Ultrasound myocardial integrated backscatter signal processing: Frequency domain versus time domain

H. Rijsterborgh*, F. Mastik, C. T. Lancée, P. Verdouw, J. Roelandt, N. Bom

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

24 Citations (Scopus)

Abstract

In the literature, different forms of measuring the ultrasound power returned by myocardial tissue are reported. Frequency domain methods will give the maximum frequency information, whereas time domain methods are limited in bandwidth, but more practical to realize. It was the purpose of this study to compare the various methods of signal processing. High frequency ultrasound signals from a pig's myocardium, digitally recorded during normal contractile performance, were analyzed by six different methods of signal processing to obtain estimates of backscatter power. The myocardial tissue characterization parameters studied were the integrated power as well as its cyclic variation during the cardiac cycle. A total number of 8109 ultrasound traces obtained in 16 pigs were processed. The study included three signal processing methods in the frequency domain: frequency compensated integrated backscatter calculated over both a large (4 MHz, method 1) as well as a small frequency bandwidth (2 MHz, method 2) and uncompensated integrated backscatter (method 3), and three methods in the time domain: high frequency signal squared and integrated (method 4), mean rectified signal level (method 5) and mean signal level after logarithmic compression and envelope detection (method 6). The random measurement variation (including beat-to-beat variation) was analyzed as well as the paired differences of the backscatter parameters obtained by the respective methods as compared with the only theoretically correct method in the time domain (method 4). The magnitudes of the random measurement variation expressed as a standard deviation (SD) were comparable (range 0.93-1.2 dB) except for method 6 (0.61 dB), where the measurement variation is decreased by the logarithmic compression. Analysis of the variation (SD) of the paired differences of absolute backscatter levels, compared with the "gold" standard, resulted in 1.8 dB (method 2), 1.2 dB (method 3), 1.2 dB (method 4), 1.3 dB (method 5) and 1.5 dB (method 6). Comparison with method 4 yielded 1.2 dB (method 1), 0.8 dB (method 2), 0.0 dB (method 3), 0.4 dB (method 5) and 1.5 dB (method 6). The paired differences of cyclic variation showed the same tendencies. From these results, it can be concluded that a limited bandwidth of high frequency signals produce only small differences in the measurements of cyclic variation, even if the applied processing method is not strictly valid theoretically. Signal processing in the frequency domain offers the advantage of an extended frequency bandwidth. Using time domain methods with a limited bandwidth, single measurements of cyclic variation may deviate as much as ±2 dB (±2 SD) from those obtained in the frequency domain using the full frequency bandwidth. However, in normal myocardial tissue these differences appear to cancel out.

Original languageEnglish
Pages (from-to)211-219
Number of pages9
JournalUltrasound in Medicine and Biology
Volume19
Issue number3
DOIs
Publication statusPublished - 1993

Bibliographical note

Funding Information:
Acknowledgement--These investigations are supported by The Netherlands Technology Foundation (STW).

Fingerprint

Dive into the research topics of 'Ultrasound myocardial integrated backscatter signal processing: Frequency domain versus time domain'. Together they form a unique fingerprint.

Cite this