Real-Time Speed of Sound Estimation for Point-of-Care Tissue Health Assessment.

Abstract

Speed-of-Sound (SoS) is a fundamental acoustic property that emerging Ultrasound (US) modalities aim to leverage for tissue health assessment and image quality improvement. Tissue SoS has been tied to tumor malignancy classification, muscle health assessment, steatotic liver classification and bone porosity measurement among other. Consequently, leveraging the tissue SoS for more accurate beamforming, not only enables higher resolution imaging, leading to more accurate cyst classification, but also correct for heavy skull aberration and defocusing during High Intensity Focused Ultrasound. Directly measuring the tissue SoS requires multi-site access, thus limiting such methods to the breast and some limbs. Current Pulse-echo SoS estimation algorithms demand high computation time or provide a single global SoS value, either constraining real-time assessment or decreasing accuracy. In this dissertation, I developed: 1) A single-shot SoS estimation algorithm that estimates the global SoS of the media by leveraging the signal consistency across channels from a single transmitting event. 2) Multiple localized SoS estimators that leverage image dissimilarity from multiple transmission events to provide the user with the SoS of the segmented regions in the image, either in stratified media, or arbitrary configurations. By developing custom image formation, segmentation, raytracing and wavefront tracking frameworks, optimal transmission schemes, and GPU acceleration on a portable, research scanner, I’m able to provide robust, accurate, SoS estimation platforms that have been thoroughly validated in vitro, ex vivo and in vivo. This dissertation research aims to bridge the gap between SoS research and other US modalities that can benefit from the SoS information, as well as the gap between the low-level-research side that develops methods for robust US tissue assessment, and the clinical research side that tracks and relates such features to conditions of clinical relevance. With the single-shot algorithm, I provide non-SoS researchers with valuable information to increase the accuracy of their algorithms without the need of long transmitting times or high computational demands, crucial for flow or real time imaging. With the other algorithms, I provide robust avenues for non-invasive tissue health assessment, as well as arming clinicians with access to intrinsic tissue features, that can be used for real time monitoring or for subsequent research that arise from the insights clinicians gain with this novel tool.