Image-Based Metrics in Ultrasound for Estimation of Global Speed-of-Sound

TL;DR

This work investigates image-based approaches for global speed-of-sound estimation in ultrasound, sidestepping complex propagation models by leveraging image quality, image similarity, and multi-frame statistics. By grid-searching a suite of 11 metrics across RF and B-mode data from simulations and tissue phantoms, it finds that image-similarity metrics (notably MSE, MI, and Correlation) yield high-accuracy global SoS estimates (often within a few m/s), while single-frame image-quality metrics require frame compounding to be effective. Among single-frame metrics, ST-Ten and Focus show promise, especially on compounded images, whereas the multi-frame CV metric performs well with many frames but struggles with small regions. Overall, image-based SoS estimation provides a fast, data-accessible alternative to physics-based methods and offers potential extensions to layered or local SoS imaging, with practical implications for improving ultrasound image formation and diagnostics.

Abstract

Accurate speed-of-sound (SoS) estimation is crucial for ultrasound image formation, yet conventional systems often rely on an assumed value for imaging. While several methods exist for SoS estimation, they typically depend on complex physical models of acoustic propagation. We propose to leverage conventional image analysis techniques and metrics, as a novel and simple approach to estimate tissue SoS. We study eleven metrics in three categories for assessing image quality, image similarity and multi-frame variation, by testing them in numerical simulations and phantom experiments. Among single-frame image quality metrics, conventional Focus and our proposed Smoothed Threshold Tenengrad metrics achieved satisfactory accuracy, however only when applied to compounded images. Image quality metrics were largely surpassed by various image comparison metrics, which exhibited errors consistently under 8 m/s even applied to a single pair of images. Particularly, Mean Square Error is a computationally efficient alternative for global estimation. Mutual Information and Correlation are found to be robust to processing small image segments, making them suitable, e.g., for multi-layer SoS estimation. The above metrics do not require access to raw channel data as they can operate on post-beamformed data, and in the case of image quality metrics they can operate on B-mode images, given that the beamforming SoS can be controlled for beamforming using a multitude of values. These image analysis based SoS estimation methods offer a computationally efficient and data-accessible alternative to conventional physics-based methods, with potential extensions to layered or local SoS imaging.

Figures (4)

• Figure 1: Pipeline overview for optimizing a comparison metric (C.M.) between two acquisitions ($R_1$ and $R_2$) using trial-and-error of difference SoS values. Note that image quality and multi-frame statistic metrics work similarly, but instead with one or several acquisitions, respectively.
• Figure 2: Sample behavior of image comparison metrics, demonstrated for Phantom 1, by normalizing each metric between their minimum and maximum for visualization purposes. The optimum value (opt) found by each metric is indicated in the legend.
• Figure 3: To study the sensitivity of image-based metrics to utilized image window, the distribution of SoS estimation absolute errors are presented for using image patches (layers) of varying sizes from 32 down to 1 mm in depth. Each layer location is perturbed vertically in increments of 0.4 mm creating four layer positions each, in order to increase statistical stability of the evaluation. The evaluation is conducted for the two Phantoms using the Dual acquisition for the comparison and image-quality (via compounding) metrics, and using 17 VS Tx frames for CV. Accordingly, the distributions at 32 summarize 8 estimations (2 phantoms $\times$ 1 32 mm-patch $\times$ 4 perturbations), whereas at 1 sumarize 256 estimations containing 32 1 mm-patches instead.
• Figure 4: The distribution of SoS estimation absolute errors are presented for using image patches (layers) of varying sizes from 32 down to 1 mm in depth, similarly to \ref{['fig:slicing_full']} but when the SoS search range is restricted around the known SoS for each experiment, i.e., $s_i\in \{c_\mathrm{GT}\pm50\}$ m/s. Each layer location is perturbed vertically in increments of 0.4 mm creating four layer positions each, in order to increase statistical stability of the evaluation. The evaluation is conducted for the two Phantoms using the Dual acquisition for the comparison and image-quality (via compounding) metrics, and using 17 VS Tx frames for CV.