In this study was assessed the precision and accuracy of a novel arthroplasty navigation tool. On-Tool Tracking (OTT) is an innovative on-board wireless device for 3D tracking using miniaturized active infrared LED reference frames. It combines proprietary hardware, software and firmware to acquire and process stereo images to track objects in 3D. OTT seeks to address three basic problems encountered in arthroplasty navigation: inconvenient cameras-markers line-of-sight, large OR footprint and high cost. This study tackles the challenging problem of how to experimentally align, independently measure and present the static 3D position of the OTT relative to its tracked target.

Static accuracy was measured by traversing the OTT over a 3D grid covering the tracking volume [Fig. 1] using an MTS 858 Bionix 5-axis test machine, with a working volume of 100×55.0×76.2 [mm] [Fig. 2]. The absolute position errors were estimated from the MTS actuated/measured versus the OTT recorded X,Y,Z coordinates. First, we registered the OTT coordinate system to that of the MTS, using a point-to-point algorithm which yielded a best-fit OTT-to-MTS 3D transformation. The data set comprised 637 points/locations; with 30 samples collected/averaged at each location. The positional error was the Euclidean (scalar) distance between the reference and measured positions. The RMS, mean, standard deviation, 95% confidence interval, and maximum error were calculated for the whole 3D volume along with three XY planes-of-interest within that volume (at 100, 130, and 160mm OTT-to-reference-frame distances). Initial calibration of the OTT stereo vision rig was made on a totally different and independent physical setup.

Table-1 summarizes the 3D errors for three XY planes-of-interest and the entire volume. The histogram in Fig.3 shows the 3D error distribution. The RMS errors increased with the OTT-to-reference-frame distance. To determine whether the error source was potentially a “scaling” problem, we decoupled the 3D error into individual axis errors [Fig.4]. The summary for all planes is shown on the chart of Fig. 5a. Fig. 5 depicts the directional errors contributed by each axis. Overall results for the OTT show a mean static accuracy of 0.481±0.253 [mm].

The results validated the static accuracy of our overall system, to sub-millimeter averages throughout, but reaching >1mm at the extremes of the measuring volume. Our errors propagated from uncertainty in registration and errors in rigid-body detection rather than just the error of localizing a single retro-reflective marking sphere or LED, as many vendors quote. This study also demonstrates the correlation of the error with the OTT-to-reference-frame (perpendicular) distance and with the proximity of the reference frame to the image edges. The error was expectedly highest in the Z-direction. The errors were mostly uniform within a given XY plane; but increased when the reference frame approached the edges of a captured image. The OTT uses very wide-angle lenses, and so the image distortion/aberration correction algorithms could never be perfect. However, the errors at the distances where the actual surgical cuts would be made (≤ 145 mm) are comparable to today's state-of-the-art systems, even with this highly compact and utilitarian technology.

Introduction

Computer aided surgery aims to improve surgical outcomes with computer guidance. Navigated Freehand bone Cutting (NFC) takes this further by eliminating the need for cumbersome mechanical jigs, while decreasing cutting time and complexity. To reduce the footprint of the NFC tracking system (currently NDI Polaris) we designed and implemented “On-Tool Tracking” (OTT), a novel miniaturized tracking system that mounts onto the cutting instruments (Fig. 1). This study investigates the accuracy of the 3D-measurements of the OTT system.