From D.Rea at qub.ac.uk Tue Apr 12 09:27:07 2005 From: D.Rea at qub.ac.uk (Derrick Rea) Date: Tue, 12 Apr 2005 14:27:07 +0100 Subject: [Physionet-users] wrsamp precision Message-ID: <007701c53f63$542e4d80$f01d758f@DREAX0DBAS> I want to use sqrs and ecgpuwave to detect QRS complexes in ecg signals we have recorded but have run into problems. The ecg signals were recorded with variable pad contact and so vary widely in their voltage range. I believe that both these algorithms will still find the QRS, adjusting their threshold levels given signal strength, but found that precision in the original signal was being lost when i converted my signal to a wfdb .dat using 'wrsamp -f 13 -F 200 -i bj3ophecgbase1.txt -o bj3ophecgbase1 1'. ie bj3ophecgbase1.txt file time ecg value 0.000000 2.747447 0.005000 2.713942 0.010000 2.613425 0.015000 2.479404 but examining this signal using 'rdsamp -r bj3ophecgbase1' shows that the ecg value is rounded to the nearest integer. sample ecg value 0 3 1 3 2 3 3 2 I can't see a switch in wrsamp that forces the full precision ecg value to be read? I've tried the -p -p switch in rdsamp thinking the precision problem may be in rdsamp but to no avail. Is there a way to ensure full precision of my file to be maintained as a wfdb .dat file without delving in the source code? I've also tried to run 'ecgpuwave -r mitdb/100 -a qrs' but find that I get a segmentation fault (core dumped). Anyone an idea why this would be happening? sqrs works well on the same file. Derrick -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.physionet.org/pipermail/physionet-users/attachments/20050412/3eff6aa9/attachment.html From george at mit.edu Tue Apr 12 12:37:36 2005 From: george at mit.edu (George B. Moody) Date: Tue, 12 Apr 2005 12:37:36 -0400 Subject: [Physionet-users] wrsamp precision In-Reply-To: <007701c53f63$542e4d80$f01d758f@DREAX0DBAS> References: <007701c53f63$542e4d80$f01d758f@DREAX0DBAS> Message-ID: <425BF950.4070604@mit.edu> The output of wrsamp is a signal file containing integers, and if you don't scale wrsamp's input, it will simply round all input values to the nearest integer, as you observed. Fortunately, wrsamp's -x option does the scaling for you. I created a short text file, bj3oph.txt, containing the following (cut-and-pasted from your note): 0.000000 2.747447 0.005000 2.713942 0.010000 2.613425 0.015000 2.479404 I then ran these commands: wrsamp -F 200 -x 10000 -G 10000 -o bj3oph 1 Hi, I'm interested in digitizing a rather large archive of existing paper ECG plots. http://www.physionet.org/physiotools/digitizing/ provides a nice writeup of existing techniques and some software packages suited to the task, although they seem more appropriate for the occasional scan & capture of an ECG. I'm wondering if anyone is familiar with a company and/or commercial product that will easily archive a significant number of ECGs? By significant number I mean on the order of hundreds if not thousands of ECG plots. In other words, does anyone know of an "industrial strength" paper ECG digitizer or company capable of providing such a service? Thanks for any input! -Len Marinis len.marinis at cogentsoftware.net http://www.cogentsoftware.net From rasch at med1.med.tum.de Mon Apr 18 15:12:32 2005 From: rasch at med1.med.tum.de (Raphael Schneider) Date: Mon, 18 Apr 2005 21:12:32 +0200 Subject: [Physionet-users] Digitizing Paper ECG Archives (en masse) In-Reply-To: <20050418180152.14307.qmail@web41510.mail.yahoo.com> References: <20050418180152.14307.qmail@web41510.mail.yahoo.com> Message-ID: <200504182112.33405.rasch@med1.med.tum.de> Hi, On Monday 18 April 2005 20:01, Len Marinis wrote: > I'm interested in digitizing a rather large archive of existing paper ECG > plots. http://www.physionet.org/physiotools/digitizing/ provides a nice > writeup of existing techniques and some software packages suited to the > task, although they seem more appropriate for the occasional scan & > capture of an ECG. > > I'm wondering if anyone is familiar with a company and/or commercial > product that will easily archive a significant number of ECGs? By > significant number I mean on the order of hundreds if not thousands of ECG > plots. In other words, does anyone know of an "industrial strength" paper > ECG digitizer or company capable of providing such a service? The only commercial product I know is 'ECGScan' from A.M.P.S. LLC (www.amps-llc.com). There is also a demo version of the program. I have no experience with the program but maybe it is helpful for you. Raphael