Signal processing strategies for enhancement of signal-to-noise ratio of thermodilution measurements

Abstract

Thermodilution measurements of clinically important cardiac parameters, such as cardiac output and stroke volume, are subject to many sources of error. The temperature fluctuations (thermal noise) normally found in the pulmonary artery constitute one of these sources of errors. To improve the signal-to-noise ratio of thermodilution flow measurements rather than increase the signal level, we investigated four signal processing strategies designed to reduce the thermal noise power. We applied the noise reduction strategies to thermal noise data, containing simulated thermodilution curves, obtained in a mock circulatory loop. We compared the accuracy and reproducibility of the curve area estimates produced by the algorithms to the area estimates obtained by numerical integration of the thermal signal. Our results show that a bandpass (BP) integration technique combined with a noise canceler can improve thermodilution curve area estimate reproducibility and accuracy. The BP integration technique improved the reproducibility of cardiac output measurements by roughly 16 dB and is directly applicable to most thermodilution hardware currently in use. The more accurate noise cancelers, combined with the BP integration technique, provided correspondingly improved signal-to-noise ratios, with the improvement ranging up to 50 dB.