The NuroSleeve comprises five main components: a custom 3D-printed splint, an external FES unit, a main control unit (MCU), a clinical software suite for configuration, and a rechargeable battery, as shown in Fig. 1. The custom firmware running on the MCU receives signals from one or more input sensors, Fig. 1: The NuroSleeve consists of the main control unit (center) which accepts different input control signals (left) to control one or more end effectors (right). Implementations are customized for each patient by the occupational therapist using the clinical configuration software.
processes the data in real time, and derives control signals for the effectors. The NuroSleeve firmware and hardware allow for user-specific sensor setup combined with personalized input/output mapping. In other words, sensor placement and user commands are customized for the user, as are the effector control strategies. The NuroSleeve can be controlled by one or more of the following control inputs: joystick input, electromyography (EMG) signals, inertial measurement unit (IMU) signals, and voice control. Current effector options include a mechanically operated forearm splint and an external two-channel clinical-grade FES device (see Fig. 1).
It is important to emphasize that the NuroSleeve device is user-specific and is personalized with the help of an occupational therapist (OT) using the clinical configuration software. The OT can exploit rehabilitation therapy principles to help identify the optimal combination of sensor inputs and effectors, and the optimal placement of each, based on the individual’s unique abilities, needs, and functional goals. The clinical software suite permits these configurations via Bluetooth and allows for device usage monitoring, real-time data collection and visualization, and command exchange between NuroSleeve and external devices, such as brain-computer interface (BCI) systems.
The NuroSleeve splint is also customized for each individual using 3D scanning and printing technologies. Each splint is built to perfectly accommodate the unique anatomy and impairment of the user to maximize comfort, efficacy, and fit. We refer to this customization process as the “Digital Orthotist” process, as it combines modern industrial design techniques with occupational therapy and orthotics know-how. The process starts with a 3D scan of the user’s impaired hand and forearm, which is then used to build a custom computer-aided designed (CAD) model of the splint. The model may be fine-tuned to suit the user as necessary and then it is 3D printed. The combination of 3D scanning and 3D printing technologies has facilitated our development of an orthosis that is lightweight, aesthetically pleasing, and form-fittingin key locations while form-adjusted in others to avoid pressure points and bony prominences [38]. The splint has an innovative clamshell design, incorporating a hinge on one side that allows the thumb and cuff sections to be opened and closed, maximizing the individual’s ability to don and doff without assistance. The use of 3D-printed rigid plastic components, rather than fabric or other soft materials, provides mechanical support and enables dynamic grasp properties, which are particularly relevant for individuals with spasticity (e.g., excessive tone) [39].
The NuroSleeve orthosis also integrates an external FES unit that can generate functional movement by stimulating paretic muscles. The application of electrical current to a person's muscles depolarizes peripheral neurons and elicits muscle contractions, allowing the person to perform a volitional movement. FES has evolved into a crucial treatment approach that clinicians may use to help individuals with stroke and SCI regains the capacity to stand, walk, reach, and grasp [40]. FES can benefit individuals by substituting or enhancing movement. Repeated muscle activation using FES may also increase voluntary motor control. This suggests that the use of FES devices improves motor recovery and can serve as a rehabilitation technique as well as assisting with ADLs [41]. The NuroSleeve is intended to be used as an assistive device, but it can be also used as a self-modulated rehabilitation device. Supplementary Table 1 lists current commercially available neuromuscular electrical stimulation devices designed for rehabilitation purposes.
Device Control Method
The NuroSleeve can control the linear actuator of the splint and/or the FES unit using any of the following control modes: (1) Manual control. A small joystick is fitted to the device at an accessible location for the user; (2) Voice Activation. A voice recognition module that does not require connection to the internet (this was added to the system following user feedback). This voice recognition module can extract and analyze the voice features of a speaker after a single calibration session with the individual. Following the calibration and setup session, the individual can use the voice control option to control the NuroSleeve independently and without the use of external resources; (3) EMG control. The device’s two EMG channels can be set up to control the linear actuator and/or FES with a multi-threshold approach, in which one or two signal thresholds are set up to trigger the effectors. The threshold values and their use in controlling the effectors can be customized via the clinical software suite; and/or (4) IMU control. The NuroSleeve can leverage up to two IMUs for splint and/or FES control. In IMU mode, the device can be operated in two different configurations: continuous or discrete. Continuous control configuration uses a multi-threshold approach and makes use of the IMU's 3D orientation data to continuously control the linear actuator and/or FES effectors. The discrete control configuration uses a tap-and-go control approach, in which the system uses the IMU 3D acceleration data to fully open or close the hand; each tap on the IMU sensor toggles between extraction and retraction of the linear actuator and/or FES stimulation. In other words, the IMU sensors can function as a toggle switch to control a two-state machine based on the status of the effector.
The Splint
Figure 2: Exploded view of NuroSleeve splint components including thumb, fingers, arm, and cuff sections.
The NuroSleeve 3D printed custom splint consists of four main sections: forearm, cuff, thumb, and fingers, as shown in Fig. 2. The fingers section facilitates the opening and closing of the hand and is assisted by a splint-mounted low-profile, lightweight electro-mechanical linear actuator (PA-07, Progressive Automations, Arlington, WA) [42]. The linear actuator has a 50mm stroke length and an integrated current limiting circuit as a safeguard mechanism to avoid overtravel. The stationary part of the linear actuator is connected to the forearm section while the actuated rod is connected to the fingers section (see Fig. 2). When the linear actuator rod is extended forward, the finger piece assists the user with grasping, when the rod is retracted, it assists with hand opening. This allows the user to achieve functional flexion and extension of the metacarpophalangeal (MCP) joint of the affected hand. The forearm section of the splint has multiple endpoint modification holes for mounting the linear actuator so that its position can be adjusted. This allows for changing the start and end position of the MCP joint’s flexion and extension while keeping the range of motion (ROM) fixed. This approach allows the OT to customize the individual’s ROM endpoints, based on their clinical conditions and functional needs. Moving the linear actuator distally along the forearm, for instance, enables the user to grasp smaller objects, whereas mounting the linear actuator proximally can facilitate the grasping of larger objects. As part of the NuroSleeve calibration process, safe and optimal hand motion and grasp are carefully verified and validated by an OT to avoid any potential for injuries, such as repeated hyperflexion or hyperextension of fingers and soft tissue injuries [43].
Digital Orthotist Process
To create a user-centric device that effectively matches the user's hand anatomy and functional requirements, we devised a digital splint design process that combines modern industrial design and occupational therapy techniques. This “digital orthotist” process (as shown in Fig. 3) begins with a 3D scan of the subject’s forearm and hand using the Creaform Go!SCAN 3D scanner [44], which features a volumetric accuracy [45] of 0.050 mm ± 0.150 mm/m [46] and uses proprietary software (VXmodel) [47] to create a watertight model by removing scan artifacts and superfluous information (e.g., chest-related data), overlapping or coarse surfaces, and holes. Once created, the individual watertight model is then imported into Rhinoceros [48], a 3D CAD software that features powerful design and modeling tools ideal for the creation of the custom 3D-printable UE splint. To improve the reliability and repeatability of the "digital orthotist” process, automation scripts are created within Grasshopper [49], a parametric programming tool and native plugin of Rhinoceros. While some common design elements, such as the hinge, endpoint modification holes, and joystick housing can be rescaled and modified for use in multiple models, the profile of the forearm, cuff, thumb, and fingers section of each model are unique, custom-designed by processing the user’s UE 3D scan data.
Once the splint customization and creation have been completed, the 3D splint files are printed on a Markforged X7 industrial 3D printer [50] using the Onyx™ micro carbon fiber-filled nylon [51]. Compared to other 3D printing materials, such as Nylon or Acrylonitrile Butadiene Styrene (ABS), Onyx™ produces orthoses with superior chemical resistance, rigidity, and flexural stress [52, 53]. All sections of the splint are printed such that the skin-contact surfaces are face up and thus not in contact with the build’s support scaffold. This ensures that the skin-contact surfaces are as smoothed as possible to reduce skin irritation. Printing a complete NuroSleeve splint requires approximately 48 hours.
After printing and assembly of the splint, fabric and straps are added at specific locations to secure the user’s forearm, hand, and fingers in place. Namely, custom Oly Fun fabric wraps are attached to the finger piece and thumb piece to create a mitten-like pocket for the fingers, Oly Fun fabric was chosen because it is non-stretchy, Fig. 3: The “Digital orthotist" process begins with a 3D scan of the individual's hand and forearm. The resulting 3D mesh is then processed and imported into Rhinoceros, where with the help of custom Grasshopper scripts is converted into a personalized 3D-printable splint. Finally, the 3D-printed splint is built and assembled.
allowing the fingers to stay open despite resistance, and it is non-woven, so it does not fray like other fabrics. It also breathes well, which reduces sweat and the likelihood of skin irritation. Finally, Rolyan straps and hook-and-loop tapes were used to secure the cuff and thumb sections to the forearm, preventing the clamshell hinge from opening. Rolyan straps, which are recognized for their softness and flexibility, were chosen to reduce the risk of causing skin irritation.
The Functional Electrical Stimulation
As depicted in Fig. 1, for FES, the NuroSleeve incorporates a two-channel commercially available neuromuscular stimulator, the Chattanooga Continuum™ [54]. The stimulator can be controlled by any of the input control signals of the NuroSleeve. By stimulating the individual's specific muscles electrically, it is possible to generate retraction or extraction at the relevant joint. The OT can alter the intensity of stimulation based on the individual's body structure and the specific muscle being stimulated.
The NuroSleeve Main Control Unit
The MCU enables control of the effectors by the various control sensors. Its core component is an Arduino Nano controller module [55] with an ATMega328p [56] microcontroller, which runs the NuroSleeve firmware and allows for data collection from up to two wired Bosch BNO055 [57] intelligent 9-axis IMUs ( for motion control), up to two MyoWare 2.0 Muscle Sensors [58], (for EMG control), an ELECHOUSE Voice Recognition Module V3 [59] (for voice control) and/or a small joystick controller (for manual control) A description of the device control logic and operating principles are provided in the section below titled "Multi-threshold Control Approach".
Bluetooth connectivity between the clinical software and external devices, such as BCI systems, has been implemented into the MCU via the onboard DSD TECH HC-05 [60] Bluetooth module, and the NuroSleeve proprietary Bluetooth communication protocol allows easy integration of third-party applications. The NuroSleeve device (including the linear actuator and the MCU) is powered by a single TalentCell 12V 3000mAh Lithium-ion (Li-ion) battery pack [61], consisting of three 18650 Li-ion batteries in series. To ensure adequate electrical safety, the battery pack has an integrated circuit for protection against over-charging, over-discharging, and short circuits.
Clinical Software Suite
The Python-based clinical software suite has been developed in-house to provide real-time management and configuration of the NuroSleeve system via Bluetooth. The software features two separate versions: a Clinical Software Suite and a User Software Suite. The Clinical Suite is designed to be used by a trained therapist and can be used to (1) customize input and output mapping; (2) adjust sensitivity and use of the input sensors; (3) visualize sensor data and settings in real-time; and (4) access and store the user compliance data from the MCU. The user suite is intended for home usage by individuals and their caregivers; it has less functionality than the Clinical Suite but permits the user to adjust the sensitivity of the sensors.
User Compliance Data Monitoring
The NuroSleeve is capable of logging user compliance data and storing it in a dedicated flash memory chip inside the MCU. This feature allows therapists and clinicians to monitor at-home device use between clinical appointments, and enabling Remote Therapeutic Monitoring (RTM), for which the U.S. Centers for Medicare and Medicaid Services (CMS) recently established payment policies [62]. The monitored compliance data can include the total time of device use and the amount of time spent opening and closing the hand and the time spent in each operation mode (i.e., joystick, voice, IMU, EMG), since the clinician’s last reset. These data which can help inform therapy goals and outcomes.
Multi-threshold Control Approach
The control strategy of the NuroSleeve can be customized to the individual's physiological needs and preferences through a quick calibration and setup phase integrated into the clinical software suite. The primary customization involves the mapping between selected input control signals (sensors) and desired outputs (effectors). Together, the individual and the treating therapist can determine which of the available inputs is most effective in controlling the effectors for that individual. If IMU inputs are selected, the sensors can be configured for continuous or toggle control. From there, the therapist and the individual can select the optimal sensor placement and activation movement for effector control. For continuous control, the effectors will be controlled by the threshold crossings of two custom-determined thresholds of the selected input signal (higher threshold and lower threshold). A positive crossing of a higher threshold triggers an effector command, while a negative crossing of a lower threshold triggers the opposite command. When the signals fall between two thresholds, the effector maintains its current state. The following equations describe how a selected sensor channel can determine the output command to a selected effector.
$$Effector\: Command\left(t\right)=\left\{\begin{array}{c}OPEN,\: if(Sig\left(t\right)>HighThr) \\ CLOSE,\: if(Sig\left(t\right)<LowThr) \\ HOLD,\: if\left(LowThr\le Sig\left(t\right)\le HighThr\right),\end{array}\right.$$
where:
\(Sig\left(t\right)\) is the current input signal channel (sensor) value at time t
\(HighThr\) is a fixed higher threshold value configured with the software
\(LowThr\) is a fixed lower threshold value configured with the software
\(OPEN, CLOSE and HOLD\) are the possible output commands at time t
If discrete control is selected, the NuroSleeve’s effectors are controlled with the IMU sensor acting as a toggle switch. In this configuration, the IMU signal of interest is acceleration and threshold crossing of a single threshold is used to change the state of the effector (OPEN/CLOSE for motor, ON/OFF for FES). This setup is ideal when the user desires to trigger the opening or closing of their hand by tapping the IMU sensor.
$$Effector\: Command\left(t\right)=\left\{\begin{array}{c}FULLY\: OPEN,\: if(Status==CLOSED\: AND\: Sig\left(t\right)>Thr)\\ FULLY\: CLOSE,\: if\left(Status==OPEN\: AND\: Sig\left(t\right)>Thr\right) \end{array}\right.$$
It is important to emphasize that even in the discrete control mode configuration, the device control commands are updated continuously, namely, a new direction command is generated and sent to the effector at every firmware runtime update (time step = 20ms).
Non-clinical Testing
A battery of electrical and mechanical bench tests was conducted with the NuroSleeve to ensure that its specifications met the device requirements. The requirements evaluated included hand splint ROM, grasp force and speed, battery life, and total life cycle of the splint. While manual measurements were carried out for most requirements, accelerated life testing was also implemented using two specific setups as shown in Fig. 5. The first accelerated life test focused on the mechanical characteristics of the device, repeatedly opening and closing the hand onto a simulated test load placed below the palm to simulate object grasping. The repeated hand movements (opening/closing) were programmed to last four seconds each, while two seconds of rest were introduced between consecutive movements, resulting in a 66% linear actuator utilization (duty cycle). To fully automate the test, two force sensors and a microcontroller were integrated into the test setup to continuously monitor the device operation and collect data, including identifying the breaking point accurately.
The second accelerated life test deployed a similar setup with the addition of two springs connected to the splint finger piece. The purpose of the springs was to induce a constant load of 13.3N in both extension and flexion to replicate the realistic hand movement force and effect of a spastic hand (which is opposed to extension at the MCP joint), which may decrease the splint's lifespan.
Figure 5: Test Bench for the NuroSleeve with added resistance with springs
To determine the battery life of the NuroSleeve, its power consumption was analyzed. In standby or FES mode, the NuroSleeve PCB with all sensors connected consumes 175mA of continuous current; movement of the linear actuator increases the current consumption to between 275mA to 375mA, depending on the torque generated. Assuming a nearly linear relationship between battery capacity and operating time due to the extremely low discharge current rate [63], the battery life of NuroSleeve is estimated to be between 8 and 17 hours.
Clinical Trial
In addition to the extensive non-clinical bench electrical and mechanical testing, the NuroSleeve has been evaluated in a clinical setting by various stakeholders including users, therapists, caregivers, and physicians. Continuous integration of stakeholder feedback is crucial to the development of a device with practical utility. The NuroSleeve is currently being evaluated in a clinical trial (NCT04798378), approved by the Thomas Jefferson University Institutional Review Board (IRB). Consented and enrolled participants complete an 8-week rehabilitation program that incorporates the device into occupational therapy sessions and ADLs at their homes.
In brief, the trial consists of an initial (pre-intervention) clinical outcome assessment session, which establishes a baseline of each participant’s UE functional ability. The Canadian Occupational Performance Measure (COPM) [64] is also administered to identify three to five activities across multiple domains (including work, self-care, and leisure) in which the participant desires to improve their functional performance or satisfaction. Each participant then undergoes 3D scanning and receives a customized NuroSleeve, after which they engage in eight weeks of outpatient occupational therapy sessions that incorporate the device. During the sessions, the OT trains the participant on how to incorporate the NuroSleeve into their selected activities at home, so they start to use the NuroSleeve independently during daily activities. After the 8-week intervention period, the standardized outcome assessments are repeated, with and without the device being worn.