Acquiring a grid of 25 EKG signals

Here is an example of using the component software to acquire a 5x5 grid of 25 EKG signals from the precordial (chest) leads. This example is similar to the one shown previously, which was used to acquire the standard 12 EKG lead signals.

Once again, the external electronics used for data acquisition was amplifier #4 used with an output attenuation of 20%. Four electrodes made from 1964 FDR silver dimes filed down to remove the embossing were attached to the 3 limb locations (right arm, left arm, left leg) and to the chest. The limb electrodes were attached dry using masking tape and 10 gauge speaker cable. The chest electrode was applied wet, using conducting gel (Neuroscan QuickGel, henceforth referred to only as "goop"), masking tape, and more flexible 14 gauge speaker wire.

The chest electrode was sequentially positioned at each point in a 5x5 grid of 25 points across the chest, one point at a time. The 5x5 grid had a spacing of 3 inches between points, and the center of the grid was located near the standard precordial lead point V4. In all future discussion, points on this grid will be referred to by their vertical and horizontal coordinates, each ranging from 1 to 5. Therefore, I call the 25 points of the grid V(1, 1) to V(5, 5), these two extremes signifying the upper right and lower left corners of the grid. This grid was large enough to completely encompass the standard precordial positions V1-V6, and to include additional parts of the chest away from these points as well.

Acquiring all 25 signals took a little over an hour, in which most of the time was spent repeatedly removing, cleaning, regooping, and reattaching the chest electrode as it was moved through each of the set of 25 points in the grid. In this example, up to about a minute of unfiltered data was acquired in individual 5 second records (epochs) from each electrode position. 1280 points were acquired in each record, which gives a sampling rate of 256 points/sec, which is sufficient to record signals up to 128 Hz. During acquisition, the AC power was turned off throughout the house so that no 60+ Hz line filtration was needed (but it could be applied to the data later). Each set of signal records was saved in a first-in-first-out (FIFO) data structure which could be perused later. Data was acquired from each grid point until the pulse rate stabilized at about 1 waveform per second.

After acquisition, each FIFO structure was perused using a slider to extract its contents, and a single 5 second record of data characteristic of the signal from that electrode position was selected and saved in a 1d plot. The resulting dataset for the entire grid consists of a set of 25 5 second raw records, one from each point in the grid.

The overall raw data, arranged by position, looks like the following. In this image, the scale of each plot is from +1V to -1V, although the Mac sound input clips at about 1.5 V peak-to-peak.

(Click for larger image)

Here is an example of using the component software to extract 4 1 second waveforms from each of the 25 5 second records shown above. Before each subset is extracted, the entire 5 second record is passed through a sharp lowpass Fourier filter to remove any frequency components above 50 Hz. Each subset of the record is synchronized about the R peak of the waveform, arbitrarily placing that peak 1/4 of a second into the subset. All four subsets are then averaged together to form a single waveform which is indicative of the EKG signal occurring at a specific electrode coordinate position. The averaging process also removes much of the spurious in-band noise which varies from subset to subset, producing a smoother curve.

Here is the filtered 5 second record for signal coordinate (3,3), and the average of the 4 subset waveforms contained in that record.

Here is the result of filtering, extracting, synchronizing, and averaging 4 subsets from each of the 5x5 grid location signals. In this image, all average waveforms have been plotted on different scales to show the variation in waveform shape with respect to electrode location. Note that some of the waveforms have inverted features (such as the inverted P and T "waves" in the first row and first column), suggesting that these signals are above and to the right of the initial depolarizing "trigger" event location, which is presumably normal Sino-Atrial depolarization in this case.

(Click for larger image)

©Copyright Sky Coyote, 2001-2002.