Tissue retraction is essential for safe and efficient surgical resection; yet, it remains one of the most repetitive and attention-intensive subtasks in minimally invasive surgery. Surgeons must constantly manipulate the tissue to maintain exposure of critical anatomy while simultaneously performing precise dissection. In this work, we present PushCVAE, a Conditional Variational Autoencoder framework that learns to generate soft-tissue, nonprehensile retraction actions directly conditioned on monocular endoscopic images. Given a singleview input, PushCVAE predicts a three-dimensional contact point and pushing trajectory that safely retracts tissue without requiring explicit depth reconstruction or geometric modeling. We evaluated PushCVAE in seven surgical procedures using the Virtuoso Endoscopy System [1], a dual-arm continuum robot: Central Airway Obstruction (CAO) removal and Benign Prostatic Hyperplasia (BPH) median lobe removal. Out of 55 total retractions across both models, PushCVAE provided adequate retraction tension in 81.8% of pushes and required surgeon intervention only one time. These results represent the first demonstration of autonomous, image-conditioned softtissue retraction during surgical resections, in both airway and prostate surgery. PushCVAE establishes a generalizable framework for learning safe, autonomous surgical subtasks from small-scale demonstrations, paving the way for adaptive, image-driven surgical assistance across anatomically diverse domains.
Motion is a longstanding challenge faced in ophthalmic imaging. Patients undergoing Optical Coherence Tomography (OCT) eye exams rely on mechanical stabilization to suppress motion during high-quality imaging, which is a significant barrier for patients with impaired mobility or movement disorders. We previously developed a mobile robotic OCT system to overcome this barrier. The system still requires subjects to adopt a stable posture, however, which precludes imaging during large-scale motion, such as in Parkinson's disease or physical activity. Thus in this work, we exploit the periodic motions commonly seen during in-place activity or movement disorders and use an autoregressive filter to predict motions 100s of ms into the future based on 5 s of historical observations. This allows the robotic system to learn, anticipate, and compensate for motion to maintain high image quality. Using this approach, we conducted experiments with phantoms and with two healthy human subjects performing movement tasks. Results show that compensation reduced eye errors below the tolerance of our scanner for $\geq \mathbf{8 0}$% of subject imaging time. Our method demonstrates fast and robust motion prediction capabilities for subjects in various motion profiles, showing promising results for future research and clinical applications.
Continuum robots are a promising technology for minimally-invasive surgeries such as cardiac catheterization. However, due to anatomical differences between patients, a single robotic tool may not be appropriate for all patients. Therefore, designing custom, patient-specific robotic tools may improve the ease of certain procedures for clinicians. In this work, the design and fabrication of a tendon-driven notchedtube robotic tool for left coronary artery access based on a patient's chest computed tomography (CT) scan is demonstrated. The fabricated robot approximated the desired planar curve with an average error of 2.77 mm with an applied tendon tension within 3% of the expected tension. Additionally, the robot was successfully navigated through a two-dimensional phantom of the aortic arch from the descending aorta to the entrance of the left coronary artery.
A handheld tendon-driven endoscopic robot has been developed for keyhole skull base neurosurgery. The robot features a compact actuation unit, a joystick-controlled handheld interface, and an actively steerable distal tip capable of multi-directional bending inside narrow anatomical corridors. To address the issue of endoscopic camera contamination during surgery, a camera-cleaning mechanism was integrated. Experimental validation under three types of contamination - fog, bone dust, and blood - demonstrated that the cleaning system significantly improved imaging clarity, particularly restoring fog-obstructed images to near-baseline visibility levels. In addition, a polar-grid-based field-of-view (FOV) test showed that the steerable robot tip increased the visible area from 1192.89 mm2 in the straight configuration to 8369.10 mm2 after bending in different directions, representing a 601.5% improvement. These results confirm the effectiveness of the proposed robotic platform in enhancing intraoperative visualization and maneuverability in minimally invasive skull base procedures.
Rehabilitation after stroke, ligament injury, or neuromuscular disorders requires repetitive, precise, and safe movement training-conditions that are difficult to maintain with traditional clinic-based systems that are bulky and therapist-dependent. This study presents a Flat Inflatable Hydraulic Artificial Muscle (fiHAM) actuator integrated into a wearable lower-limb rehabilitation device designed to provide compact, high-force, and adaptive motion assistance. The fiHAM actuator is fabricated from biaxially oriented polypropylene (BOPP) film pouches that are heat-sealed using a custom G-code pattern and filled with water as an incompressible hydraulic fluid. This design enables a force density up to 2.3 × greater and a response speed 40% faster than comparable pneumatic artificial muscles. Bench experiments demonstrated blocked forces exceeding 180 N, stroke lengths up to 30 mm, and repeatability within ±3% over more than 50 actuation cycles. Integrated with a lightweight body-weight support (BWS) module, the wearable device can unload 20-30% of a user's body weight, reducing lower-limb stress during assisted gait. Intitial testing confirmed smooth torque generation, accurate angular tracking within ±0.6°, and no detectable leakage under continuous operation. Compared to pneumatic soft actuators, the fiHAM system achieved higher stiffness, faster response, and improved payload performance while remaining lightweight and portable. This work demonstrates a low-cost and scalable manufacturing process using printable polymer films and programmable heat-sealing, offering a practical path toward personalized, hydraulically actuated rehabilitation devices. The proposed system bridges the gap between laboratory prototypes and real-world wearable therapy solutions, enabling continuous, safe, and adaptive at-home rehabilitation with clinical-grade precision.
Precise identification and localization of steerable robotic needles made of nitinol is critical for MRI-guided interventions. However, nitinol needles are challenging to identify in MR images due to strong susceptibility artifacts and a lack of intrinsic signal of nitinol structures. This study aims to develop a deep learning-based method to identify nitinol needles in MR images. We developed a simulation-driven deep learning approach featuring a U-Net model trained on simulated MR images containing susceptibility artifacts induced by nitinol needle segments. These simulations incorporated variations in segment size, orientation, and location. The trained model was applied to experimentally acquired MR slices to identify and localize the needle segments in individual slides. By stacking the 2D results, the full 3D needle structures were identified. The experimental results show that the average shape estimation error is $0.57 \pm 0.1 ~\text{mm}$ along the curved section of the needle and the average tip orientation estimation error is $0.2 \pm 1.5^{\circ}$. Our approach offers a robust solution for robotic needle tracking in image-guided interventions and holds promise for clinical use in MRI-guided surgery with steerable robotic needles.
Lumbar epidural injection requires precise needle placement to ensure efficient drug delivery into the epidural space. MRI-compatible robotic systems offer unique advantages for this procedure by combining the advantages of intraoperative MRI guidance with the precision and dexterity of robotic assistance. This paper presents a 4-degree-of-freedom (DOF) MRI-compatible robotic system designed to assist surgeons in performing lumbar injections with improved accuracy and consistency. The proposed system features a modular architecture comprising an actuation unit, a two-layer linkage mechanism, and a needle placer. The kinematics of the system were derived, and a control framework incorporating backlash compensation was implemented. A workspace analysis was conducted, with the effective workspace found to be ±72.9 mm (medial-lateral) and ±32.5 mm (superior-inferior), and a tilting range exceeding 25°. As a proof of concept, the prototype was experimentally evaluated to validate its mechanical performance. The results demonstrated sub-millimeter precision with an average deviation from a mean location of 0.33 mm, confirming the feasibility of accurate and repeatable needle guidance, and marking a step toward clinical translation of robot-assisted MRI-guided lumbar injection procedures.
Continuum robotic tools present a promising alternative to rigid surgical tools for performing minimally invasive surgeries (MIS). Robotic tools developed for MIS are characterized by small outer diameters (ODs). Due to this constraint, most studies utilize metal tubes to realize the body of the robot. While these robotic tools have exhibited acceptable performance, these devices are susceptible to failure due to kinking and breakage if not designed or handled appropriately. Robotic tools fabricated using soft materials, due to their inherent compliance, can exhibit relatively higher deformation without failure. However, fabricating surgical robots with small outer diameters using soft materials can be challenging using existing techniques such as 3D-printing and molding. In this study, we explore the usage of thermoplastic polyurethane tubing with an OD of $<4 ~\text{mm}$ in combination with laser micromachining to fabricate a steerable robotic cannula. The repeatability in joint motion of a tubing sample after laser micromachining with a notch pattern is experimentally evaluated. The tubing is then machined with two bending joints capable of generating bending in opposite directions. Models to estimate the behavior of the robot are derived and experimentally validated. The models exhibited RMSE of $3.72^{\circ}, 3.16^{\circ}$, and 7.11° for the proximal joint, distal joint, and roll joint deflection, respectively. These values corresponded to $6.15 \%, 4.77 \%$, and 4.9 % of the maximum output angles exhibited by the proximal, distal, and roll joints, respectively. Finally, the proposed robotic cannula is demonstrated as a steerable neuroendoscope inside a pediatric brain phantom model.
The predominant paradigm for robotic minimallyinvasive surgery places the main surgeon at a console, teleoperating robotic instruments inside the patient, with an assistant surgeon providing support at the bedside. We present an open-source simulation platform for training this robotic surgery team that integrates the console from a da Vinci Research Kit (dVRK) for the main surgeon, a haptic device (repurposed from an existing simulator) for the assistant surgeon, a simulation environment built on the Asynchronous Multi-Body Framework (AMBF), and an Augmented Reality (AR) interface on a head-mounted display (HMD) to provide a common training environment. Our system emulates a realistic surgical scenario in which the dVRK console enables the surgeon to control the patient-side manipulators while the haptic device provides both control and tactile feedback for the first assistant (FA) operating a virtual laparoscopic grasper, with the HMD providing visualization of the extra-corporeal parts of the robot. A user study with 6 teams (12 subjects total), each performing a retraction and suturing task with both a physical system and our proposed simulator, demonstrates that our simulation platform effectively replicates key aspects of team-based surgical training.
We introduce SurgiDiff, a hierarchical recommender framework designed to assist surgeons by combining diffusion-based motion generation with large language model (LLM) reasoning. An ensemble of diffusion models generates diverse, uncertainty-aware trajectories, while the LLM evaluates their motion quality and task success to recommend the safest and most effective option for execution. Tested across tasks such as needle reach, needle pick, gauze retrieve, and peg transfer, SurgiDiff produces smooth, stable, and realistic trajectories that outperform state-of-the-art baselines, offering interpretable, safety-aligned guidance to support surgical decision-making.
Conventional endovascular catheters, built from semi-rigid polymers, offer limited tip steerability and rely on base manipulation, which hinders navigation through tortuous vasculature and can compromise patient safety. This work presents a soft robotic catheter combining a passive flexible shaft with a dexterous, mesoscale tip composed of two hydraulic actuators, one for bending and the other for torsion, fabricated from hyperelastic silicone with fiber/fabric reinforcement. Independent, decoupled control is enabled via coaxial fluid supply tubes, and module catheter design can be reconfigured to tailor workspace and task performance. Motion characterization shows up to 90° bending at 0.25 mL injected water volume and 360° axial rotation at 0.30 mL injected water volume, with block tests measuring 53.8 mN tip force (bending) and $1.3 \text{mN} \cdot \mathrm{m}$ torque (torsion). Decoupled, multi-DOF articulation was demonstrated for tool manipulation and navigation on the benchtop and in an aortic phantom, highlighting the dexterous and safe motion capability of the soft robotic catheters.
Tendon-driven robotically steerable guidewires have great potential for clinical use due to their ability to exhibit the required dexterity to navigate tortuous vessels and scope for miniaturization. Tendon-driven guidewires (TDGs) predominantly utilize nitinol tubes with machined segments at their distal section that act as bending joints. Furthermore, the tubes utilized in TDGs are hollow to accommodate the tendon for actuation. Due to the lumen of the tube, reduced amount of material at the distal segment, and the sub-mm outer diameter (OD) of the tube, the visibility of TDGs under fluoroscopy is low, which may impede the operator's ability to steer the guidewire within the vasculature. To address this drawback, in this paper, we present a method to increase the radiopacity of TDGs by using electroplating. The machined tubes used to realize TDGs are coated with gold, a material commonly used to increase the radiopacity of medical devices, using electroplating. A workflow is presented to coat the tubes with gold. Two samples of two guidewires with different ODs were utilized in this study. The effect of gold plating on the radiopacity and the mechanical behavior of utilized guidewires prior to and post-electroplating is presented. It was found that through electroplating, the radiopacity of TDGs can be increased without significantly affecting their mechanical behavior. For the sample which showed the maximum variation in the best-fit slopes for the curvature vs. force data, the slope before and after coating was $152.5 ~\mathrm{m}^{-1} / \mathrm{N}$ and $145.9 ~\mathrm{m}^{-1} / \mathrm{N}$, respectively. This result implies minimal variation in the stiffness of the samples post gold electroplating.
Accurate sensorless force estimation in teleoperated surgical robots remains challenging due to complex internal dynamics and high static friction, particularly in systems with harmonic drives such as the da Vinci Research Kit Si (dVRK-Si). This work investigates the application of dithering, a small, high-frequency oscillating signal at the robot joints, to reduce friction-related uncertainty and enhance force estimation accuracy. We first experimentally characterize the low-velocity friction behavior of the dVRK-Si Patient Side Manipulator (PSM-Si), identifying the Coulomb friction bands for its first three joints. A joint-space dithering scheme is then designed and implemented, with signal frequency and amplitude optimized through system identification and accelerometer-based feedback. Static and quasi-static experiments demonstrate that dithering improves the sensitivity and stability of Cartesian force estimation, reducing mean absolute errors by 50-80% depending on the operating condition. The results confirm that dithering effectively mitigates frictioninduced disturbances in surgical robots, enabling more precise force estimation without hardware modification and providing a pathway toward enhanced haptic feedback in teleoperated surgical systems.
Surgical robots are typically operated through large, stationary consoles that restrict surgeon mobility and hinder workflow efficiency. To address these limitations, this paper presents a novel mixed-reality (MR) teleoperation and visualization framework for the da Vinci Research Kit (dVRK). The system leverages hand and head tracking from the Microsoft HoloLens 2, combined with voice commands, to control both surgical instruments and the endoscopic camera while simultaneously displaying endoscopic video on the headset. A key innovation is the use of relative orientation control to overcome hand motion limitations, enhanced by virtual instrument shafts aligned with the physical instruments to improve intuitiveness. The framework was evaluated using the Laparoscopic Skills Training and Testing (LASTT) method with 11 participants. Experimental results show that although the proposed MR-based teleoperation system achieved lower task performance than the traditional console, participants demonstrated rapid improvement and reported comparable overall scores in the System Usability Scale (SUS) and NASA Task Load Index (NASA-TLX) questionnaires. The proposed system, in addition, exhibited lower latency than prior MRbased teleoperation approaches, demonstrating its potential for mobile and immersive robotic surgery.
Quantifying surgical expertise is essential in surgical training programs and often requires capturing both the task outcome and the underlying dynamic motion of the system. This study presents a framework for objectively evaluating laparoscopic camera navigation skills based on dynamic metrics derived from trajectory analysis. The position and orientation of a 30-degree laparoscope was recorded while 59 participants performed a standardized navigation task on a 3D-printed maze. Motion metrics for the assessment included idle time, total time, dimensionless jerk, rotation smoothness, space coverage, backtracking percentage, sample power entropy, directional changes, and speed entropy. Of these metrics, five were statistically significant: 1) Idle time (p-value = 3.7e-05), 2) total time (p-value = 3.7e-05), 3) dimensionless jerk (p-value = 0.0004), rotation smoothness (p-value = 0.0007), and space coverage (p-value = 0.0024). Results highlighted how experts demonstrate controlled variability in their movements and smoother motion patterns, while novices tend toward more unpredictable and exploratory movements. These findings help establish that analyzing dynamic metrics provides quantitative insight into camera control behavior between surgeons at different skill levels. As prior research has focused on traditional surgical skills rather than laparoscopic camera navigation skills, this work is a novel complement to more conventional time-based measures, proving it is possible to use dynamic metrics to automate feedback in surgical training programs.
Minimally invasive surgery (MIS) is preferred over open surgery due to reduced patient trauma, shorter hospital stays, and fewer procedural complications. However, current tools used in MIS, often comprising passive structures, limit the surgeon's ability to maneuver within the body. Continuum robots have been extensively studied to address these challenges due to their inherent compliance and ability to navigate complex anatomical regions. In this work, we develop a 2.14 mm outer diameter (OD) hydraulic polymerbased continuum robot. A hydraulic actuator $(\approx \mathbf{1. 0 8 ~ m m ~ O D) }$ is routed inside a notched polyimide (PI) tube using 3Dprinted blocks to ensure proper force transmission. We further compare three designs of the robot to quantify the impact of actuator routing and attachment. The motion of each design is characterized, and the proposed design exhibits the largest planar motion and reduces joint twisting up to 58.8% compared to other designs. The hysteretic behavior of the proposed robotic joint is modeled using a Preisach model to relate the actuation force to the tip deflection, resulting in a tip deflection RMSE of 2.26°. Lastly, the robot motion is demonstrated within an aortic arch phantom to show robotic navigation through confined anatomical regions. This research shows the potential of the proposed robot for MIS.
Lung cancer remains a leading cause of cancer mortality worldwide, and minimally invasive thoracic surgery has become the preferred approach for early-stage cases. Video-Assisted Thoracic Surgery (VATS) enables safe and oncologically effective resections through small incisions. To further reduce invasiveness, a magnetically anchored endoscope has been developed to share a single uniportal incision with surgical instruments. However, the proximity of ferromagnetic tools causes undesired magnetic interference, compromising anchoring stability and surgical safety. This study investigates optimized magnetic configurations to mitigate instrument-endoscope interference while maintaining sufficient anchoring and translational forces. Using finite element modeling (FEM), multiple magnet arrangements were analyzed, and a triangular Halbach pair configuration was identified to minimize stray field interactions with surgical instruments. The optimized design achieved stable anchoring with reduced lateral disturbance compared to conventional coaxial configurations. The magnetic interference was reduced by a maximum of 93% with the triangular prism array design, while maintaining the desired anchoring and translational performance. Experimental validation using fabricated magnets confirmed the simulation results and demonstrated effective interference suppression. The findings provide a practical magnetic design guideline for single-port magnetically anchored thoracoscopic systems, contributing to safer and more compact minimally invasive thoracic surgery.
Traditional surgical robotic grasping end effectors do not scale down to needlescopic (3 mm and smaller) sizes well, and performance challenges are compounded when wrist degrees of freedom are included. Current wristed needlescopic grasper designs typically produce low grasp forces, making it challenging to retract tissue or firmly hold suture needles. In this paper, we propose a new design that achieves grasp forces at least six times higher than the repurposed flexible graspers used in prior concentric tube robotics research. Our new design exceeds even the grasp force of the much larger da Vinci Surgical System. We achieve this by combining a nitinol push-pull actuation tube with distal metal cam joints that amplify grasp force. The result is a tool with concentric-tube-type wrist joints capable of not only strong grasps, but also active jaw opening (e.g., for blunt dissection), that exceed the requirements of even larger-scale laparoscopic surgery. We validate our new grasper experimentally, comparing it against three commercially available graspers of the types previously attached to concentric tube robots, as well as to a da Vinci Surgical tool. We find that even at 3 mm diameter, our prototype surpasses the maximum grasp forces of an $\mathbf{8. 5 ~ m m}$ da Vinci tool, while providing more than 4 N of active opening force, and affording torsional stiffness many multiples higher than prior concentric tube end effectors. An additional useful feature of our design that is not present in prior grasper designs is the provision of an open central lumen useful for delivering tools such as laser fibers into the surgical field.
Suturing is valuable in many surgical procedures, yet is challenging to implement in mm-scale instruments due to their small end effectors with limited grasp force capabilities. To address this, in this paper, we present a new miniature needle grasper design and use it to demonstrate needle-andthread suturing using concentric tube robots (CTR) for the first time. We compare the new grasper's performance with two conventional endoscopic forceps graspers experimentally. Our design achieved a maximum needle-throwing force of 7.05 N, more than twice that of existing graspers, while maintaining a stable grasp throughout piercing motions. We then apply this technique to the practical surgical application of internal tracheal stent fixation. Our experiments demonstrate that a single suture can increase fixation strength by over an order of magnitude. These results demonstrate the feasibility of enhancing current minimally invasive procedures with thrown sutures within confined anatomical sites.
Human-robot interaction (HRI) in healthcare has expanded rapidly, but it still lacks a systematic taxonomic framework. This study developed a four-dimensional taxonomy for HRI in healthcare by analyzing existing literature, including 49 articles from 2020 to 2025, using QDA Miner and WordStat. The taxonomy encompasses user types including patients, healthcare staff, caregivers, children, and older adults; robot types including surgical, socially assistive, physically assistive, and service robots with functional and appearance subcategories; interaction elements including task based, social, and therapeutic interactions; and context including acute care centers, community based, specialized units, research centers, institutional care units, and telehealth. In this analysis, child subcategories (child codes) along with parent categories revealed prominent and least studied research areas. Furthermore, multidimensional co-occurrence analysis showed clustering of various dimensions and both connected and isolated components of the taxonomy. Overall, this proposed taxonomic framework helps to identify the research gaps, potential research opportunities, and a roadmap for advancing HRI in healthcare.
