The results of our research and development work are regularly published in renowned journals and presented at scientific conferences.
This study evaluated the feasibility and potential benefit of Przypominajka—a sensor glove paired with a tablet application—in 16 post-stroke patients. After more than two weeks of exercise, the device group improved more than the paper-instruction control group, including a clinically meaningful change in FMA-UE. The findings support sensor-assisted exercise as a useful supplement to upper-limb rehabilitation.
The paper presents Balonikotron, a soft balloon-actuated device for assisted wrist extension. A model linking force, angle, pressure, and volume agreed closely with measured forces (r = 0.936), supporting safety monitoring. In a four-week pilot study of 12 patients, the device group showed significant improvements in functional independence and upper-limb function.
This review examines scales, tests, and instruments for objectively evaluating post-stroke hand rehabilitation. It organizes key parameters and discusses combining established tools such as EMG, goniometers, and dynamometers with vision systems, pose estimation, and artificial intelligence. The goal is to help teams select outcome measures and improve the comparability of therapy results.
This review summarizes research from 2019–2022 on immersive virtual reality delivered through head-mounted displays for post-stroke upper-limb rehabilitation. The technology enables individualized, task-oriented motor training in a virtual environment and may supplement conventional rehabilitation. The article discusses the evidence base and suitability of this approach for stroke patients.
The paper describes Przypominajka v2, a self-training support system combining a sensor glove, on-device exercise classification, a tablet interface, and a therapist web application. Classification accuracy was high for the healthy participants studied but lower for new users. A patient case study showed a positive response, alongside usability issues involving glove fitting and instruction clarity.
The authors present an automated hand-measurement system combining computer vision, a haptic device, and a web interface. It measures hand size, finger and wrist range of motion, and force. Testing with five healthy participants showed agreement with selected manual measurements and potential for precise post-stroke rehabilitation support.
This study evaluated the effectiveness and safety of four weeks of Balonikotron-assisted therapy in 12 patients after ischemic stroke, randomly assigned to experimental and control groups. After 20 sessions, upper-limb function, spasticity, and independence were assessed with FMA-UE, MAS, and the Barthel Index. The experimental group showed significant improvements with large effect sizes, while the control group showed no significant changes.
This case study examined Balonikotron use by a 53-year-old patient after ischemic stroke. The patient completed three 45-minute wrist-extension sessions per week, with assessment using the Ashworth Scale, VAS, range-of-motion measurements, joint circumferences, and sensory testing. The results indicated relationships among muscle tone, pain, and range of motion, and supported robotics as a possible complement to conventional rehabilitation.
Robotic rehabilitation devices that utilize embedded systems, such as the Stretchbox—a device designed for hand rehabilitation through the actuation of rollers stretching a rubber band by BLDC motors—demand controllers that are safe, scalable, and understandable at a high level. Behavior trees satisfy these requirements and have found broad applications in robotics and some pioneering applications in the medical sector. Despite their widespread use, there has been an absence of behavior tree frameworks for microcontrollers programmed with MicroPython, which is an attractive language for embedded systems and the chosen platform for the Stretchbox’s high-level controller. This paper introduces a behavior tree implementation for MicroPython tailored as a high-level controller for the Stretchbox device. The implementation includes standard behavior tree nodes, specialized nodes, and classes designed for interfacing with microcontroller peripherals and for Bluetooth communication with the tablet-based user interface. This system provides a comprehensible and straightforward method for controlling the Stretchbox device for rehabilitation, establishing a well-defined hierarchy for event management. Furthermore, the solution is designed to be extensible, allowing for adding new therapeutic behaviors and integrating extra devices. As an open-source library, the implementation facilitates the adoption of behavior trees in other MicroPython-based projects. The detailed explanation of the approach in applying behavior trees to the Stretchbox serves as a practical demonstration of how behavior trees can be effectively used in controlling rehabilitation devices.
Soft robotics experiments necessitate advanced measurement techniques to track robot pose and deformations accurately. Electromagnetic tracking systems provide substantial benefits over vision-based systems, especially in capturing both the position and orientation of a robot’s surface without occlusion issues. However, these systems are typically designed for applications in controlled environments, with multiple guidelines stated by the producers, necessitating an evaluation of their effectiveness in different contexts. This paper presents a series of foundational experiments to assess the capabilities and limitations of such a system in soft robotics. We investigate the impact of distance from the base station, resistance to environmental interference, effects of metallic elements, and the accuracy of rotational measurements. The results demonstrate a significant influence of positional accuracy relative to the distance from the transmitter, with a tenfold decrease in accuracy observed. Additionally, the presence of metallic objects in close proximity (less than 20 mm) to the probes adversely affects accuracy, as do large metallic elements such as construction materials. This effect is particularly pronounced when the sampling frequency of the system aligns with multiples of the electrical network frequency, resulting in more than a tenfold reduction in accuracy.