Adaptive Models for Multi-Covariate Imaging of Sub-Resolution Targets (MIST).

Journal Article (Journal Article)

Multi-covariate imaging of sub-resolution targets (MIST) is a statistical, model-based image formation technique that smooths speckles and reduces clutter. MIST decomposes the measured covariance of the element signals into modeled contributions from mainlobe, sidelobes, and noise. MIST covariance models are derived from the well-known autocorrelation relationship between transmit apodization and backscatter covariance. During in vivo imaging, the effective transmit aperture often deviates from the applied apodization due to nonlinear propagation and wavefront aberration. Previously, the backscatter correlation length provided a first-order measure of these patient-specific effects. In this work, we generalize and extend this approach by developing data-adaptive covariance estimation, parameterization, and model-formation techniques. We performed MIST imaging using these adaptive models and evaluated the performance gains using 152 tissue-harmonic scans of fetal targets acquired from 15 healthy pregnant subjects. Compared to standard MIST imaging, the contrast-to-noise ratio (CNR) is improved by a median of 8.3%, and the speckle signal-to-noise ratio (SNR) is improved by a median of 9.7%. The median CNR and SNR gains over B-mode are improved from 29.4% to 40.4% and 24.7% to 38.3%, respectively. We present a versatile empirical function that can parameterize an arbitrary speckle covariance and estimate the effective coherent aperture size and higher order coherence loss. We studied the performance of the proposed methods as a function of input parameters. The implications of system-independent MIST implementation are discussed.

Full Text

Duke Authors

Cited Authors

  • Ahmed, R; Flint, KM; Morgan, MR; Trahey, GE; Walker, WF

Published Date

  • July 2022

Published In

Volume / Issue

  • 69 / 7

Start / End Page

  • 2303 - 2317

PubMed ID

  • 35613063

Electronic International Standard Serial Number (EISSN)

  • 1525-8955

International Standard Serial Number (ISSN)

  • 0885-3010

Digital Object Identifier (DOI)

  • 10.1109/tuffc.2022.3178035

Language

  • eng