Real-Time Automated Point Process Method for Detection and Correction of Erroneous and Ectopic Heartbeats

Introduction

This page allows scientists and researchers to test our algorithm for the detection and correction of erroneous and ectopic heartbeats.

Permission is hereby granted, free of charge, to any person to use it for academic purposes, subject to the conditions outlined in section “Conditions of use” below.

Copyright (C) Luca Citi, Emery N Brown, and Riccardo Barbieri, 2010-2013.
All Rights Reserved.

{lciti,enb,barbieri}@neurostat.mit.edu    Neuro Cardiovascular Signal Processing Unit

Usage

When you upload a file with a series of R events, our server will process it and return a results file with the processed series. Please make sure that the file content and the file name do not contain sensitive data. The communication with our server is not encrypted and the data file might reside on our server beetween reboots.

The input file should consist of a table saved in a text file. Each row should correspond to a heartbeat and can be made of several fields delimited by TABs, spaces, or commas. One column must correspond to the series of the times of occurrence (in seconds) of R-events (or equivalently other fiducial points of a heart contraction measured from ECG, BVP, ...). Additional columns can correspond to other parameters, such as the series of RR intervals, the type of R event, and so on. All these additional columns will be ignored by our software. Our tool will try to guess which column corresponds to the series of R-events by looking at the one with with the highest numerical value in the last row.
It is important that the column with the time of the R-events is formatted with enough significant digits that at least millisecond accuracy is guaranteed for all events. In other words, fprintf formats like "%.5g" should be avoided because they might result in the first R time being printed as 1.7891 and the last being 1140.7, i.e., with only 100ms accuracy. Use something like "%.3f" instead.

The input file can be created, for example, from data available from the physionet ATM. Choose a record of interest, with heartbeats annotations, then choose "time format" in "seconds", choose a section of the record that you would like to analyze (e.g., 1 hour), then choose "Show RR intervals as text" from the dropdown menu on the right. If you cut and paste the resulting table in a text file you obtain a valid input for our tool. Alternatively, you can the WFDB command ann2rr -r nsr2db/nsr001 -a ecg -f 0 -t 600 -v s -i s3 -V s -w -W > rr.txt to create a file in the correct format. Another example of valid file is Y2.pre from Physionet.

Upload the file using the button below and wait for our server to process it. If the processing is successful you'll be asked for a location on your disk where to save the output file. This file is a text file containing a table containing the corrected time of R-events in the first column, and the action taken for that beat in the second column: 0 means the beat time was identified as correct and not modified, ≠0 that the beat was detected as erroneous and a correction was attemped. Please note that the first minute of the record is used by the algorithm to learn the statistics of the signal. For this reason the detection of erroneous beats is made with a simpler method. The performance of the algorithm should be assessed on the signal after the first minute.

File:

Conditions of use

1. The service is provided "as is" without warranty, either expressed or implied, including, but not limited to, the warranty of correctness, fitness, or anything else whatsoever. The licensee agrees that the use of the software is at their sole risk. In no event shall the copyright owners be liable for any direct, indirect or consequential damages however connected with the use of this service.
2. Any work making use of this tool must explicitly acknowledge the copyright owners. In case of a scientific publication reporting the use of this tool, citation of a relevant paper authored by the copyright owners is the preferred form of acknowledgement.
3. Permission to use this tool for non-academic purposes, including but not limited to, commercial, medical, military purposes, is NOT granted. Such permission, where possible, may be obtained subject to the express prior written consent from the copyright owners.