Automatic breach detection during spine pedicle drilling based on vibroacoustic sensing

Area of Science:

• Spinal Surgery
• Surgical Robotics
• Biomedical Engineering

Background:

• Pedicle drilling is critical in spinal surgery, with breaches risking damage to adjacent nerves and vessels.
• Current image-guided methods are radiation-intensive and rely on pre-operative data.
• Accurate intraoperative breach detection is essential for patient safety.

Purpose of the Study:

• To develop and evaluate a novel, radiation-free algorithm for detecting cortical breaches during pedicle drilling.
• To utilize a non-visual sensor setup combined with deep learning for real-time breach detection.
• To assess the efficacy of various sensor combinations for optimal performance.

Main Methods:

• Integrated multiple vibroacoustic sensors (microphones, accelerometers) and an optical tracking system.
• Collected data from cadaveric human spines (L5-T10) during pedicle drilling.
• Developed an automatic labeling method and employed deep learning on mel-spectrograms for breach classification.

Main Results:

• Individual sensors achieved high breach recall: contact microphones (85.8%) and uni-axial accelerometers (81.0%).
• A data fusion model combining contact microphones and uni-axial accelerometers achieved a 98% breach recall.
• The proposed method demonstrated significant potential for intraoperative breach detection.

Conclusions:

• Non-visual sensor fusion with deep learning offers a promising radiation-free solution for detecting breaches during pedicle drilling.
• This technology can help prevent screw misplacement and accidental bone breaches in spinal surgery.
• The approach has potential for broader applications in other surgical procedures.