Abstract
Accurate identification of the location the vertebra and corresponding pedicle is critical during pedicle screw insertion for percutaneous spinal fusion surgery. Currently, two dimensional (2D) fluoroscopy based navigation systems have extensive usage in spinal fusion surgery. Relying on 2D projection images for screw guidance results in high misplacement rates. Furthermore, fluoroscopy-based guidance exposes the surgical staff and patient to harmful ionizing radiation. Real-time non-radiation-based ultrasound (US) is a potential alternative to intra-operative fluoroscopy. However, accurate interpretation of noisy US data and manual operation of the transducer during data collection remains a challenge. In this work we investigate the potential of using multi-modal deep convolutional neural network (CNN) architectures for fully automatic identification of vertebra level and pedicle from US data. Our proposed network achieves 93.54% vertebra identification accuracy on in vivo US data collected from 27 subjects.
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call