ScanLite: Infrastructure-free ultrasound reconstruction fusing image processing and deep inertial tracking
PI: Mellon, Stephen
Department: NDORMS – Botnar
CAT&MAUS is a system that has been developed to reconstruct and track bony motion of joints in 3D space to allow clinicians to assess internal joint structures during dynamic movement. It currently uses a handheld ultrasound (US) probe which is tracked by a VICON motion capture (MOCAP) system, allowing 3-dimensional images to be built up according to the movement of the probe as it is manipulated by a sonographer.
The VICON system requires external cameras and reflective markers attached to the probe to track its movement. Although this system works well, its inclusion adds cost and limits portability, which have been identified as barriers to CAT&MAUS being clinically adopted. The aim of the current study is to develop a probe-tracking method that does not rely on infrastructure such as external cameras.
In addition to the ultrasound (US) transducer, many commercially available US probes now contain a sensor called an IMU (inertial measurement unit). This sensor can measure the probe rotations and accelerations but the signal from an IMU is insufficient for accurate tracking of the US probe position because it tends to drift and diverge away from the ground truth. The nature of this drift has been the subject of research in robotics. It has been found that by training artificial intelligence with the signal from an IMU and the ground truth data, it is possible to reduce the drift.