Wavelet analysis of precordial doppler ultrasound on venous air embolism

Wavelet analysis of precordial doppler ultrasound on venous air embolism

陳鎮邦(Chun-bong Chan)

香港大學/Doctor of Philosophy/Electrical and Electronic Engineering/博士(1997年)

並列摘要

(Uncorrected OCR)
Abstract of thesis entitled
Wavelet Analysis of Precordial Doppler Ultrasound on
Venous Air Embolism
submitted by
Chan Chun Bong Brent
for the degree of Doctor of Philosophy at The University of Hong Kong
in March 1997
Venous air embolism (VAE) is a life-threatening complication commonly encountered during neurosurgical procedures with patients in sitting position. Normal incision below the heart level causes bleeding but incision wounds above the heart level can be of sub-atmospheric pressure, resulting in air being sucked to the venous system, which is known as VAE. The non-invasive precordial Doppler ultrasound (PDU) is widely adopted for VAE monitoring but it requires constant listening to the Doppler heart sound (DHS) by an attentive anesthesiologist for detecting VAE and cannot quantify the embolic air volume entering the heart chamber. This quantitative information is valuable as clinical treatment will be prescribed only when clinically significant embolic air volume (above 1 ml) is present.
In this thesis, processing of DHS signals using wavelet analysis (WA) for overcoming various shortcomings of the current PDU method was studied. Over twenty anesthetized dogs were tried in a number of simulation experiments of VAE
with bolus injection (0.01 to 0.80 ml) and continuous infusion (0.80 to 9.60 ml) of air via the right external jugular veins. Results showed an important observation among all the subjects that the contrast of embolic to normal DHS signal power was substantially enhanced by the WA, facilitating automatic detection of embolic DHS by a simple thresholding on the energy profile of the wavelet-processed DHS signals. The embolic signal power was further found to be linearly related to the air volume administered by either bolus injection or continuous infusion in log-log scale. Besides, there was statistically no difference (P > 0.05) in each slope and intercept of the regression lines for both types of injection. Hence, it can be deduced that the embolic power at large air volume is proportionate to that of small volume, enabling estimation of clinically significant embolic air volume based on prior injection of sub-clinical air volume.
A novel fast detection algorithm for VAE was developed based on the above findings, which was further implemented as a real-time VAE monitor with detection range up to 10 ml, sufficient to cover both sub-clinical and clinically significant embolic air volume. The sensitivity was found to be 0.0007 ml/kg, even better than the currently most sensitive (0.02 ml/kg) but invasive VAE monitoring method of transesophageal echocardiography. Furthermore, the real-time trace of the embolic power rendered by the monitor provided a readily-comprehensible record of the development of VAE during surgery, which is valuable to anesthesiologists for assessing the risk of VAE and planning proper clinical treatment. Though dogs instead of humans were tried in the present simulation experiments of VAE due to the ethical issue, the higher heart rate of dogs as compared with that of adult humans actually imposed a more stringent computational requirement on the monitor. Hence, VAE monitoring in humans using the real-time monitor should also be feasible.

並列關鍵字

Doppler ultrasonography. ； Wavelets (Mathematics) ； Embolism.