Premium
SU‐D‐210‐05: The Accuracy of Raw and B‐Mode Image Data for Ultrasound Speckle Tracking in Radiation Therapy
Author(s) -
O'shea T,
Bamber J,
Harris E
Publication year - 2015
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4923920
Subject(s) - speckle pattern , imaging phantom , frame rate , match moving , physics , ultrasound , signal (programming language) , optics , computer vision , artificial intelligence , computer science , acoustics , motion (physics) , programming language
Purpose: For ultrasound speckle tracking there is some evidence that the envelope‐detected signal (the main step in B‐mode image formation) may be more accurate than raw ultrasound data for tracking larger inter‐frame tissue motion. This study investigates the accuracy of raw radio‐frequency (RF) versus non‐logarithmic compressed envelope‐detected (B‐mode) data for ultrasound speckle tracking in the context of image‐guided radiation therapy. Methods: Transperineal ultrasound RF data was acquired (with a 7.5 MHz linear transducer operating at a 12 Hz frame rate) from a speckle phantom moving with realistic intra‐fraction prostate motion derived from a commercial tracking system. A normalised cross‐correlation template matching algorithm was used to track speckle motion at the focus using (i) the RF signal and (ii) the B‐mode signal. A range of imaging rates (0.5 to 12 Hz) were simulated by decimating the imaging sequences, therefore simulating larger to smaller inter‐frame displacements. Motion estimation accuracy was quantified by comparison with known phantom motion. Results: The differences between RF and B‐mode motion estimation accuracy (2D mean and 95% errors relative to ground truth displacements) were less than 0.01 mm for stable and persistent motion types and 0.2 mm for transient motion for imaging rates of 0.5 to 12 Hz. The mean correlation for all motion types and imaging rates was 0.851 and 0.845 for RF and B‐mode data, respectively. Data type is expected to have most impact on axial (Superior‐Inferior) motion estimation. Axial differences were <0.004 mm for stable and persistent motion and <0.3 mm for transient motion (axial mean errors were lowest for B‐mode in all cases). Conclusions: Using the RF or B‐mode signal for speckle motion estimation is comparable for translational prostate motion. B‐mode image formation may involve other signal‐processing steps which also influence motion estimation accuracy. A similar study for respiratory‐induced motion would also be prudent. This work is support by Cancer Research UK Programme Grant C33589/A19727.