An assistive upper-limb exoskeleton controlled by multi-modal interfaces for severely impaired patients: development and experimental assessment

Highlights

• The Bridge exoskeleton is an active upper limb assistive device for disabled people.

• The device is controlled with interfaces suitable for severely impaired patients.

• Joystick-based and vocal interfaces are used to detect user’s intention.

• An inverse kinematic algorithm tracks the hand position and drives joints angles.

• With Bridge, users with muscular dystrophy increased their upper limb functionalities.

Abstract

Active exoskeletons can help adults with muscular dystrophy regain independence and self-esteem, which have been limited due to their severe and progressive muscular weakness. A four degrees-of-freedom fully actuated upper limb exoskeleton, equipped with a spring-based anti-gravity system, has been designed, prototyped, and tested on end-users. While wearing the exoskeleton, the user directly controls the system by actively driving the end-effector position (i.e., the hand) using a joystick or vocal control. The exoskeleton’s kinematic model has been determined so that, given a desired user’s position in the task-space, a differential inverse kinematics algorithm computes the desired joint-space motion trajectories. The dynamic model was investigated in the vertical plane, demonstrating that gravity torques were considerably higher than velocity-induced and inertia torques, which have been therefore neglected. A pilot study on 14 Muscular Dystrophy patients was conducted. Outcome measures included: (i) externally-assessed functional benefit evaluated through the Performance of Upper Limbs module, (ii) self-perceived functional benefit assessed through the ABILHAND questionnaire, and (iii) usability of the system assessed through the System Usability Scale. All participants strongly increased their range of motion, and they were able to perform activities that were not possible without the exoskeleton, such as feeding. The externally-assessed and self-perceived functional improvements were statistically improved when wearing the exoskeleton (PUL p-value$=0.001$, ABILHAND p-value$=0.005$). System usability was evaluated to be excellent. Patients’ feedbacks were encouraging and outlined future development steps.

Introduction

“From the patient’s point of view, improvement is measured by regaining lost abilities and by being able to do something – anything – today, she/he couldn’t do yesterday” [1]. People who have muscular dystrophy (MD) progressively lose the ability to walk, stand, and control the function of their arms. This hinders them from performing activities of daily life (ADLs), participating socially, and being independent [2], [3]. Thanks to improved medical care and technical possibilities, life expectancy has increased rapidly [4]. Consequently, most of these patients do not have functional arm movements for more than half of their life, if unsupported. Arm disability has been demonstrated to play a key role in reducing patients’ autonomy and worsening quality of life [5], [6]. In recent decades, some efforts have been made in the field of upper limb assistive devices (ADs). A recent systematic review (14 studies, 184 participants) investigated assistive devices effects on upper limb functionalities of patients with neuromuscular disorders [7], [8]. This study demonstrated that assistive devices significantly increase the ability to perform daily life activities for people affected by neuromuscular diseases, with an associated large effect size in favor of ADs. However, ADs assistance should be tailored to the residual ability of target end-users. To this aim, we divide users with respect to residual ability as evaluated by clinical scales [9]: (i) slightly impaired (i.e., muscular weakness when performing movements and reduced range of motion), (ii) moderately impaired (i.e., from partial to active movement in the absence of gravity), and (iii) severely impaired (i.e., no contraction or traces of contraction which do not result in movement). Recently, we carried out a crossover clinical trial with the two most diffused upper limb assistive devices in the market targeting dystrophic patients [9]. The results showed that, depending on the level of disability, different outcomes are observed for the two tested devices, but none of them cope with the severely impaired end-users.

Assistive devices can be classified as passive, semi-active, or fully-active devices [9]. Most of the exoskeletons developed to assist upper limbs are passive solutions, which require the user to have residual ability to pilot the device. To help the user to fulfill the desired action, an assistive device might move the user’s arm in the space to reach the target (as in the case of exoskeletons or end-effector devices) or independently fulfill the action (as in the case of external manipulators). When coming to exoskeletons, passive devices include the passive A-gear [10], the Exoskeletal Meal Assistance System (EMAS III) [11], or the commercial device Wrex (Jaeco, USA). Wrex has been successfully tested on people with MD with mild impairment, increasing upper limb functionality [9], [12], [13]. These solutions, however, are not suitable for severely impaired patients, given the active contribution required to the user [9].

Some semi-active ADs are available on the market, such as Armon Ayura (Microgravity Products BV, The Netherlands) [14], Electrical Top/Help (Focal Meditech BV, The Netherlands) [15], or the Neater Arm Support (Neater Solutions Ltd, United Kingdom). These devices do provide upper limb(s) support through a forearm brace that provides anti-gravity support. The compensation level is usually remotely controlled and can be adjusted according to users’ needs. However, they still require the user’s muscular effort to drive the device and fulfill the movements. A recent study on MD patients demonstrated their effectiveness on moderately impaired subjects but not on users with very poor residual functional ability [9], [15]. These devices are characterized by an imperfect gravity compensation, and they require sufficient muscle strength to overcome inertia [16].

For the assistance of severely impaired patients, a fully active device is the only solution. While there is a considerable number of powered upper limb exoskeleton in the rehabilitation field (e.g., Armeo Power — Hocoma, Switzerland; Harmony SHR — Harmonic Bionics, USA), few efforts have been devoted to the development of fully active exoskeletons in the assistive technologies field. Kooren and colleagues [17] developed the Active A-Gear, a wearable arm exoskeleton with five degrees-of-freedom (DOFs). The human–machine interface of this device consists of a force sensor attached between the forearm link and the arm to measure the interaction forces between the user and device. However, this device was tested only on one healthy subject. Another example is the A-Arm [18], a planar fully-active arm assistive device. The end-user controls it with a force sensor or through the EMG surface signals acquired from four different muscles (biceps, triceps, deltoid anterior, and deltoid posterior). The A-Arm was tested on one person with MD in a severely impaired condition, and it increased the functional workspace of the user’s arm. However, it only supports movements in the horizontal plane without allowing the participant to support hand-to-mouth movements, critical when coming to daily life activities (e.g., drinking, eating, etc.).

Force and EMG-based controlling strategies have been originally developed for rehabilitation purposes [19], [20], [21]. Such interfaces have been recently tested on three MD patients [22]. In the study, one patient can be classified as moderately impaired (i.e., he/she can raise the hand to the mouth but cannot raise a glass of water, following the Brooke scale classification), and two patients are severely impaired (i.e., cannot raise the hand to the mouth). While the less impaired patients succeeded in using both control interfaces, the most severely impaired subject (i.e., the one who has no useful function of the hand) could not effectively use the force-based control interface. Indeed, the force-based method requires users to have enough force in their arms to be detected by force sensors, which are not reliable for severely impaired patients. Indeed, considering MD patients, the loss of motor function occurs from proximal to distal level. The force-based control requires effort from muscles in the proximal part of the arm. Hence, they become insufficient to target the most compromised patients, especially to move in the anti-gravity direction. Finally, this human–machine interface is generally felt by patients as more fatiguing [23]. EMG control, instead, has been shown as an intuitive and natural control interface for robotic arm supports in adults with MD until the last stage of the disease [24]. However, it was found by MD patients as less intuitive, and the EMG-based interface presents several practical issues (e.g., poor long-term stability of the measurements, high sensitivity to electrode location, the time required to place the electrodes, and uncomfortable feeling due to multiple electrodes in contact with the skin for a long period of time [23]). Despite the technological advancement, patients often still experience problems controlling myoelectric ADs in daily life [25]. A relevant limitation of this approach is that it allows the simultaneous control of only one or two DOFs arm movements. In fact, patients could not simultaneously control the arm support with three DOFs [22].

Nowadays, a possible alternative to fulfill an action completely assisted by the robot are external manipulators, such as the iARM or the JACO robotic arms. They can be mounted on the user’s wheelchair to facilitate the accomplishment of ADLs in people with upper limb impairments [26], [27]. The most common human–machine interface of the JACO system is a three-axis joystick that allows the user to move the robot’s end-effector in the Cartesian space and modify the hand’s pose, and grasp objects [28]. The joystick is constituted by a cylindrical driving system and five pushbuttons permitting to switch between different control modes. However, it is very difficult or even impossible for certain users to use this kind of joystick [29]. Indeed, it is not sensitive, and its big dimensions require wide hand movements that severely impaired patients do not preserve. A possible alternative is a voice-based control, which was demonstrated to be an easy and intuitive alternative for the weakest subjects [29]. However, external manipulators do not preserve a direct interaction of the subject’s arm with the environment. These systems are not directly connected to the users’ arm and completely substitute the users’ action. Differently from the previously described exoskeletons and end-effector systems, external manipulators do not render movement proprioception. This solution could be less accepted by patients, who feel deprived of being part of the action itself.

This study’s specific research question is the design, development, and test of an upper limb assistive device for severely disabled people to fulfill the desired action towards better independence in daily life activities. We followed human-centered (or user-centered) design processes for interactive systems, an approach to interactive system development that focuses specifically on making systems usable. Usability goals, user characteristics, environment, tasks, and workflow of a product, service, or process are given extensive attention at each stage of the design process [30]. In this view, we performed a focus group with three dystrophic patients (end-users), three caregivers, two clinicians, and three engineers to derive specific requirements, which we did not find in any commercial product, nor prototype described in the literature at best of our knowledge. Specifically: (i) the assistive device should move the user’s own arm with no or very little residual force, assuring natural movement proprioception afferent feedbacks; (ii) the upper limb motion should always be under the direct control of the end-user, without using specific pre-defined trajectories or pre-defined actions to be performed; (iii) severely impaired end-users should test the prototype for usability assessment; (iv) the assistance provided by the device should be adequate to permit the end-users to fulfill activities of daily living.

For these reasons, we here propose a new assistive upper limb exoskeleton (i.e., the Bridge exoskeleton). The aim is to provide the user with an AD that detects movement intention and drives the user’s arm in the world space coordinates without performing any muscular forces. This work presents the exoskeleton system design and the efficacy validation on a pilot group of end-users. Section 2 introduces the mechatronic system design and modeling, the control system, and the human–machine interfaces. The clinical experimental protocol is described in Section 3, while the results are reported in Section 4. Finally, Sections 5 Discussion, 6 Conclusion draw, respectively, the discussion and conclusion of the work.