Multi-channel A-mode ultrasound can be used to track bone motion without radiation. However, using multiple transducers increases setup complexity, while unreliable or missing signals may reduce tracking accuracy. This project investigates whether missing ultrasound channels can be reconstructed from a smaller set of available measurements.
Using an existing dataset of synchronized ultrasound signals, transducer poses, and reference bone motion, the student will develop a learning-based method to predict missing channels—for example, reconstructing two additional channels from six measured channels. The method will explore spatial relationships between transducers and temporal information from ultrasound sequences.
The reconstructed channels will be integrated into an existing bone-tracking pipeline. Performance will be compared with both the complete-channel system and a system using only the available channels, evaluating signal reconstruction, bone-depth estimation, and bone-motion accuracy.
Main activities
-
Process and analyse the existing multi-channel ultrasound dataset.
-
Develop and evaluate a virtual-channel reconstruction method.
-
Investigate its contribution to bone tracking under different channel configurations.
Expected background: Biomedical engineering, robotics, electrical engineering, or a related field; experience with Python and an interest in deep learning and medical ultrasound.