Trackerless 3D Ultrasound Stitching for Computer-Assisted Orthopaedic Surgery and Pelvic Fractures

Abstract

In pelvic fracture surgeries, percutaneous screws must be placed accurately for effective fixation and to prevent damage to surrounding tissue structures. Fluoroscopy is currently used to image the pelvis to provide guidance, but this produces harmful ionising radiation and does not allow three-dimensional (3D) visualisation. Ultrasound offers three-dimensional, non-ionising, real-time, and inexpensive imaging. It has thus emerged as an alternative to fluoroscopy for intraoperative imaging in computer-assisted orthopaedic surgery (CAOS). However, ultrasound-based surgical guidance is challenging because ultrasound produces inherently noisy images with limited field-of-view. While several techniques have been proposed to improve bone clarity in ultrasound scans, there is limited work on enhancing ultrasound’s field-of-view for CAOS. In particular, improving the field-of-view for surgical guidance for pelvic fracture surgeries would be needed to achieve accurate and reliable registration to preoperative data, and accurate screw placement in the pelvis.We propose and evaluate the feasibility of a trackerless method for stitching volumetric ultrasound to achieve an extended field-of-view. Stitching is performed using corresponding features in the overlap between three ultrasound volumes, extracted using an implementation of the 3D scale-invariant feature transform. The volumes are processed using confidence-map weighted phase symmetry detection. Alignment between the volumes is calculated using coherent point drift rigid registration.We succeeded in extending the field-of-view of 3D ultrasound by creating a 39×43×115mm volume from three initial overlapping volumes, with reasonable overall accuracy. We show a mean post-registration surface error of 0.54mm, compared to 0.33mm achieved by previous tracking-based stitching. Our method achieved a mean distance error of 5.1%, compared to 2% in a similar tracked and 3D SIFT-based technique. Our stitching method does not use tracking, thus contributing towards simpler surgical navigation.

Exploiting multimodal CNN architecture for automated teeth segmentation on dental panoramic X-ray images

“…The threshold-based segmentation method suffers in identifying root boundaries. The level set methods [26][27][28] were constructed using different boundary-based methods to perform tooth segmentation. These methods can deal with sharp curves and corners, and variations of teeth shapes but their computation complexity is high and often trapped in local minima.…”

Recent Trends, Technical Concepts and Components of Computer-Assisted Orthopedic Surgery Systems: A Comprehensive Review

“…3D fluoroscopy represents a relatively new technology; it allows for 3D reconstruction by using multiple intraoperative fluoroscopy images. Unfortunately, this method requires an increased dose of radiation and patient exposure [142,143,144]. Compared with 2D fluoroscopy, 3D fluoroscopy gives better image quality.…”