Journal of Medical Robotics Research

This study concerns the development and evaluation of a simple and ergonomic robotic system for Magnetic Resonance Imaging (MRI)-guided needle breast biopsy with lateral needle approach. The device comprises two piezoelectrically actuated linear motion stages intended to align a needle supporter with the target for manual needle insertion. The device demonstrated submillimeter accuracy and safe operation within a 3 T clinical MRI scanner. In phantom studies, tumor simulators of varying sizes were successfully targeted in both laboratory and MRI settings.

Soft actuators are a new generation of robotic actuators designed for safer and more adaptable physical human-robot interaction, that can be triggered by various stimulating mechanisms, including pneumatic, electric, electromagnetic, light, magnetic, and thermal sources. Among the different types of soft actuators, thermoresponsive ones that utilize heat as the stimulus show great potential due to their ability to deliver a relatively high force-to-size ratio without the need for external air pumps, tethers, high voltage sources, or complex designs. However, a major drawback of such actuators is their limited bandwidth. Traditional methods rely on Joule heating for actuation, with the actuator deflating when the heat source is turned off and ambient temperature takes over. Recently, the Peltier mechanism has been introduced as an alternative approach for active heating and cooling. This research paper presents a comparative analysis of the Peltier and flexible heater mechanisms in terms of the bandwidth and energy consumption of phase-change thermo-active soft actuators. The study aims to assess the potential of Peltier-based actuation in addressing the bandwidth limitations observed in traditional soft actuators. The findings reveal that Peltier-based actuation can significantly improve actuation speed in thermoresponsive soft actuators. However, it is important to note that the performance of Peltier-based actuators decreases after a few cycles unless additional measures, such as the use of an external fan, are implemented. This increase in performance comes at the cost of higher energy consumption, which should be carefully considered in practical applications.

Closed-form forward and inverse position, velocity and acceleration kinematics problems solutions are developed for a redundant medical robot family. Developed solutions allow one to choose a specific joint as a set one. Solutions work in constant time, suitable for real-time operation. Methods for joint limit and singularity avoidance are proposed. Simulation experiments are performed on simulated KUKA MED 14 r820 robot.

Robot-assisted arthroscopic surgery has been receiving growing attention in the field of orthopedic surgery. Most of the existing robot-assisted surgical systems in orthopedics place more focus on open surgery than minimally invasive surgery (MIS). In traditional arthroscopic surgery, the surgeon needs to hold an arthroscope with one hand while performing the surgical operations with the other hand, which can restrict the dexterity of the surgical performance and increase the cognitive load. On the other hand, the surgeon heavily relies on the arthroscope view when conducting the surgery, whereas the arthroscope view is a largely localized view and lacks depth information. To assist the surgeon in both scenarios, in this work, we develop a two-arm robotic system. The left-arm robot is used as a robot-assisted arthroscope holder, and it can hold the arthroscope still at a designated pose and reject all other potential disturbances, while also allowing the operator to move it via a pedal switch whenever needed. The left-arm robot is implemented with an impedance controller and a gravity iterative learning (Git) scheme, where the former can provide compliant robot behavior, thus ensuring a safe human–robot interaction, while the latter can accurately learn for gravity compensation. The right-arm robot is used as a robot-assisted surgical tool, providing virtual fixture (VF) assistance and haptic feedback during the surgery. The right-arm robot is implemented with a point-based VF algorithm, which can generate VF directly from point clouds in any shape, render force feedback, and deliver it to the operator. Furthermore, the VF, the bone, and the surgical tool position are visualized in a 3D digital environment as additional visual feedback for the operator. A series of experiments are conducted to evaluate the effectiveness of the prototype. The results demonstrate that both arms can provide satisfactory assistance as designed.

Superelastic notched tube instruments are being increasingly employed in surgical robotics. Due to their superelastic material properties, they can achieve small bending radii even for large bending angles, which makes them very interesting for minimally invasive interventions, typically dealing with confined spaces. Several works have aimed to model the behavior of such tubes in regard to the actuation force. The most extensive considers the superelastic material properties as well as the friction between the instrument and the tendon. However, no model is able to capture both the hysteresis behavior of the superelastic material as well as the hysteresis behavior induced by the friction. This text proposed a quasi-static method for modeling the behavior of nitinol notched tube instruments accounting for both material- and friction-induced hysteresis. The model is validated on four samples and compared to state of the art models. Results show an average notch position root mean square error of 1.06 mm (5.5%) and 0.46 mm (2.4%) for the method with measured parameters and optimized parameters, respectively.

Magnetic actuation systems that utilize external electromagnetic coils present a promising approach for controlling untethered small-scale robots, notably in medical applications. However, these systems frequently experience actuation singularities at which robotic control over the agent is lost or severely diminished. Such singularities may lead to excessively large current or rapid current changes that significantly affect the robots’ stability in specific areas of the workspace. In order to effectively mitigate these actuation singularities, this paper introduces a closed-loop control method using a novel nested electromagnetic coil system, leveraging finite element analysis-based magnetic field data and a novel controller combining a singularity solver and optimal current control method. The electromagnetic system has 12 coils nested in four coil units, each having three concentric coils with different outer radius, surrounding a central workspace. We call the 12-coil system Variable Outer Radius Individually Addressable Coil Stacks (VORIACS). Here, we experimentally evaluate the method for closed-loop trajectory tracking of a soft magnetic capsule endoscope (PillCam™) and a permanent magnet robot inside the VORIACS system at different magnetic field orientations. The results demonstrate that the VORIACS system with a singularity solver reduces root mean square (RMS) position tracking errors of the PillCam™by 46% to 50% and orientation tracking errors by 41% to 42% compared to a standard 4-coil system. Moreover, the system can achieve a notable 60.73% reduction in RMS position error, with sub-millimeter precision control when controlling the permanent magnet robot. We further manipulate the PillCam™for tracking a medical application-inspired complex trajectory, showcasing the system’s potential in precise control of magnetic devices for medical use.

Efforts to develop an artificial intestine for treatment of short bowel syndrome have thus far failed to achieve coordinated peristalsis. To address this, we introduce an implantable, biomimetic, magnetically-actuated peristaltic pumping apparatus. The system consists of a magnetic pump surrounding an intestinal graft and a rotating external magnetic field generator that produces peristaltic motion in the graft as the field direction alternates. We performed computational finite element modeling to simulate deformation and stress concentration on the pump. We then tested the ability of the system to produce fluid flow through porcine subintestinal submucosa at actuation frequencies from 0 Hz to 4 Hz. The system achieved a maximal flow rate of 6.03 mL/min at 3 Hz actuation frequency, which is in the physiological range of human intestinal flow rates. This device represents a novel proof-of-concept for achieving untethered peristalsis in an implantable intestinal graft that is primed for preclinical testing in animal models of short bowel syndrome.

Trocar insertion is a critical first step of all minimally invasive surgery; however, it also carries a high risk for errors. Studies suggest that entry errors are the most common complication in laparoscopic surgery with 4% of errors leading to patient fatality. Surgeon error due to excessive force is often the cause for entry errors; however, adequate training has been shown to reduce the risk of these surgical errors. In practice, institutions lack widespread and relatively inexpensive means to train surgeons for trocar entry that does not involve patient risk. In our prior work, we presented a simple Stewart platform haptic device with a numerical model to simulate key force characteristics of trocar insertion. Evaluation in our first study was limited to device characterization. In this paper, we present a more robust haptic mechanism with higher fidelity linear actuators, an increased workspace, and tissue visualization to accompany haptic cues. We also present a novel upper module that allows for a sudden drop of the trocar after the final puncture event to create a more realistic simulation. We performed a user study with eight novices to investigate how well the device and visualization train users in the trocar insertion procedure. By the end of the experiment, subjects using the device had a normalized error reduction of roughly 85% on average, relative to themselves. This device shows potential for widespread training of trocar insertion, possibly leading to fewer complications and deaths following the procedure. Finally, our upper module also represents an innovative addition for traditional admittance-type haptic device designs, not typically capable of accurately representing motion in free space.

In minimally invasive telesurgery, obtaining accurate force information is difficult due to the complexities of in-vivo end effector force sensing. This constrains development and implementation of haptic feedback and force-based automated performance metrics, respectively. Vision-based force sensing approaches using deep learning are a promising alternative to intrinsic end effector force sensing. However, they have limited ability to generalize to novel scenarios, and require learning on high-quality force sensor training data that can be difficult to obtain. To address these challenges, this paper presents a novel vision-based contact-conditional approach for force estimation in telesurgical environments. Our method leverages supervised learning with human labels and end effector position data to train deep neural networks. Predictions from these trained models are optionally combined with robot joint torque information to estimate forces indirectly from visual data. We benchmark our method against ground truth force sensor data and demonstrate generality by fine-tuning to novel surgical scenarios in a data-efficient manner. Our methods demonstrated greater than 90% accuracy on contact detection and less than 10% force prediction error. These results suggest potential usefulness of contact-conditional force estimation for sensory substitution haptic feedback and tissue handling skill evaluation in clinical settings.

Accurate endoscope navigation is vital for precise sampling and drug delivery during endoluminal procedures. Current systems rely on sensorized endoscopes for real-time bronchoscope tracking, but they are costly and require specialised settings. Camera pose estimation offers an alternative for endoscope motion tracking and navigation. In this study, we assess state-of-the-art monocular pose estimation methods for endoluminal navigation. We analyze multiple datasets and diverse feature tracking frameworks, including SIFT, ORB, BRISK, FREAK, AKAZE, SuperPoint, LoFTR, and EndoSfMLearner. Our findings reveal that, while modern deep learning-based feature detectors like SuperPoint and LoFTR yield denser features compared to traditional methods, their overall quality and reliability remain suboptimal. This insight is crucial for understanding the limitations of current monocular odometry in endoluminal environments. We also emphasize the challenges in generalizing learned frameworks developed for other odometry applications to endoscopic applications, stressing the need for more publicly available datasets. To fill this gap, we introduce a clinical bronchoscopy dataset with ground truth camera pose data. This dataset serves as a realistic benchmark for improving feature-based odometry in surgical navigation and encourages tailored framework development for learned pose estimation in endoluminal interventions.

Soft-tissue needle steering, where a deformable needle is inserted into the tissue to guide its tip to a desired position, is a common minimally invasive surgery (MIS) procedure. The diverse types of needles and complex tissue dynamics limit the use of existing approaches that utilize models of the needle and the tissue for automating the task. In this work, we employ a data-driven approach using deep reinforcement learning (DRL) to achieve autonomous needle steering by viewing it as a multi-goal reinforcement learning problem. Human interventions are incorporated during training to accelerate learning and reduce catastrophic failures. Generative adversarial imitation learning (GAIL) is combined with regular DRL by utilizing a hindsight relabeling scheme for human interventions to encourage the agent to imitate human behavior. To emulate the sim-to-real process, an agent is first trained in a simplistic simulation environment for needle steering and then transferred to a sophisticated one considered as the real world with fine-tuning (sim-to-sim). Experimental results show that with human interventions, the proposed method outperforms the other compared DRL approaches and can achieve good performance with only 2,000 training steps in the complex simulation environment, achieving an average return comparable to that of a 55,000-step agent trained from scratch.