This aim of this start-up project was to gain experience in body surface potential mapping. A commercial 32-channel amplifier (Elettronica Trentina) was utilized, making use of a direct memory access channel (DMA) of a Compaq PC 486 (1000 Hz/frame). Body surface isopotential potential maps were reconstructed from sparce irregular grids using B-spline and minimal curvature interpolation methods, and then numerous parameters were extracted.

In addition, after developing the required techniques, isochrone and isointegral maps were reconstructed, to enable a more precise serial comparison of electrocardiographic activity on the surface of the human torso. After removing artefacts and signal averaging, data was analyzed by mainly self-written software on a Compaq PC and on a SUN Sparc 10 workstation - Unix, OpenWindows. The two systems were linked via Ethernet. Some of the visualization on the Sun was done by custom software, provided by Philips, Germany, Hamburg lab.

The system was used later by the hosting laboratory for detection of alograft rejection after heart transplantation.

Reducing the number of leads in body surface potential mapping

We first conducted a thorough literature study. Work by other investigators (R. Lux, J. Abildskov, B. Taccardi and others) have shown that reducing the number of leads down to 30 is feasible. As a starting point for our reduced grid selection we used the dense non-regular 219 lead grid of Taccardi. This grid contains 157 electrodes on the front- and 62 electrodes on the backside. The hosting lab was also kindly provided earlier with 6 averaged 219 lead recordings, taken by Professor Taccardi with this grid - 3 normal and 3 pathological ones. We first reconstructed the full 219 lead maps. Then we reconstructed the same maps with several 32 lead sets, that we chose guided by several special criteria like requirement for higher electrode density on the front side, preference for the 12 standard and the Kornreich leads etc. The set complied with the grid proposed earlier by C. Zywietz, for regular sampling of information content (using entropy). The reconstructed maps were compared to the ones reconstructed from the dense (219 lead) grids, at different time instants. We chose the subset that showed the least statistical difference.

To be able to place the electrodes precisely, we prepared a special vest in 3 sizes taylored for different body habitats. One of these vests, with the 32 electrodes attached, is shown for the front side and the rear side of the torso, and during recording.

Variability of body surface maps in normal patients in serial comparison of multichannel recordings and maps

A method for quantifying the fluctuations of the QRS electrocardiogram over the unrolled thoracic surface in normal subjects is presented. Serial comparisons were performed on 32-lead ECG recordings and body surface maps from 7 healthy men, using as similarity measures correlation- and rms-difference in both time-signal and 2D map domain. Recordings were made on one and the same day, and on different days. The time-signal correlation- and rms-differences were plotted as 2D distributions. Correlation was higher on the front side of the torso and rms-differences were largest in the precordial area. The average time-signal correlation calculated over the QRS in comparisons of signals taken on different-days was 0.9717, whereas the corresponding rms-differences amounted to 57.8 mV. Additionally, 2D comparisons between reconstructed map frames at different time instants in the QRS were carried out, using the 2D versions of the correlation and rms-difference. The averaged results of 2D comparisons between isopotential map frames were very close to these values. The averaged results of the 2D comparisons were close to the time-signal comparison values (respectively 0.9696 and 68.9 mV). Finally, 2D comparisons between isointegral maps (mapping the QRS signal integral) were performed. They yielded even higher correlations (0.98, 3.453 mV.s). Although the number of subjects studied was not high, the investigation brought to light important variability ranges that may serve as a basis when detecting pathologies. Besides, the topology of normal ECG variability on the body surface was unveiled. This study confirmed the pertinence of correlation and rms-difference as measures of time-signal and map similarity.

• Spatial distribution of the rms differences (in mV) between QRS complexes of one and the same patient recorded: (a,c) on the same day, and (b,d) on different days.
• Spatial distributions of the time domain cross-correlations between QRS complexes of the same patient, estimated by comparison of recordings from: (a,c) the same day, and (b,d) different days.

In a three-way ANOVA model we analyzed the variance of various comparison estimates, as influenced by the factors "patient", "day" (same/different) and "recording" (first/second). The strongest impact had "day". A significant contribution of the factor "patient" and its interaction with "day" was established, indicating that a portion of the day-to-day variance was caused by the patient himself, possibly through physiological fluctuations. This proved that not all variations originate from altered electrode sites.

Related PUBLICATIONS:

1. Gramatikov, B. Variability maps of body surface ECG in normal subjects. Physiological Measurement, IOP publishing, Vol. 16, 1995, pp. 239-252.

2. Gramatikov, B. Comparison of ECG maps reconstructed from irregular grids of limited number of electrodes. 39th International Colloquium of the Technical University Ilmenau, Sept.27-30, 1994, Ilmenau, Germany. In Proceedings, B-2.3.1., Vol.2, pp. 411-415.

3. Gramatikov, B. Serial comparison of multilead ECG recordings used for mapping. Proceedings of the Bulgarian Seventh National Conference on Biomedical Physics and Engineering (with international participation), Sofia, 17-19 Oct 1996, in English, pp. 117-120.

4. Gramatikov, B. Electrocardiographic surface mapping of the human torso. Fourth National Conference of applied science "Electronic Technology - ET'95" of The Technical University of Sofia. 27-29 Sept. 1995, Sozopol, Bulgaria. In Proceedings, pp. 122-126 (in Bulgarian language).