George B. Moody
Harvard-MIT Division of Health Sciences and Technology
If your data are in printed form only, how can they be recovered for further analysis? Although PhysioToolkit does not currently include software for this application, many visitors are interested in digitizing plots such as ECG recordings on paper or data plots from publications. This page offers a very brief survey of resources that may be helpful.
Algorithms for digitizing paper ECGs (which can be applied to other XY plots as well) are discussed in an excellent paper by JT Wang & DP Mital, A microcomputer-based prototype for ECG paper record conversion, J Network and Comput Appl 19(3):295-308 (1996). In addition to the authors' detailed account of their own techniques, the paper includes a good bibliography.
Another relevant paper is by T Kao et al., Computer analysis of the electrocardiograms from ECG paper recordings, Proc. 23rd Annual EMBS, 2001.
Scanning a paper ECG
Use a color scanner with a resolution of at least 600 dpi, and scan the original chart (which should have a red or other distinctively colored grid). The results will be much worse if you have to work from black and white images, because the grid will interfere with recovery of the signals.
Any scanner software should work. Avoid the use of lossy compressed formats (JPEG, GIF); PNG is a good choice, especially if you plan to use ecg2png (below).
Cleaning up the scanned image (separating signals from the grid, etc.)
Larry Widman has written an open-source program for Linux called ecg2png that does this job very well; it can remove the grid or leave it in place as you prefer.
An additional benefit of using ecg2png is that it produces a very compact 2- or 3-color image that displays quickly and cleanly in a web browser. Larry's earlier work on digitizing paper ECGs is cited in both of the papers mentioned above.
Converting a raster image to X-Y pairs
There are a few free programs that can do this:
- Engauge Digitizer: This is free (GPL) software for Linux and MS-Windows, which should be portable to any other platform supported by the Qt toolkit.
- WinDig: Free (binary only) for MS-Windows; very easy to use and highly recommended.
- DIGITIZE ("Computer Program to Digitize X-Y Plots for UNIX"): formerly available from http://www.anl.gov/LabDB/Current/Ext/H021-text.001.html but not on-line as of March 2003. (If you know a current URL for this software, please let me know!)
- NIH Image: This is open-source image-processing software for the Macintosh (MS-Windows and Java ports are also available). See Recovering Data from Line Plots in the NIH Image manual for detailed instructions.
Among the commercial programs worth considering for this application are OPR (for MS-DOS; this seems to be no longer available) and FlexiTrace for MacOS). There are other commercial programs for raster-to-vector conversion that are less well-suited to this task but might be useable as an alternative. Some of these are intended for GIS applications (e.g., digitizing paper maps); others are sold by CAD software vendors and are designed to allow engineering drawings to be scanned and imported into CAD programs; still others, such as Adobe Streamline, are intended for graphic artists who want to be able to manipulate line drawings without introducing raster artifacts. These are not likely to work as well in this application as a specialized program such as Engauge Digitizer, WinDig, DIGITIZE, OPR, or FlexiTrace, because they don't exploit the knowledge that the curves to be traced have the form of a function (i.e., that there is a unique ordinate for each abscissa).
Useful resources for more information about this subject are:
If you would like help understanding, using, or downloading content, please see our Frequently Asked Questions.
If you have any comments, feedback, or particular questions regarding this page, please send them to the webmaster.
Comments and issues can also be raised on PhysioNet's GitHub page.
Updated Thursday, 9 July 2015 at 12:11 BRT