ON-LINE STATIC EIT RECONSTRUCTION SERVER
Static image reconstruction from the data measured in vivo is still intriguing problem in EIT. Static imaging here is visualization of absolute conductivity distribution inside body (i.e. real anatomy), not just imaging of difference between two states such as inspiration and expiration (so-called dynamic imaging). During a few years our research group develops fast and robust static EIT imaging method, which is used currently with the experimental EIT systems tested in several Russian medical centers. We develop this server to enable researchers from international EIT community to try our algorithms on the data collected with different instruments and to estimate advantages of static in vivo EIT imaging. This also would help us to improve the methods due to new experience and feedback from the users. The algorithms are described in [paper 1 and paper 2].
Your can upload your data set, and our server will reconstruct static ("absolute") EIT image from it. Keep in mind that equal distances between electrodes and perfect contact of all electrodes with the body is much more important at static reconstruction than at familiar dynamic imaging. The result will be shown in your internet browser. The input data can be represented in text or binary format. The text file format enables to represent data measured with three different strategies:
polar (opposite) current injection and voltage measurements on pairs of adjacent electrodes (currently used in Moscow "TE-1" system)
dipole (adjacent) current injection and voltage measurements on pairs of adjacent electrodes (like in Sheffield systems)
trigonometric current patterns with measurements relative to additional grounded electrode (like in Rensselaer "ACT3" system)
The format of text files looks like simplified version of EIT data exchange format suggested in [P. Record and P. Riu, Raw data interchange format for EIT, Clin. Phys. Physiol. Meas., 1992, 13, Suppl. A, 201-7]. The file must consist of:
First line with measuring strategy keyword: POLAR, DIPOLE (see Note below) or TRIGON.
Line with number of electrodes, for example 16. This number must be even.
N lines with raw EIT data, where N is number of electrodes indicated in line 2.
Each line of the raw data corresponds to one current injection pattern and must consist of N floating point or integer numbers representing the measured potential differences. For polar and dipole injection these are potential differences between adjacent electrodes starting from the first electrode (absolute numbers of electrodes are used!). First line correspond to current injection through first and opposite (polar injection) or first and second (dipole injection) electrodes. For the pairs, which include injecting electrodes, zero values are supposed. It is possible also to use zero values where actual values can be found from reciprocity, but this is not recommended. For trigonometric injection, the values are potentials on each electrode relative to additional grounded electrode. The first line corresponds to the cos lowest frequency pattern, the N/2-th line corresponds to sin lowest frequency pattern. As only N-1 trigonometric current patterns are possible, the last line of data set in this case should consist of zeros. Certainly, trigonometric injection doesn't seem suitable for static imaging due to large influence of contact impedances, and this is least tested part of our software.
The rest lines of the text file can be used for comments, notes and so on.
You can view examples of the text files for polar, dipole and trigonometric injection, use them for testing the server and find more details on HOW TO PREPARE INPUT FILE FOR THE ON-LINE EIT RECONSTRUCTION SERVER page.
The binary data formats include now raw data for polar injection with relative numbering of electrodes (192 floats in PC format, 4 bytes each) and native (".sdb") files from the Moscow "TE-1" system. Multiframe binary files are acceptable.
Note. Actually all input data formats are converted to the polar injection strategy before image reconstruction. This provides the best suppressing of measurement errors and noise. There is possibility to reconstruct images directly from the data measured with dipole injection using keyword TRUEDIP instead of DIPOLE (see example truedip.txt). In some cases (for accurately measured data) this gives better resolution, but we recommend to start from the standard procedure.
You can add the reconstructed images with necessary explanations to our EIT Gallery or just see what other people have put there.
This server is under development, and we will be grateful for any suggestions, contributions and bug reports, which you can direct to Alexander Korjenevsky.