Background
Over the last decades, many robotic devices have been developed for upper extremity rehabilitation after neurological disorders, for example, current established systems
include the MIT-Manus [1], the Assisted Rehabilitation and Measurement (ARM)Guide [2], the Mirror Image Motion Enabler (MIME) [3], the Bi-Manu-Track [4] and the ARMin [5]. Although the design and development of all these robotic devices have been extensively reported only a few studies were performed as part of a regular rehabilitation program and mainly focused on the effectiveness of specific training sessions or specific patient groups [6-8]. The main goal of these devices is to increase the intensity and quality of rehabilitation therapy [9] by providing well-controlled and highly repeatable conditions as well as optimized assistance to the patient [10,11]. In addition these devices are able to reduce the work load of the therapist by assisting specific movements of the patients and supporting the weight of the patients arm during therapy [12].
In the field of SCI rehabilitation passive arm orthoses are receiving increased interest, such as the Therapy Wilmington Robotic Exoskeleton (T-WREX) [13-15] and its modified and commercialized version, the Armeo Spring (Hocoma AG, Volketswil, Switzerland). These non-robotic, gravity support systems are based on an
ergonomic arm exoskeleton with integrated springs. Such devices cradle the entire arm, from shoulder to the hand, and counterbalance the weight of the patients' arm. They enhance any residual function and neuromuscular control and assist active movement across a large 3-D workspace providing an augmented feedback [16]. As there are no actuators implemented in these devices, all movements are generated by the users themselves.
The passive orthoses and robotic devices are equipped with sensors responsible for the assessment of their multiple degrees of freedom as well as to display the movement of different joints. Therefore, enormous amounts of data are collected during training that could be used not only to monitor the training session (intensity, duration, frequency etc.) but also to follow changes in the functional impairment. Recently studies have started to focus on the effectiveness of training with a gravity compensation device in different patient groups [16-18]. However, psychometric properties (reliability and validation) that account for clinical and patient-relevant aspects (such as the influence of the positioning of the patient) have not been sufficiently addressed.