Embolic Doppler ultrasound signal detection using discrete wavelet transform

Area of Science:

• Biomedical Engineering
• Signal Processing
• Medical Ultrasound

Background:

• Asymptomatic circulating emboli are detectable using Doppler ultrasound.
• Embolic signals (ES) are characterized as short-duration transient signals.
• Wavelet transform offers optimal time-frequency resolution for analyzing transient signals.

Purpose of the Study:

• To propose and evaluate a detection system for embolic signals (ES) using the discrete wavelet transform (DWT).
• To identify parameters useful for describing embolic signals (ES).
• To assess the feasibility of DWT for distinguishing ES from artifacts and Doppler speckle.

Main Methods:

• A fast discrete wavelet transform (DWT) algorithm utilizing Daubechies eighth-order wavelet filters across eight scales was employed.
• Two independent datasets, each containing embolic signals (ES), artifacts, and Doppler speckle (DS), were used for evaluation.
• Threshold values were optimized using the first dataset to differentiate signal types based on DWT coefficients at different scales.

Main Results:

• DWT coefficients for artifacts predominantly appeared at higher scales (5-8), while ES and DS were concentrated at lower scales (1-4).
• The DWT inherently filtered out most artifacts during the transform process.
• Detection accuracy for ES was high: 98/100 in the first dataset, 95/100 in the second dataset, and 198/202 in a third dataset.

Conclusions:

• The discrete wavelet transform (DWT) is a feasible and effective method for detecting embolic signals (ES) in Doppler ultrasound.
• The DWT system demonstrates high sensitivity and specificity in differentiating ES from artifacts and Doppler speckle.
• This approach offers a robust tool for the analysis and detection of asymptomatic circulating emboli.