data Atrial fibrillation (AF)
R-peak detection extraction specific model
Paced beat (P)
Training phase
Testing phase
Pre-processing and Feature Matching
ECG data
R-peak detection extraction algorithm
Classification of
ECG arrhythmia
Pre-processing & R-peak detection: Filtering – Differentiation – Squaring – Moving
Window Integration – Adaptive Thresholding
Feature Extraction: Dual Tree Complex Wavelet Transform (DTCWT)
Classifier: Artificial Neural Network (ANN)
Private &
Confidential
Work done
A sample ECG signal Filtered output
Detected R-peaks of the ECG signal
Differentiator output Moving average filtered output
Private &
Confidential
Work done
Confusion matrix for the Classification results
Class Confusion matrix Class Confusion matrix
Database used
[DWT] [DTCWT]
Physionet
N AF P N AF P
100 series: Normal
200 series: Abnormal
N 65501 850 957 N 65998 642 668 mixed with normal
segments
AF 1122 5933 400 AF 976 5920 224 Length of ECG > 30 min
P 141 541 4297 P 87 54 4838 No. of ECG : 44
Private &
Work done Confidential
Classification performance of DWT and DTCWT based method
Method Class Performance matrix
Acc (%) Sen (%) Spe (%) Ppr (%)
N 96.15 97.32 89.84 98.11
DWT-MLP
AF 96.35 79.58 98.08 81.01
P 96.44 86.30 98.18 76.00
N 97.02 98.05 91.45 98.41
DTCWT-MLP
AF 97.50 82.56 99.04 89.84
P 98.58 97.17 98.67 82.98
Private &
Confidential
Patent / Publication
• Prototype development is still under progress. At its
completion, patent documentation will be done.
• Manuscript is under preparation.
Private &
Project Milestones Confidential
Activity Status
Study of ECG data, basic algorithms, Completed
use of standard database
Development of data acquisition In progress
system, Development of wireless
module and its incorporation into the
device; Display of acquired signal
Development of ECG analysis software In progress
Field data collection from developed -
prototype and analysis of the same
Incorporation of analysis tool as -
backend software
Testing for real-time diagnosis -
Amount of work completed (estimated): 30 percent
Private &
Fund Utilization Confidential
Source of fund Amount (Rs. In Lakhs)
MHRD 37.00
ICMR 26.20 Total: 63.2 L
Expenditure Heads Amount (Rs. In Lakhs)
Equipment 23.00
Salary 4.70
Consumable + Travel + Contingency 1.90
Institute overhead + NIT Rourkella transfer 12.30
Blocked for NIT Rourkella transfer 8.20 Total: 50.1 L
Private &
Confidential
Thank You
The document is meant only for expert reviewers of IMPRINT-India.
It may contain intellectual property matters which is yet to be protected.
This is not to be put in public domain.
