Abstracts Track 2023

Area 1 - Fundamentals

Nr: 5

Hight Accuracy 3D Tracking by Monocular Vision: An Application to Surgical Navigation of Humeral Intramedullary Nail


Christian Elmo and Pierre Vacher

Abstract: In displaced humerus fractures, surgeons may use osteosynthesis devices such as intramedullary nails to restore bone continuity and function to the shoulder. These nails have fixation areas embedded within the humerus and no longer visible during surgery. Usually, an external ancillary device guides the surgeon during these operations. These devices are expensive, single-use or in case of reusage, require specific cleaning, sterilisation and storage operations. In this work, we are interested in the navigation of a intramedullary nail of the proximal humerus by tracking a locator in monocular vision. The objective is to guide the surgeon so that he can find a non-visible drilling axis belonging to the intramedullary nail during a surgical operation. This device is based on Localized Object (LOOB) composed of an arrangement of planar Aruco markers. These markers are randomly positioned and glued on a volume of any geometry. A specific algorithm has been developed to find the relative position of the glued-markers in regards to each other and constitute the calibration step of the LOOB. This calibration is based on a bundle adjustment method where the initial guess is found using the graph theory. After this offline calibration step, the location of the LOOB and any object rigidly attached to it can be tracked in real-time by monocular vision. In our application, LOOBs have a double-decahedron geometry (h = 76 mm) to allow the observation of a significant number of markers (at least two non-coplanar markers) whatever the orientation and the position of the object in regards to the camera. One LOOB is rigidly attached to the nail (LOOB-n) and another to the drilling device (LOOB-d). The camera used has a resolution of 3.1 megapixel, it is positioned at about 1 m distance from the object and observes a field of about 1 m 3 . Adapted procedures are used to identifiy the various drilling axes necessary for the fixation of the centromedullary nail from LOOB-n and the drilling axis from LOOB-d. In real time, the positions and orientations of both drilling device and intramedullary nail are known. These informations are represented in the form of a virtual scene to the surgeon. Under the previously stated imaging conditions, the measurement uncertainties in terms of accuracy associated with the locator alone are on average 0.002 mm with a standard deviation of 0.04 mm in position and on average 0.01 ◦ with a standard deviation of 0.14 ◦ in rotation. Tests were also carried out to qualify the uncertainties in terms of accuracy after identification of an axis. Intramedullary nail implantation tests assisted by the presented navigation tool were carried out in the anatomy laboratory on cadaveric specimens. A comparison between the drilling locations performed during the operation andthose observed a posteriori using a 3D CT scan was performed. The results obtained in terms of uncertainties show great promise for in vivo surgeries.