An Intrinsic Shape Estimation Algorithm for Flexible Ultrasound Probes Intended for Clinical Applications

Recent advances in ultrasonic transducer technologies have made it possible to build flexible ultrasound arrays that can conform to body parts. However, in order to perform beamforming and image reconstruction, it is necessary to estimate the transducer shape. While external shape measurement devices could be used, we propose here an algorithm that uses intrinsic characteristics of reconstructed ultrasound images to iteratively estimate the array shape. The algorithm maximizes an image focus metric based on gradient energy in the region of interest. We simulated three realistic shoulder sonograms created by a relatively large 98-mm, 384-element ultrasound array with varying profile shapes representing anatomically realistic curved array shapes (radii of curvature: 62–125 mm). The average relative position estimation error was less than 1% across the nine studied cases, and the median Complex Wavelet Structural Similarity index between reference and reconstructed images was 98% (range: 84%-99%). The proposed method could facilitate implementing flexible ultrasound probes in clinical settings.