Super-resolution spectral analysis for ultrasound scatter characterization

Abstract

Parametric Bayesian spectral estimation methods have been previously utilized to improve frequency resolution. Ultrasound signals have been tested in such methods resulting in higher precision frequency detection compared to common non-parametric spectral estimation methods based on the Fourier transform. Such a technique using a reversible jump Markov Chain Monte Carlo algorithm has been developed to fully characterize signals and in addition to frequency, to provide amplitude and noise estimation. The analysis of this method is demonstrated with a real copper sphere ultrasound scatter signal. Based on typical diagnostic ultrasound data between 1.2–4.5 MHz the new spectral estimation achieves 110 kHz minimum frequency resolution. This is at least twice the resolution of Fourier based methods, resulting in revealing new frequencies. The method may be used in the entire range of ultrasound imaging modalities and may help provide improved sensitivity, reproducibility and spatial resolution.