Fast and robust extraction of surrogate respiratory signal from intra-operative liver ultrasound images

Abstract

Our method can use 2D ultrasound to track natural landmarks from the liver as surrogate respiratory signal and hence provide a feasible solution to replace special tracking devices.

Subject-specific liver motion modeling in MRI: a feasibility study on spatiotemporal prediction

“…As mentioned before, a 4D dynamic sequence with a built-in navigator echo signal would probably be the optimal solution, but this is difficult to realize in practice. Alternatively, ultrasound tracking might also be used in practice for motion management (Harris et al 2010, Rubin et al 2012, Wu et al 2013, but this would pose specific requirements on the hardware. Using the respiratory signal as an indirect measurement of liver motion is the most practical solution.…”

Technical Note: Automatic real‐time ultrasound tracking of respiratory signal using selective filtering and dynamic template matching

“…Our early work 6 proposed an adaptive template matching (ATM) strategy with only adaption of the search center during the frame-by-frame matching process. In order to further improve efficiency and robustness of the template matching, we herein devise a new DTM method, which is characterized by three adaptive properties: dynamic template block, adaptive search center, and gradually expanding search range, where the latter two properties form a dynamic search window.…”

“…The underlying similarity metric is normalized cross correlation (CC) and the block size is fixed as 65 × 65 pixels for robust and accurate template matching. 6 2.B.1.a. Dynamic template block.…”

“…The methods extract the breathing signal using different motion estimation techniques, such as block similarity (BS), 3 manifold learning (ML), 4,5 and template matching (TM). 6,7 BS provides real-time signal tracking capability but is usually sensitive to image noises and artifacts and not directly relevant to breathing. Although extracting more accurate signal, ML is suitable for offline applications due to its batch-processing mode.…”

“…In this paper, we present an ultrasound-based method to automatically identify the liver boundary and in real-time extract its translational motion as the respiratory signal using a dynamic template matching (DTM) method. As extension of the previous work, 6,7 our new contributions include (1) a new dynamic template matching method to estimate the respiratory motion, which features dynamic template block and dynamic search window; (2) a novel signal extraction method to optimally extract accurate the respiratory signal from motion estimation results.…”