P3A-4 Compounding of Strain Data Non-Invasively Obtained at Large Beam Steered Angles

Abstract

Non-invasive vulnerable plaque detection is beneficial for prevention of stroke and myocardial infarction. A recent study [1] has shown, that non-invasive vulnerable plaque detection in the carotid arteries is feasible by ultrasound elastography. However, the quality of the strain images is low, because contrary to intravascular strain imaging, lateral strain input is required for reconstructing radial and circumferential strain images. This study focuses on reducing the distortion caused by the lateral component by constructing radial and circumferential strain images from mainly axial strain data acquired at multiple insonification angles. Beam steering techniques [2] were applied to achieve insonification angles of up to 45deg. Most of the compound radial and circumferential strain images were estimated by projecting axial strain data in radial direction. A segment calculated by principal component analysis was added to obtain full 360deg compound strain images. Compound radial and circumferential strain images were constructed for three gelatin vessel phantoms; a homogeneous phantom with a concentric lumen, a homogeneous phantom with an eccentric lumen, and a heterogeneous phantom with an eccentric soft plaque. Three radial and circumferential strain images were constructed for the concentric phantom. The first was constructed by principal component analysis of 0deg data only. The second consisted of data obtained at -45, 0 and 45 degrees beam steering, and the third was made up of data acquired at -45, -30, -22, -15, 0, 15, 22, 30, and 45 degrees beam steering. Assuming a radial strain decay, elastographic signal-to-noise ratio's (SNR <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e</sub> ), and contrast-to-noise ratio's (CNR <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e</sub> ) of these images were determined and compared. Finally, compound radial and circumferential strain images of the two eccentric phantoms were constructed from data acquired at the complete range of insonification angles, and compared. The image quality of the radial and circumferential compound strain images improved considerably compared to the radial and circumferential strain image constructed from 0deg information only. SNR <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e</sub> and CNR <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e</sub> , increases of up to 4.4 dB and 9.9 dB were measured respectively. Image quality of the elastogram improved with the number of beam steered angles used for the compounding. By comparing the strain images of the two eccentric phantoms the soft plaque region could be identified.