Ultrasonic arterial vibrometry with wavelet based detection and estimation

Thesis (Ph. D.)--University of Washington, 2000Since the invention of the stethoscope, detection of vibrations and sounds from the body has been a touchstone of diagnosis. However, the method is limited to sounds transmitted to the skin with no means to determine the anatomic and physiological source of the sounds save the cunning of the examiner.To aid diagnosis, researchers have attempted to correlate the characteristics of the sounds to the type and significance of the disorder, and they have sought to understand the mechanisms that generate the vibrations producing the sound. Despite three decades of research these goals have met limited success. Important obstacles have been the inability to non-invasively and efficiently (1) detect the vibrations and (2) fully measure the vibration characteristics.This document introduces an extension to current techniques of detecting and measuring arterial vibrations. Using ultrasound quadrature phase (complex) demodulation methods similar to those of ultrasonic color flow imaging, we have developed a system to detect and measure tissue vibrations with amplitude excursions as small as 30 nanometers. Morlet wavelet analysis, combined with non-parametric binary hypothesis testing, permits sensitive and specific detection and measurement of short duration vibrations amidst clutter and time-varying, colored noise. Specifically, the wavelet spectra will consist of Gaussian noise under the null hypothesis that vibration and blood flow are absent. Determination of the presence of vibration is then based on the expected distributions of the novel normalized wavelet power spectra or normalized cross wavelet spectra associated with the Gaussian noise, as derived in this document. Receiver operating curves for the detector were generated from simulated vibrations. These curves reveal expected detection rates in excess of 99.5% at false alarm rates below 1% for signal to noise ratios as small as 0.5. Subsequent to detection, the Morlet wavelet is also used for characterizing the detected vibration. Vibrations and their quantified characteristics are displayed on conventional ultrasound images to aid the clinician's diagnosis.The detector and estimator have been successfully applied to phantom (physical model) and in vivo vibrations. A few examples are shown, including an arterial bleed. The current system is offline, though a real-time implementation will be straightforward