Open Collections

An intraoperative position assessment system for pedicle screw insertion surgeries

University of British Columbia

Pedicle screw fixation is a common practice in spinal surgeries. Metal implants are inserted into the structure of interest (pedicles) to provide stabilization. The major surgical challenge is to ensure that the implants are within a narrow safe zone otherwise serious postoperative complications or potential revision surgeries can occur. Currently, implant malplacement rates are rather high resulting in a considerable rate of revision surgery associated with this operation. The most common technique for verifying the position of the inserted implants during the surgery is to acquire intraoperative X-rays and visually assess these two dimensional images to estimate the three-dimensional position of each implant. This is an erroneous process due to the lack of the third dimension and can result in misjudgment. In this thesis, we developed an accurate and automated system for intraoperative post-placement) pedicle screw position assessment. To this end, we first designed a semi-automatic pipeline that consisted of the main computational modules required for such a system and evaluated its performance in identifying screw breaches. After demonstrating satisfactory performance (average root mean square error better than 0.8 mm and 1.3 degrees), we focused on improving the autonomy of the system by incorporating more advanced processing modules with the aim of reducing interference with the existing surgical workflow and ultimately facilitating clinical adoption. For this purpose, we developed automatic and application-specific approaches for the following system modules: intraoperative X-ray calibration (average error 0.4 mm), spinal 2D-3D registration (capture range for a two view setup ∼ 12.5 mm and 31 degrees), implant segmentation and pose estimation (errors less than 2 degrees and 2 mm), X-ray inpainting (up to 85% increase in registration capture range) and preoperative pedicle region localization (on average 90% accuracy), which all helped us in achieving the end goal of system autonomy. The performance of individual modules were validated on clinical or clinically realistic data. This system can potentially improve the ability of a surgeon to assess pedicle breaches intraoperatively and decide whether or not a screw needs to be repositioned before closing the patient; therefore, potentially reducing the current rate of revision surgeries.

Text

2020-05-20

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/74541

Biomedical Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/