Related Experiment Video
Updated: Jul 10, 2025
06:24
A Spine Robotic-Assisted Navigation System for Pedicle Screw Placement
Published on: May 11, 2020
8.9K
Automatic registration with continuous pose updates for marker-less surgical navigation in spine surgery
Florentin Liebmann1, Marco von Atzigen1, Dominik Stütz2
1Research in Orthopedic Computer Science, Balgrist University Hospital, University of Zurich, Zurich, Switzerland; Laboratory for Orthopaedic Biomechanics, ETH Zurich, Zurich, Switzerland.
Medical Image Analysis
|November 22, 2023
Summary
This study introduces a novel marker-less system for radiation-free pedicle screw placement during spinal fusion. The approach uses deep learning and augmented reality for accurate, automatic registration and real-time surgical guidance.
Area of Science:
- Medical Imaging
- Computer-Assisted Surgery
- Deep Learning
Background:
- Established pedicle screw placement systems have accuracy but suffer from time-consuming, radiation-exposed registration and non-intuitive guidance.
- Current methods lack in-situ, surgeon-centric feedback, highlighting the need for radiation-free, automated registration and navigation.
Purpose of the Study:
- To develop and validate a marker-less, radiation-free system for automatic registration and augmented reality-guided navigation in lumbar spinal fusion.
- To address the limitations of conventional surgical navigation systems.
Main Methods:
- A deep neural network was trained for automatic lumbar spine segmentation and orientation prediction.
- The system refines vertebral pose in real-time using GPU acceleration and integrates with an augmented reality navigation system for surgeon-centric guidance.
- A marker-less approach was employed to eliminate the need for intraoperative markers.
Main Results:
- Median successful registrations of 100% on a public dataset, with median errors of 2.7 mm (target registration), 1.6° (screw trajectory), and 2.3 mm (screw entry point).
- Ex-vivo validation demonstrated 100% screw accuracy and a median target registration error of 1.0 mm.
- The system successfully handled surgeon occlusions during real-time updates.
Conclusions:
- The developed marker-less, deep learning-based system offers accurate, radiation-free registration and augmented reality guidance for lumbar spinal fusion.
- Results meet clinical demands, showcasing the potential of RGB-D data for automated registration and AR guidance in surgery.
