Safe capsule-colon interaction for magnetically driven robotic capsules is important in clinical applications. This work presents a solution based on the amplitude information of the magnetic field to adjust the distance between the interacting magnets, in order to prevent the magnetic forces exerted on the capsule robot and the pressure on the intestine walls from being overlarge, which may cause large deformation of the colon. As the first step, the geometry of the internal magnet embedded in the capsule is optimized to approach a near-spherical amplitude of the magnetic field based on the dipole model. Next, mathematical mapping from magnetic field amplitude to the interaction distance between the magnets is presented with constraint derivation and implementation. Then, a strategy to adjust the distance between the interacting magnets is provided based on the mapping using the magnetic field information. Finally, experiments are designed to validate the pose-independent interaction distance adjustment. Compared with the previous work, the proposed solution enables the quick interaction distance adjustment between the magnets to enhance the safety of capsule-colon interaction in the magnetically driven capsule endoscopies, since the interaction distance is derived straightforwardly from the magnetic field signals, without requiring the prerequisite implementation of capsule localization.

Catheter-based cardiac ablation is a minimally invasive procedure for treating atrial fibrillation (AF). Electrophysiologists perform the procedure under image guidance during which the contact force between the heart tissue and the catheter tip determines the quality of lesions created. This paper describes a novel multi-modal contact force estimator based on Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs). The estimator takes the shape and optical flow of the deflectable distal section as two modalities since frames and motion between frames complement each other to capture the long context in the video frames of the catheter. The angle between the tissue and the catheter tip is considered a complement of the extracted shape. The data acquisition platform measures the two-degrees-of-freedom contact force and video data as the catheter motion is constrained in the imaging plane. The images are captured via a camera that simulates single-view fluoroscopy for experimental purposes. In this sensor-free procedure, the features of the images and optical flow modalities are extracted through transfer learning. Long Short-Term Memory Networks (LSTMs) with a memory fusion network (MFN) are implemented to consider time dependency and hysteresis due to friction. The architecture integrates spatial and temporal networks. Late fusion with the concatenation of LSTMs, transformer decoders, and Gated Recurrent Units (GRUs) are implemented to verify the feasibility of the proposed network-based approach and its superiority over single-modality networks. The resulting mean absolute error, which accounted for only 2.84% of the total magnitude, was obtained by collecting data under more realistic circumstances in contrast to previous research studies. The decrease in error is considerably better than that achieved by individual modalities and late fusion with concatenation. These results emphasize the practicality and relevance of utilizing a multimodal network in real-world scenarios.

Soft actuators and robotic devices designed for rehabilitation and assistance are a rapidly growing field of research. Their inherent flexibility enhances comfort and usability without restricting the user's natural range of motion. However, despite these advantages, there are still several challenges that need to be addressed before these systems can be commercialized. This paper presents a comprehensive review of the latest developments in soft wearable robots, also known as exosuits. Soft exosuits are composed of two main components: actuation mechanisms (how forces/torques are generated) and physical interfaces (how and where the robot is anchored to the body). This paper reviews the advances in these two areas, while categorizing exosuits based on the intended assisted joint, assisted degrees of freedom (DOF), and device type. The systematic literature review follows the PRISMA guidelines to summarize the relevant studies and investigate their related physical interface, actuation mechanism and its design. Several limitations were identified in these areas, and insights into potential future research directions are presented. In the future, the goal should be to develop an untethered assistive device that can provide assistance to multiple joints while having a low form factor, an intuitive and natural interface, and being comfortable for the user.

Deep brain stimulation (DBS) has demonstrated therapeutic efficacy in the treatment of neurological and psychiatric disorders. Currently, DBS devices employ an openloop configuration, requiring manual adjustment of electrical stimulation to address patient needs. For this reason, closed-loop DBS is being developed, delivering appropriate treatment ondemand based on internal signal monitoring. A key challenge in current research is the complexity of interpreting the measured signals and delivering appropriate interventions, currently no miniaturised closed-loop DBS device has on-board artificial intelligence (AI) to meet this need. This paper presents a new miniaturised device, named SmartStim, that uses AI to monitor dynamically changing brain biomarkers. In addition, the AI decides if the output stimulator is required for treatment. This device has two key components: the hardware module (neural sensor unit, processor, and neurostimulator) and a software module (data processing, AI, and firmware). The neural sensor unit is comprised of two subcomponents. The first is a potentiostat that can perform impedance analysis, and the second is a dedicated fast scan cyclic voltammetry (FSCV) front-end that can perform scan rates up to 1000 V/s. This device can output currentcontrolled stimulation waveforms in a frequency range of 5 Hz–200 Hz, a current range of 1 μA to 10 mA, with active charge balancing. Five experiments were conducted to validate SmartStim: static resistive load test, emulated brain resistance test, static electrochemical cell test, impedance test, and dynamic serotonin test. These experiments confirm the potential for SmartStim to identify neurochemical patterns in a mouse brain using AI.

Robot-assisted endovascular intervention has emerged for improving the outcomes of cardiovascular interventions. However, the current segmentation methods are affected with low and varied contrast values of endovascular tools in the angiogram, and background noise, both of which affect segmentation performance. Thus, surgical scene analytics are characterized with slow tool visualization and response during endovascular navigation. In this study, a multi-lateral branched network (MLB-Net) is proposed for pixel-level segmentation of guidewire in angiograms recorded during robot-assisted cardiovascular catheterization. The network has an encoder with lateral separable convolutions and depth-wise attention, and decoder with improved loss function. Feature maps extracted during end-to-end fully supervised training were optimized for guidewire segmentation. The MLB-Net was trained and validated with multiple angiogram sequences obtained during series of robot-assisted catheterization in rabbit model. Validation studies show a robust performance, characterized with mean IoU of 84.89% and area under curve of 90.64%. In addition, the model offered fast (15.28 frame/second) and reliable segmentation performance in new angiograms obtained during additional trials carried out in pig and human phantom models. Furthermore, we evaluated the MLB-Net by comparing it with existing state-of-the-art networks. Based on our rabbit dataset, the MLB-Net offers better segmentation experience over DeepLabV3+, SegNet, and U-Net which are commonly used for medical image segmentation. Also, MLB-Net generalized well under incremental training. This study contributes a new model for fast tool segmentation, tracking and visualization and during endovascular catheterization.

Deterioration of the lower limb venous circulation can cause several health problems (e.g., varicose veins, deep vein thrombosis, and cardiac arrhythmia). However, it is not easy to avoid owing to a sedentary life, which is one of the characteristics of modern people. Several studies have focused on improving lower limb blood circulation, and some have used robotic devices. These robots, however, not only the lead to poor blood circulation, but also are too bulky and heavy to use in daily life. In this study, a new type of lower limb venous circulation robot called Healthy-CAL is proposed. By applying the semi auto locking linkage mechanism, the size of Healthy-CAL was reduced while the operating range of the robot was maximized. In addition, a foot mapping algorithm was applied to Healthy-CAL, which autonomously recognizes the user’s ankle joint range of motion and allows users to perform maximized ankle exercises regardless of the sitting position. An experiment was conducted with eight participants to demonstrate the effectiveness of the proposed foot-mapping algorithm. We also used Doppler ultrasound in five participants to confirm the improvement in venous circulation with Healthy-CAL.

As surgical interventions trend towards minimally invasive approaches, Concentric Tube Robots (CTRs) have been explored for various interventions such as brain, eye, fetoscopic, lung, cardiac, and prostate surgeries. Arranged concentrically, each tube is rotated and translated independently to move the robot end-effector position, making kinematics and control challenging. Classical model-based approaches have been previously investigated with developments in deep learning-based approaches outperforming more classical approaches in both forward kinematics and shape estimation. We propose a deep reinforcement learning approach to control where we generalize across two to four systems, an element not yet achieved in any other deep learning approach for CTRs. In this way, we explore the likely robustness of the control approach. Also investigated is the impact of rotational constraints applied on tube actuation and the effects on error metrics. We evaluate inverse kinematics errors and tracking errors for path-following tasks and compare the results to those achieved using state-of-the-art methods. Additionally, as current results are performed in simulation, we also investigate a domain transfer approach known as domain randomization and evaluate error metrics as an initial step toward hardware implementation. Finally, we compare our method to a Jacobian approach found in the literature.

Cyclograms are useful for describing interjoint coordination in clinical gait analyses. In this paper, we establish a model for generating hip-knee cyclograms at different walking speeds by using B-Spline curves, which could provide reference standards for evaluating hip-knee joint coordination. We use the shape similarity between individual and reference cyclograms to develop a hip-knee joint coordination assessment metric: the dynamic time warping-shape context (DTW-SC). The DTW algorithm is used to match corresponding points between the hip-knee cyclograms, and the matched results are visualized to show abnormal changes in gait patterns. The distance between corresponding points is determined through shape context descriptors and applied to quantify the hip-knee joint coordination of the subjects. To verify the effectiveness of the proposed approach, a stereo vision-based portable lower limb motion capture system is developed to collect hip and knee joint angle data from five healthy and five hemiplegic subjects walking at comfortable speeds. Experimental results show that the range of the modeling error is [0.84°, 2.76°]. DTW-SC metric can visualize and quantify the abnormal patterns of the subject’s hip-knee joint coordination. The work is meaningful in evaluating the hip-knee joint coordination of hemiplegic patients.

It was observed that repetitive and task-oriented movements can strengthen muscles and improve the walking capabilities among patients experiencing gait impairments due to neurological disorders. However, the traditional physiotherapy is laborious, may not provide desired cadence and gait patterns, and due to the lack of therapists, it remains inaccessible to many patients.In the present work, we propose a novel low-cost treadmill-based underactuated gait rehabilitation exoskeleton that can provide naturalistic lower limb movements during walking. The exoskeleton mechanism has been devised by constraining a planar serial chain formed by revolute joints to resemble the Stephenson III six-bar linkage satisfying the adaptability and the end-effector trajectory requirements. The synthesized mechanism matches the concurrent knee and ankle joint movements during human walking relative to the hip, in terms of position and time. The kinematics and dynamic force analysis of the mechanism formulating the position, velocity, acceleration, and static torque have been presented. A full working lightweight exoskeleton prototype with a single actuated joint has been constructed. A pilot study with a healthy human subject has been performed. As a result, it is found that the proposed exoskeleton can provide naturalistic human motion trajectories suitable for rehabilitation purposes.

Wearable robots hold great promise in supporting people during activities of daily living. Personalizing the assistance provided by this technology for not just level ground walking, but also other real-life activities like uphill and downhill walking, will foster its wider adoption. In this paper, we propose an approach to personalize the assistance delivered by a wearable robot – the Myosuit, a tendon-driven soft robot for mobility assistance – for downhill walking. We use a human-in-the-loop method with candidate profiles of 15-steps which aims to minimize knee extensor muscle activity, utilizing the covariance matrix adaptation evolution strategy algorithm (CMA-ES) and a cost function based on normalized RMS of surface EMG signals. By utilizing CMA-ES, we vary the magnitude and timing of Myosuit’s support to find the most appropriate assistance profiles. We compared the optimal profiles to assistance off (transparency) condition and a general trajectory derived using a model-based approach. Relative to the transparency condition, the average muscle activity was reduced by 12.67% (SD: ±8.73; t(20) = 5.25, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$p\leq.01$ </tex-math></inline-formula> ) across all participants and muscle groups. The results of our study may support future development of personalized assistance algorithms for wearable robots and lead to better adoption of this technology.