Surgical robots edge toward autonomy in orthopedics and neurosurgery

Surgical robotics is approaching a turning point. According to an editorial published in Frontiers in Robotics and AI on 20 October 2026, the field’s central challenge is no longer simply building machines that can hold a drill steady, but creating systems whose capabilities, clinical efficacy and efficiency are good enough to deploy them in increasingly complex and critical surgical tasks, close to fragile structures, in tissue that is deformable and dynamic. The editorial, written by Philipp Fürnstahl of Balgrist University Hospital in Zürich, Tom Vercauteren of King’s College London, Emmanuel Benjamin Vander Poorten and Ruixuan Li of KU Leuven, Matthias Harders of the University of Innsbruck, Fabio Carrillo of the OR-X Translational Center for Surgery in Zürich and Reza Monfaredi of Children’s National Hospital in Washington, DC, introduces a research collection on robotics in orthopedics and neurosurgery that illustrates both how far the technology has come and how much stands between today’s prototypes and routine clinical autonomy.

The editors frame the problem in starkly technical terms. Better designs, better sensing approaches, more sophisticated algorithms and advances in system intelligence and autonomy are all needed, they argue, because current robotic technologies, while augmenting human capabilities, still depend heavily on direct human control. Achieving greater autonomy requires extensive research combining practical approaches with advanced robotic artificial intelligence techniques, and every autonomous decision based on real-time surgical data must uphold the highest standards in safety. The stakes are unusually high in the two specialties the collection covers. In orthopedics and neurosurgery, robots must deal both with soft, deformable tissue and with hard, bone tissue. While operating in the vicinity of fragile and critical structures, they may need to exert large forces, which further complicates safety considerations. Maintaining surgical and functional accuracy beyond purely geometric accuracy in such a dynamic environment, the editors note, remains an open challenge.

What unites the works gathered in the collection, according to the editorial, is the close integration of technical innovation with clearly identified unmet clinical needs, supported by rigorous validation in synthetic or ex-vivo settings. That last qualifier matters. The studies described were tested on synthetic models, animal tissue or cadaveric human specimens rather than in live patients, so the editors describe them as suggesting strong potential for establishing tangible real-world advancements in orthopedic and neurosurgical robot-assisted interventions, not as demonstrating clinical benefit in the operating room. Within that scope, the collection spans the spine from the lumbar region to the cervical vertebrae, minimally invasive cement injection, and high-force fracture realignment in the femur.

The collection opens with a mini-review by Cho and colleagues on human-centered assurance in robot-assisted orthopedics and neurosurgery. After summarizing the different ways artificial intelligence is being incorporated into next-generation robotic platforms, the work introduces assurance frameworks that can help surgeons effectively maintain oversight of complex AI-supported platforms. Human-centered assurance, the editorial reports, is argued to be a key enabler for the safe adoption of increasingly autonomous surgical systems. Cho and colleagues go on to sketch several essential research directions aimed at understanding how human responsibility and control can be maintained in the upcoming and increasingly autonomous orthopedic and neurosurgical interventions. In other words, before any robot takes a larger share of the decision-making, the field is being asked to build the machinery of accountability first.

The first experimental study highlighted, by Davoodi and colleagues, targets pedicle screw placement in the lumbar spine, a common fixation procedure in which screws must be drilled along precise trajectories into the vertebral bodies. The team focused on an optical navigation and registration approach as an alternative to methods that rely on intra-operative fluoroscopy and CT registration. Validated on both synthetic models and ex-vivo bovine spines, the approach produced average position errors of 2.61 millimeters and 1.51 millimeters respectively, and orientation errors of 2.65 degrees and 2.30 degrees. Those numbers, the editorial notes, demonstrate the potential of the approach to reduce radiation exposure for both patients and clinicians, since optical tracking could substitute for repeated X-ray imaging. The authors conclude that although challenges remain, robots equipped with optical tracking show considerable potential for clinical use.

While Davoodi and colleagues concentrated on the lumbar spine, a study by Cong and colleagues extends robotic screw placement to cervical spine procedures, which are considerably more delicate. The team reported achieving 97.2 percent clinically acceptable robot-assisted placements of anterior transpedicular screws in the subaxial cervical spine using five human specimens. The study used the uOssa-Nav-s750 system, a robotic framework developed by Wuhan United Imaging Surgical that employs 3D C-arm imaging, point-based rigid registration and semi-automatic trajectory alignment. Once the robot is automatically aligned to the preplanned screw trajectory, the surgeon performs the final fine adjustment and locking before drilling. The editorial is careful about what this shows: the ex-vivo study on human specimens suggests that a robotic approach is feasible even for challenging interventions such as cervical anterior transpedicular screw placement, where current procedures are conducted manually. The technique offers demonstrated biomechanical advantages but is considered technically highly demanding today, and the editors note that, provided further validation is conducted, robotic approaches may help broaden its adoption in clinical settings in the future.

A third study, by Inigo and colleagues, addresses a different spinal procedure with a different imaging constraint: vertebroplasty, a minimally invasive intervention used to stabilize weakened or fractured vertebral bodies. Unlike pedicle screw placement, where the goal is to accurately drill trajectories into which screws are subsequently placed, vertebroplasty requires direct needle insertion through the pedicle into the vertebral body in one go; bone cement is then injected through the needle’s hollow channel to strengthen the fractured vertebra. Crucially, vertebroplasty typically cannot rely on pre-operative CT. To counter this lack of information, the team’s Spine-DART system combines atlas priors with deep learning, differential rendering and a multi-view learned similarity loss, ultimately requiring only two arbitrary X-ray images for three-dimensional vertebral reconstruction and for planning three-dimensional transpedicular paths. That makes the system highly practical in real intraoperative settings, the editorial argues. Trained on the colon and NMDID data sets, the method was tested on the VerSe-small data set and in a single cadaver experiment, achieving competitive reconstruction results and plans considered clinically viable in the majority of cases.

In a study by Gaytan and colleagues, the authors describe the ability of the Robossis V2 system to support femur fracture reduction, assisting in the re-alignment of the different femur bone fragments after fracture to restore anatomical position and promote healing. The paper describes the system’s mechanical design and focuses on the improved workspace of Robossis V2 compared with its V1 predecessor, showing workspace coverage well beyond the clinically required alignment ranges. Further work on intuitive control and workflow integration is planned to optimize surgeon-robot collaboration in what the editorial calls a high-force, safety-critical application, a reminder that in orthopedics the robot’s environment includes not only delicate anatomy but enormous mechanical loads.

Stepping back from the individual studies, the editors draw a broader conclusion about where the field is heading. Advancing the capabilities of surgical robotics, they write, demands interdisciplinary research across multiple fronts. Aside from progress in actuation, sensing and control, a key challenge is improving the real-time understanding of the surgical scene to deal with deformation, physiological phenomena and uncertainty. The collection showed a major recent focus on developing new clinical workflows that exploit diverse real-time sensors and real-time data processing to enable real-time, AI-empowered decision-making in complex surgical environments. As surgical robot capabilities improve and systems gain autonomy, ensuring their reliability through novel control and planning strategies is described as crucial for handling complex situations without compromising patient safety. Seamless data integration from visual and non-visual sources is deemed essential for accurate decision-making.

The editorial closes on the human side of the equation. In parallel with increased autonomy, improvements in human-machine collaboration are needed to ensure robustness, safety and efficiency of operation. Beyond more intuitive interfaces that enhance human oversight, improved communication of intention among autonomous agents could strengthen human assurance. That, in turn, may facilitate smoother downstream transitions and support subsequent regulatory processes, which the editors describe as critical to bringing orthopedic and neurosurgical interventions to a new level of sophistication. For now, the evidence base remains preclinical, resting on synthetic models, bovine spines, cadaveric specimens and a single cadaver experiment, but the collection maps a coherent route from laboratory validation toward the operating theater, with safety, oversight and radiation-sparing navigation as its recurring themes.

Subject of Research: An editorial introducing a research collection on robotics in orthopedic and neurosurgical surgery, spanning human-centered assurance, optical navigation for pedicle screws, cervical screw placement, image-sparse vertebroplasty planning and femur fracture reduction.

Article Title: Editorial: Robotics in orthopedics and neurosurgery

Article References: Fürnstahl, P., Vercauteren, T., Vander Poorten, E. B., Li, R., Harders, M., Carrillo, F., & Monfaredi, R. (2026). Editorial: Robotics in orthopedics and neurosurgery. Frontiers in Robotics and AI, 13, Article 1982252. https://doi.org/10.3389/frobt.2026.1982252

Image Credits: AI Generated

DOI: 10.3389/frobt.2026.1982252

Keywords: surgical robotics, orthopedics, neurosurgery, pedicle screw placement, vertebroplasty, optical navigation, autonomy, human-robot collaboration, intraoperative imaging, real-time sensing, robotic perception, minimal invasive surgery