Quantitative ultrasound imaging of bone: anatomical images, tissue structural quality, and pulsatile blood flow

TL;DR

This work tackles the need for a non-ionizing, portable method to jointly assess bone anatomy, tissue quality, and intraosseous blood flow. It introduces a bone-corrected ultrasound reconstruction framework that accounts for cortical refraction and wave-speed anisotropy to simultaneously image bone interfaces, estimate $C^{radial}$ and $C^{axial}$, and map pulsatile flow. In vivo tibial validation against pQCT shows strong cortical-interface alignment ($R^2=0.80$) and significant correlations between wave speeds and BMD ($R^2$ ~ 0.45–0.49), while a ray-selection strategy enhances blood-flow detection and suppresses interface reflections. The approach offers a portable, real-time triad of biomarkers—anatomy, tissue quality, and blood flow—that could aid in diagnosing osteoporosis, fracture healing delays, and osteonecrosis at the point of care, though motion sensitivity and coverage remain challenges to address.

Abstract

We propose an ultrasound approach which provides, with one single examination and one single device, access to three bone biomarkers: anatomy, tissue quality and blood flow. It unlocks ultrasound imaging inside bone by accounting for ultrasound wave speed heterogeneity and anisotropic wave refraction. This study reports the first \emph{in vivo} evaluation with a comparison to peripheral Quantitative Computed Tomography (pQCT) and modulations of blood flow. Anatomical multi-layer bone-corrected reconstruction was validated at the tibia of healthy volunteers against pQCT and showed agreement on bone cortex interfaces. Estimation of axial and radial ultrasound wave speeds in cortical bone tissue (i.e. along the tissue symmetry axis and normal to it) demonstrated good reproducibility and positive correlation with bone mineral density measured by pQCT. Pulsatile blood flow was mapped and quantified in cortical and medullary regions. A directional ray selection method was developed to enhance blood signal extraction by reducing strong specular reflections originating from the outer and inner surfaces of the bone cortex. Physiological and non-physiological modulations of blood flow, namely head-up/head-down tilt table maneuvers and arterial occlusions, demonstrated the method sensitivity to blood flow variations. For the first time, reactive hyperemia was observed inside bone cortex. These results demonstrate the feasibility of a portable, non-ionizing, and quantitative ultrasound approach for structural, anatomical, and vascular characterization of bone tissue. This approach may offer new diagnostic capabilities for bone disorders, for instance osteoporosis, delayed fracture healing or osteonecrosis.

Figures (6)

• Figure 1: Simulated point spread functions (PSF) in a medium mimicking a longitudinal view of the tibia diaphysis. (a) Simulation setup showing the simulated medium containing two point targets. (b) Raw data for transmission with element #48, plotted as a function of time and the receivers positions. (c) Homogeneous ray tracing scheme and (d) corresponding B-mode image obtained. Artefacts resulting from mode-converted shear waves (SW) are pointed with white arrows. (e) Multi-layered ray tracing assuming an isotropic bone layer and (f) corresponding image. (g) Multi-layered ray tracing assuming an anisotropic bone layer and (h) corresponding image.
• Figure 2: (a) View of the pQCT scanner used to acquire transverse reference images of the tibia. (b) Example of a pQCT slice showing the tibial and fibular cortices. (c) Schematic representation of the transverse imaging plane. (d) Ultrasound B-mode image reconstructed under the assumption of a homogeneous medium and (e) using the proposed multi-layered model including lens, skin, anisotropic cortical bone, and marrow. (f) Cortical thickness measured on the ultrasound images as a function of cortical thickness measured on pQCT slices. (g–j) Superposition of pQCT and ultrasound images for four different volunteers. The outer cortical interface (periosteum) segmented from the ultrasound image is overlaid in cyan on the grayscale ultrasound image.
• Figure 3: In vivo measurement of radial and axial wave speed in the tibial cortex. (a–e) Estimation of radial wave speed using auto-focusing (a) and axial wave speed using head-wave tracking (e). (b–f) Reproducibility between visits. (c–g) Wave speed measurements across two age groups. (d–h) Correlation between ultrasound wave speed and bone mineral density (BMD) from pQCT.
• Figure 4: Ray selection method and intraosseous blood flow imaging. (a) Ray tracing strategy for the PSF tilted on the right side, and (b) the corresponding image. (c) Ray tracing strategy for the PSF tilted on the left side, and (d) the corresponding image. (e) Final image obtained using ray selection. (f) Image obtained using the aberration correction framework presented above. (g) In vivo Power Doppler image of the tibial cortex assuming an homogeneous wave speed model, (h) applying the aberration correction and (i) applying the ray selection and aberration correction methods. Movie S1 provided as a supplementary material shows the temporal evolution of panel (i) and plots of Power Doppler signals over time in some specific regions..
• Figure 5: Modulation of intraosseous blood flow using Power Doppler imaging. (e) Timeline of the occlusion experiment with Doppler acquisitions. (f) Power Doppler (PD) signals during occlusion and hyperemia for one volunteer. Dashed lines show the 90$^{th}$ percentile. (g) Summary of the occlusion experiment results across all volunteers. (h) Aggregated data showing median ratios between different experimental phases. (i) Timeline of the tilt experiment with Doppler acquisitions. (j) Power Doppler signals acquired in head-up (HU) and head-down (HD) positions for one volunteer. (k) Summary of the tilt experiment results across all volunteers. (l) Aggregated data showing the overall effect of tilt on intra-osseous blood flow.