Matrix a-mode ultrasound configuration for reliable and accurate bone surface detection
Dennis Christie (ET-BE), Rene Fluit (University of Groningen), Guillaume Durandau (McGill University), Massimo Sartori (ET-BE), Nico Verdonschot (ET-BE).
Abstract
A combination of Amplitude-mode (A-mode) ultrasound (US) and motion capture offers a new possibility to estimate 3D bone pose and kinematics in a non-invasive and non-radiative manner. The proposed method employs US transducers on the skin to measure the depth (skin-to-bone distance), while motion capture tracks their 3D pose. However, detecting bone surface from an A-mode’s 1D US signal is a challenging task. The conventional method for selecting the bone peak relies on the subjective judgment and experience of the user. One possible solution to address this issue is to extend the spatial information within the US signal by utilizing a 2D matrix configuration which will provide a useful spatial context in identifying bone peak. The purpose of this study is to investigate the advantages of this configuration over a single transducer in bone surface detection. An experiment was conducted with a human cadaver's right lower leg. US measurements with 3x3 A-mode matrix configuration and CT scans were performed simultaneously to obtain the depth estimation and the ground truth, respectively. The matrix configuration enables us to leverage data from the adjacent transducers and apply surface estimation to identify and correct the misidentified bone surface point to some extent. The result shows that estimation with surface yields statistically significant (p<0.05) lower depth errors compared to using a single transducer; from 0.62 to 0.45 mm absolute median error, mostly by removing anomaly detections. This method can be applied automatically by this configuration hence eliminates reliance on user subjectivity.