Joint Beamforming and Feature Detection for Enhanced Visualization of Spinal Bone Surfaces in Ultrasound Images

TL;DR

This paper introduces a novel ultrasound imaging framework combining eigenspace minimum variance beamforming and phase symmetry detection to improve the visualization and localization of spinal bone surfaces, achieving sharper boundaries and reduced noise.

Contribution

The study presents a new combined beamforming and feature detection approach that enhances bone surface visualization in ultrasound images, outperforming standard methods in clarity and noise reduction.

Findings

Bone localization error comparable to standard methods

20% reduction in intensity distribution variability around bone surface

60% noise reduction inside bone shadow

Abstract

We propose a framework for extracting the bone surface from B-mode images employing the eigenspace minimum variance (ESMV) beamformer and a ridge detection method. We show that an ESMV beamformer with a rank-1 signal subspace can preserve the bone anatomy and enhance the edges, despite an image which is less visually appealing due to some speckle pattern distortion. The beamformed images are post-processed using the phase symmetry (PS) technique. We validate this framework by registering the ultrasound images of a vertebra (in a water bath) against the corresponding Computed Tomography (CT) dataset. The results show a bone localization error in the same order of magnitude as the standard delay-and-sum (DAS) technique, but with approximately 20% smaller standard deviation (STD) of the image intensity distribution around the bone surface. This indicates a sharper bone surface detection. Further, the noise level inside the bone shadow is reduced by 60%. In in-vivo experiments, this framework is used for imaging the spinal anatomy. We show that PS images obtained from this beamformer setup have sharper bone boundaries in comparison with the standard DAS ones, and they are reasonably well separated from the surrounding soft tissue.