The first, robot-assisted osteotomy, uses an optical tracking system and a collaborative robotic arm to guide the bone saw. The hand-eye calibration procedure determines the exact pose of the robot’s end effector relative to the tracking marker. During the procedure, the surgeon tracks the surface of the bone with an optical probe. The acquired point cloud is registered to the preoperative CT model using the Super4PCS algorithm. The robotic arm then moves the slotted jig to the planned osteotomy plane. The slot physically restrains the saw blade to ensure accurate angle and depth, while the tracking marker compensates for slight patient movements.
“Our system covers the entire process from preoperative planning to osteotomy and finally segment positioning and fixation within a single integrated framework,” explains co-corresponding author Professor Jian Yang. RAMRS consists of two modules.
“This allows us to control the angle without the need for invasive custom guides,” says Professor Jingfan Fan.
The second module, Augmented Reality Guided Reconstruction (ARR), addresses the challenge of aligning the fibular segment to the defective mandible. A quick response (QR) marker is attached to the mandible. However, hand-picking the four corners of the marker with a probe may result in inaccuracies due to lever arm deformation and hand tremors. So the team developed a rotating caliper-based correction (RCC) method. The algorithm projects the selected points onto the optimal plane, finds the minimum area enclosing the rectangle, and applies the known 50 mm side length of the QR marker. “RCC reduces corner displacement by more than 25% compared to traditional least squares or constrained optimization,” said Dr. Long Shao. After this improvement, a virtual mandible reconstruction model will be overlaid on the live video feed, allowing surgeons to intuitively project color-coded locations where each fibular segment should be placed.
The team evaluated RAMRS on 25 fibulas and 25 mandibular specimens, then eight cadaver legs and eight cadaver heads to mimic clinical conditions. For fibular osteotomy, RAMRS achieved a mean angular error of 3.19° (SD 1.39°) and centroid distance error of 1.28 mm (SD 0.59 mm). Compared with previously reported methods, this angular error was 9% lower than computer modeling assistance, 42% lower than contour-aligned AR, and 22% lower than the combination of osteotomy and reconstruction preformed plates. For mandibular osteotomy, the minutiae distance error was 1.18 mm (SD 0.59 mm) and the volumetric error was only 4.34%, which was significantly better than image-guided sagittal saw (8.55%).
In the reconstruction stage, the RCC method reduced the AR fusion error to 1.26 mm (SD 0.42 mm). This was a 42-48% improvement compared to the two baseline methods. The final reconstructed mandible showed an intercondylar distance error of 1.38 mm, a mandibular angular distance error of 1.36 mm, and an angular error of 3.62°. All metrics were comparable to or better than published clinical series using pre-bent plates or CAD-CAM guidance.
Cadaver experiments confirmed that the accuracy of the system was maintained even in the presence of soft tissue. The osteotomy error remained less than 2 mm, and the reconstruction error remained small despite the reduction of the registrable bone surface.
This is a major step forward, as most previous research has stopped at plastic phantoms. We validated RAMRS under conditions that closely resemble actual surgery, including soft tissue exclusion and limited exposure. ”
Professor Tao Chan, Clinical Collaborator
The authors acknowledge that intraoperative blood contamination may still affect marker tracking and AR overlay. Future studies will integrate more robust image enhancement algorithms and add distance warning mechanisms to protect critical structures such as facial arteries. Research into regulatory preparation and standardized workflows is also planned.
“Our vision is a seamless ‘see-cut, see-place’ workflow where the robot handles geometrically demanding osteotomies and AR provides real-time visual guidance for reconstruction,” concludes Professor Yang. “RAMRS has demonstrated that combining robotics and augmented reality can improve both accuracy and ease of use in complex maxillofacial surgery.”
Authors of this paper include Sifan Cao, Jingfan Fan, Long Shao, Qing Sun, Tao Xu, Danni Ai, Tianyu Fu, Deqiang Xiao, Hong Song, Tao Zhang, and Jian Yang.
sauce:
Beijing Institute of Technology
Reference magazines:
Cao Cao, S. Others. (2026). Robot-assisted osteotomy and reconstruction with AR guidance in maxillofacial reconstructive surgery. Cyborgs and bionic systems. DOI: 10.34133/cbsystems.0590. https://spj.science.org/doi/10.34133/cbsystems.0590
