Haptic Robot-assisted Surgery Improves Accuracy of Wide Resection of Bone Tumors: A Pilot Study

Abstract

Background

Accurate reproduction of the preoperative plan at the time of surgery is critical for wide resection of primary bone tumors. Robotic technology can potentially help the surgeon reproduce a given preoperative plan, but yielding control of cutting instruments to a robot introduces potentially serious complications. We developed a novel passive (“haptics”) robot-assisted resection technique for primary bone sarcomas that takes advantage of robotic accuracy while still leaving control of the cutting instrument in the hands of the surgeon.

Questions/Purposes

We asked whether this technique would enable a preoperative resection plan to be reproduced more accurately than a standard manual technique.

Methods

A joint-sparing hemimetaphyseal resection was precisely outlined on the three-dimensionally reconstructed image of a representative Sawbones femur. The indicated resection was performed on 12 Sawbones specimens using the standard manual technique on six specimens and the haptic robotic technique on six specimens. Postresection images were quantitatively analyzed to determine the accuracy of the resections compared to the preoperative plan, which included measuring the maximum linear deviation of the cuts from the preoperative plan and the angular deviation of the resection planes from the target planes.

Results

Compared with the manual technique, the robotic technique resulted in a mean improvement of 7.8 mm of maximum linear deviation from the preoperative plan and 7.9° improvement in pitch and 4.6° improvement in roll for the angular deviation from the target planes.

Conclusions

The haptic robot-assisted technique improved the accuracy of simulated wide resections of bone tumors compared with manual techniques.

Clinical Relevance

Haptic robot-assisted technology has the potential to enhance primary bone tumor resection. Further bench and clinical studies, including comparisons with recently introduced computer navigation technology, are warranted.

Appendix

Calculation of various parameters of accuracy

We used reverse engineering software (Version 12.0; Studio, Geomagic, Inc, Research Triangle Park, NC, USA) to analyze the pre- and postresection three-dimensionally reconstructed laser scan images. The preoperative image contained the three target planes of the resection, ie, the three planes that were defined as part of the preoperative plan as the superior, inferior, and vertical cut planes. For each specimen, the standard best alignment function of the software was used to perfectly superimpose the preoperative and postoperative images of the Sawbones specimen; this placed the preoperative and postoperative scans in a common coordinate system. We then used the software to highlight the cortical rim of each of the three cuts, deliberately excluding the cancellous surfaces to avoid the analysis problems associated with the cancellous voids of the cancellous surface. (These voids resulted in the software selecting points not only on the surface of the cancellous bone, but also in the depths of the voids, the latter of which were clearly not in the plane of the cut.) We believed this technique was well justified, because it was unlikely that a horizontal saw blade would produce cuts in the cancellous surface that deviated wildly from the plane of the cuts defined by just the cortical rims. The small values recorded for the deviation from flatness for both the robotic and manual cuts (see Results section) support this assumption.

The cortical rims of each of the three cuts—superior, inferior, and vertical—were exported as discrete points (relative to the common coordinate system created after the preoperative and postoperative images were superimposed) and imported into a technical computing software program (Matlab; MathWorks, Natick, MA, USA) for further analysis. This software was used to help analyze the various geometric relationships between points on the resection plane (and the best-fit plane) and the target planes. Using this software, the perpendicular distance between each point and the corresponding target plane was calculated. Additionally, the angular deviations between the target plane and best-fit planes were calculated.

The data were reported in a form that was consistent with the standards of the International Organization of Standardization (ISO) (Fig. A-1) [13]. According to these standards, the location error is defined as the perpendicular distance from the target plane to the point on the cut surface furthest from the target plane. For simplicity, in the text we usually refer to location error as the maximum deviation from the preoperative plan. Although it is not an ISO standard, we also calculated the mean deviation of the points from the target plane by calculating the mean of the (absolute) distance of each point from the target plane. The flatness was calculated by measuring the amplitude, or peak-to-peak distance, of the cut surface.

Fig. A-1

Location errors, flatness, and parallelism.

The pitch and roll errors were defined as angular deviations between the best-fit plane and the target plane along the axis of the blade and the front edge of the blade, in that order. Specifically, the roll error refers to a rotation about the AP axis of the bone; the pitch error refers to a rotation about the superoinferior axis for the medial cut and the mediolateral axis for the superior and inferior cuts. For the best-fit planes through the superior and inferior limbs of the cuts, the deviation from parallel was defined as the angle between these two planes.