A dual-modal fiducial marker with single-view x-ray pose recoverability for MR-guided surgical navigation
Year: 2026
Type: article
Abstract: In orthopedic surgery, accurate intraoperative localization of implants and instruments is essential. Traditional surgical navigation markers rely on a fixed external reference frame, but in mixed reality (MR) surgical navigation the tracking camera moves with the surgeon’s head, making conventional markers unsuitable. This study proposes that markers for MR navigation must satisfy three design constraints: omnidirectional recognizability, X-ray recognizability, and dual-modal compatibility. Based on these constraints, we designed a marker consisting of a central cube and four cylindrical axes with unequal diameters, and proved analytically that its six-degree-of-freedom (6-DoF) pose can be determined from a single X-ray image. A pose estimation algorithm combining geometric optimization and deep learning-based feature extraction was used to validate dual-modal pose estimation, in which the marker pose is estimated independently from a single X-ray or optical image. On 100 X-ray and 194 optical images, the X-ray recognition error was 1.58° ± 1.26° (coronal plane) and 1.90° ± 1.50° (sagittal plane), with 94% of samples achieving a composite angular error below 5°; the optical recognition success rate was 90.2%, with an overall angular error of 1.52° ± 0.98°. Error distributions were symmetric and zero-centered (signed means within ± 0.02°); X-ray pose estimation remained solvable even at extreme sagittal tilt (up to ~ 40°), while rare optical failures (2.3%) could be reliably detected. These results provide preliminary evidence of dual-modal marker pose estimation from single-view X-ray and optical images; further validation is required before clinical translation.
Source: Scientific Reports
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,690
Award: G6012327474
OpenAlex ID: W7167700533
