This paper introduces a robotic biopsy device for capsule endoscopes. The proposed device consists of three modules for the complete process of biopsy, which includes monitoring the intestinal wall by a tissue monitoring module (TMM), aligning onto a polyp by an anchor module (AM), and sampling of the polyp tissue by a biopsy module (BM). The TMM utilizes a trigonal mirror as well as an on-board camera; since the TMM continuously takes images through lateral apertures, an operator such as a medical doctor is able to anchor the capsule endoscope onto the polyp and biopsy it with the visual feedback in real-time. When the operator finds a polyp using the TMM and the frontal camera of a capsule endoscope, then the AM is used to approach the polyp for biopsy. When the AM is in use, outriggers are extruded by shape-memory-alloy (SMA) springs, which results in the swelling of capsule endoscope body. In addition, an alignment module, which is a part of the AM, rotates the body of the capsule endoscope such that the biopsy razor can be aligned onto the polyp. Then, the BM excises a part of the polyp and seals the aperture, and the capsule endoscope continues exploring the intestine. The concept and working principles of the proposed device are introduced in this paper and are verified by a prototype that successfully integrates the three modules.

In exoskeletal robots, the quantification of the user’s muscular effort is important to recognize the user’s motion intentions and evaluate motor abilities. In this paper, we attempt to estimate users’ muscular efforts accurately using joint torque sensor which contains the measurements of dynamic effect of human body such as the inertial, Coriolis, and gravitational torques as well as torque by active muscular effort. It is important to extract the dynamic effects of the user’s limb accurately from the measured torque. The user’s limb dynamics are formulated and a convenient method of identifying user-specific parameters is suggested for estimating the user’s muscular torque in robotic exoskeletons. Experiments were carried out on a wheelchair-integrated lower limb exoskeleton, EXOwheel, which was equipped with torque sensors in the hip and knee joints. The proposed methods were evaluated by 10 healthy participants during body weight-supported gait training. The experimental results show that the torque sensors are to estimate the muscular torque accurately in cases of relaxed and activated muscle conditions.

Cerebrovascular accidents (CVA) cause a range of impairments in coordination, such as a spectrum of walking impairments ranging from mild gait imbalance to complete loss of mobility. Patients with CVA need personalized approaches tailored to their degree of walking impairment for effective rehabilitation. This paper aims to evaluate the validity of using various machine learning (ML) and deep learning (DL) classification models (support vector machine, Decision Tree, Perceptron, Light Gradient Boosting Machine, AutoGluon, SuperTML, and TabNet) for automated classification of walking assistant devices for CVA patients. We reviewed a total of 383 CVA patients’ (1623 observations) prescription data for eight different walking assistant devices from five hospitals. Among the classification models, the advanced tree-based classification models (LightGBM and tree models in AutoGluon) achieved classification results of over $90\%$ accuracy, recall, precision, and F1-score. In particular, AutoGluon not only presented the highest predictive performance (almost $92\%$ in accuracy, recall, precision, and F1-score, and $86.8\%$ in balanced accuracy) but also demonstrated that the classification performances of the tree-based models were higher than that of the other models on its leaderboard. Therefore, we believe that tree-based classification models have potential as practical diagnosis tools for medical rehabilitation.

The present study suggests a novel method of strengthening specific muscle groups in the lower limb during a functional movement. When the foot of a user wearing an exoskeletal robot follows a given path, the contribution of each muscle group to generate the motion changes along the path of the trajectory. The efficiency of muscle training, which is defined as the ratio of the work of a specific muscle group to that of all groups, can be maximized by changing the training load along the path. Based on a musculoskeletal model, the contribution of each muscle group along the path can be calculated as a function of its position. When a specific muscle group is chosen for exercise or rehabilitation, the efficiency of training can be maximized by setting high load where its contribution is high, and low load where its contribution is low. By doing so, the user can exercise longer with the same amount of energy consumption. The EXOWheel which features a lower limb exoskeleton is employed to verify the method and the hamstring muscle group is selected as the specific muscle group. Three healthy subjects participate in the experiment, and electromyogram sensors are employed to monitor the muscle power. The results indicate that the efficiency of the hamstring muscle group for a given circular foot path under the optimal training load is 46.2% compared with 32.5% with a constant load, and it means the 42% higher efficiency of the specific muscle group training.