- The paper demonstrates that autonomous robotic ultrasonic sacral osteotomy (RUSO) achieves sub-millimeter (0.11 mm) trajectory accuracy, compared to clinical manual techniques associated with 2–4 mm errors and more.
- %This suggests RUSO enables safer disruption of sacral tumors close to spinal nerves and other critical structures.
- The study reveals RUSO's autonomous cutting results better depth control, reducing over-penetration from about 75% to nearly 5—, precise and reliable regulation, and a marginal performance variability.
Clinical motivation
En bloc sacrectomy remains the only curative treatment for chemoresistant sacral tumors such as chordoma and chondrosarcoma. The osteotomy path must be executed with sub-millimeter precision because it lies in close proximity to the S1–S4 nerve roots, whose injury causes irreversible loss of bowel, bladder, and sexual function; the procedure also carries high risks of massive hemorrhage and infection. Computer-assisted navigation (CAN) improves outcomes—Bosma et al. reported 81% negative-margin rates versus 50% conventionally—but execution remains manual and operator-dependent. Across published CAN-guided sacral and pelvic osteotomies, mean execution errors of 2–4 mm persist, with documented maximum deviations up to 18.5 mm and tumor-directed errors approaching 20 mm. The paper's central premise is that these deviations stem from manual handling of the osteotome itself, motivating fully autonomous tool control.
System architecture
The proposed RUSO system integrates three components: a seven-DoF KUKA LBR Med 14 manipulator holding a Misonix BoneScalpel ultrasonic osteotome, and an NDI Polaris Vega optical tracking system used both for calibration/registration and for recording executed trajectories. All components are registered into the robot base frame {S} through a standard calibration pipeline:
- Hand-eye calibration: an AX=XB formulation solves jointly for the tracker-to-base transform and the end-effector-to-tracking-rigid-body transform from N robot poses.
- Pivot calibration: localizes the osteotome tip relative to the tracked rigid body by pivoting the tip in a fixed divot.
- Tip calibration: recovers the tip-to-end-effector transform using an optically tracked digitizer.
Once calibrated, the operator registers the cut start point on the phantom via digitization, after which the robot autonomously executes insertion, translation along the planned path at constant speed, and retraction, repeated in passes to reach target depth.
Experimental protocol
Comparisons were performed on PCF 15 Sawbones phantoms under identical conditions: a clinically derived straight 100 mm transverse trajectory (consistent with reported mean adult sacral widths of 84–126 mm), at target depths of 4 mm (representative cortical transection) and 8 mm (cortical plus cancellous). Manual trials (MUSO) were performed by a trained operator standing at a workbench, allowed multiple passes; robotic trials (RUSO) ran autonomously at controlled cutting speeds of 1–3 mm/s. Four metrics were evaluated: trajectory RMSE against the ideal line, physically measured cut length, active cutting time, and achieved depth computed as a per-bin minimum-z profile over discretized trajectory bins—an appropriate treatment of multi-pass cuts that avoids misleading simple averages.
Quantitative results
| Condition |
Target depth (mm) |
Speed (mm/s) |
RMSE (mm) |
Length (mm) |
Time (s) |
Depth (mm) |
| MUSO M1 |
4 |
1.68 ± 0.15 |
1.10 ± 0.26 |
97.4 ± 1.2 |
58.25 ± 4.86 |
7.0 ± 0.8 |
| MUSO M2 |
8 |
0.91 ± 0.19 |
0.93 ± 0.21 |
97.9 ± 1.5 |
111.50 ± 27.16 |
15.9 ± 2.2 |
| RUSO R1–R3 (4 mm) |
4 |
1–3 |
0.10–0.13 |
100.9–101.6 |
44–127 |
4.20–4.21 |
| RUSO R4 (8 mm) |
8 |
3 |
0.09 ± 0.05 |
102.2 ± 0.1 |
90.33 ± 0.58 |
8.06 ± 0.01 |
Three findings dominate. First, trajectory accuracy: RUSO achieved 0.11 mm mean RMSE, an order-of-magnitude improvement over MUSO (0.93–1.10 mm). Second, and most consequential, depth control: the manual operator over-penetrated by 75% on the 4 mm task (7.0 mm achieved) and doubled the target on the 8 mm task (16.0 mm achieved), attributed to excessive applied force during multi-pass cutting; the robotic system held depths within roughly 5% of target (4.2 mm and 8.1 mm) independent of speed. Given that the anterior sacral cortex is only millimeters thick, this magnitude of over-penetration would directly threaten nerve roots in vivo. Third, efficiency: RUSO completed the 4 mm task deterministically in 45 s at 3 mm/s with no accuracy degradation, whereas manual timing was variable and slower for the deeper task (≈111 s) despite producing visibly wider, jagged cuts.
Limitations
The authors appropriately frame this as a preliminary study, and several constraints bear directly on interpretation. All experiments used homogeneous polyurethane Sawbones phantoms rather than cadaveric or clinical bone, so the depth-control result has not yet been demonstrated against realistic, variable sacral anatomy or soft-tissue interaction forces. The trajectory is a single straight line, whereas clinically required sacrectomy paths are non-linear and curve around structures such as the sacroiliac joint; performance on such paths is untested here. The comparison involves a single experienced operator over a small number of trials (four manual, three robotic per condition) without formal statistical testing, limiting generalization about the manual baseline. Finally, the autonomous framework relies on preoperative point registration and open-loop execution; intraoperative sensing, replanning, and force feedback are not addressed in this study.
Conclusion
This paper provides quantitative evidence that an autonomous robotic system coupling a seven-DoF manipulator with an ultrasonic osteotome can substantially outperform manual execution on a representative sacral osteotomy task in phantoms: tenfold better trajectory accuracy, near-exact depth fidelity where manual technique over-penetrated by up to 100%, and deterministic procedure times. The results establish a foundation for robotic sacrectomy while leaving open validation on anatomically realistic specimens, complex non-linear trajectories, and closed-loop intraoperative control.