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

A research team from ETH Zurich, led by Roman Denkin and Orcun Goksel, has developed a novel, image-based approach to a fundamental problem in medical ultrasound: accurately estimating the speed-of-sound (SoS) within tissue. Conventional ultrasound systems typically assume a single, average SoS value for all patients, which degrades image clarity and diagnostic accuracy when tissue properties vary. The new method sidesteps complex physics-based models by applying eleven conventional image analysis metrics—categorized for quality, similarity, and variation—directly to standard B-mode ultrasound images or post-beamformed data. This eliminates the need for proprietary raw channel data, making the technique widely accessible.

In simulations and phantom experiments, differential image comparison metrics, particularly mutual information and correlation, proved most robust. They achieved estimation errors consistently under 8 meters per second, even when processing relatively small image patches, enabling focal estimation. The study demonstrated clinical utility with an in vivo application for breast density classification based on derived SoS maps. By offering a data-accessible and computationally efficient alternative to existing methods, this research paves the way for retrofitting existing ultrasound machines with more accurate, patient-specific imaging capabilities, potentially improving diagnostic outcomes without requiring new hardware.