Abstract
Background: A thorough review is needed for the first-order 0.05 Hz high-pass filter, which was introduced almost fifty years ago before modern techniques were available. We quantify the effectiveness of inverse filtering and assess the changes that the filter imposes on the electrocardiogram (ECG) and particularly on the diagnostically relevant ST-segment.
Methods: The method of inverse filtering was verified by the use of 1620 clinical 10 second ECGs (set 1) recorded without any high-pass filter (DC-coupled). The signals were filtered, rounded, and inversely filtered, and the root-mean-square (RMS) error and maximum error were found. For the high-pass filter evaluation, 1339 different clinical resting ECGs (set 2) were measured in non-filtered and filtered versions by an industry-grade automated measurement and interpretation program (ETM V1.12.0.0). The differences in P, Q, R, S, J, J10, J20, J40, J60, J80, and T amplitudes and ST-segment averages were compared against the QRS integral. The slope of the best linear fit and the correlation coefficient were found for each comparison. Secondly, absolute changes in the ST-segment measurements were computed with a significance level of ±25 μV.
Results: Original ECGs were reconstructed with an RMS error of less than 0.5 μV and a maximum error of ±1 μV (set 1). A clear correlation was found between QRS integral and deviations to the ST-segment (set 2, see Table 1). Any T-wave deviations were poorly described by QRS integral. No correlation was found between QRS integral and changes to other amplitude measurements. 50 ECGs (4%) had significant J10 amplitude deviations.
Methods: The method of inverse filtering was verified by the use of 1620 clinical 10 second ECGs (set 1) recorded without any high-pass filter (DC-coupled). The signals were filtered, rounded, and inversely filtered, and the root-mean-square (RMS) error and maximum error were found. For the high-pass filter evaluation, 1339 different clinical resting ECGs (set 2) were measured in non-filtered and filtered versions by an industry-grade automated measurement and interpretation program (ETM V1.12.0.0). The differences in P, Q, R, S, J, J10, J20, J40, J60, J80, and T amplitudes and ST-segment averages were compared against the QRS integral. The slope of the best linear fit and the correlation coefficient were found for each comparison. Secondly, absolute changes in the ST-segment measurements were computed with a significance level of ±25 μV.
Results: Original ECGs were reconstructed with an RMS error of less than 0.5 μV and a maximum error of ±1 μV (set 1). A clear correlation was found between QRS integral and deviations to the ST-segment (set 2, see Table 1). Any T-wave deviations were poorly described by QRS integral. No correlation was found between QRS integral and changes to other amplitude measurements. 50 ECGs (4%) had significant J10 amplitude deviations.
| Original language | English |
|---|---|
| Journal | Journal of Electrocardiology |
| Volume | 49 |
| Issue number | 6 |
| Pages (from-to) | 936 |
| Number of pages | 1 |
| ISSN | 0022-0736 |
| DOIs | |
| Publication status | Published - 2016 |
| Event | ISCE symposium - Omni Tucson National Resort, Tuscon, Arizon, United States Duration: 13 Apr 2016 → 17 Apr 2016 Conference number: 41 |
Conference
| Conference | ISCE symposium |
|---|---|
| Number | 41 |
| Location | Omni Tucson National Resort |
| Country/Territory | United States |
| City | Tuscon, Arizon |
| Period | 13/04/2016 → 17/04/2016 |