Context
Each year more than 13 million people worldwide have a stroke, with approximately two-thirds having persistent upper limb paresis and one-third presenting with hemispatial neglect [1, 2]. Intensive rehabilitation using challenging and individualized tasks enhance functional recovery after stroke [3]. Emerging techniques promote intensive rehabilitation and allow simultaneous motor and cognitive training, complementing conventional approaches.
Task difficulty adaptation during robotic rehabilitation
Recent guidelines recommend robot-assisted therapy (RAT) to improve upper limb strength, function and activities of daily living after stroke [4]. A review on control strategies, presented various types of task-difficulty adaptation mechanisms described in scientific literature for robotic neurorehabilitation [5]. Among these adaptive mechanisms, some are configured to automatically adjust task-difficulty using data derived from the robotic device as input to system decision-making [6]. Most commonly used computerized systems often rely on “assist-as-needed” guidance algorithms to adjust motor task-difficulty [5]. However, the way these systems’ effectiveness is validated varies in scientific literature and remains vaguely described in many cases [7, 8]. More specifically, it would be worthwhile to assess the pertinence of the decisions made by the system during training. Are the parameters determining task-difficulty during RAT well adapted to the functional profiles of subjects after a stroke?
Dynamic difficulty adjustment (DDA) during serious games training
Serious games also constitute an effective approach to stimulate upper limb recovery after stroke [9]. RAT can be combined to serious games to continuously and automatically adapt task-difficulty to match individual participants’ impairments and immediate performance during training [10]. Depending on device and game characteristics, different types of task-difficulty adaptation mechanisms have been previously described for serious games in stroke rehabilitation [11]. For example, serious games implemented on virtual reality systems can adapt game difficulty using preestablished increments, configured at the beginning of each session by a therapist [12]. Other virtual reality tools allow progressive difficulty adjustment based on individual performance [13]. This type of difficulty regulation mechanism, known as dynamic difficulty adjustment (DDA), has also been described for serious games implemented on robotic systems [14]. The objective of DDA mechanisms is to adjust task-difficulty automatically, in real time, according to user performance, creating feasible, yet challenging tasks, keeping the game in constant balance [13, 14].
DDA mechanisms present two main advantages. First, game characteristics dynamically adjust to match participants’ individual degree of impairment and performance in order to maintain an optimal challenge according to neurorehabilitation principle of increasing difficulty [15]. Secondly, DDA mechanisms lead to individualised levels of task-difficulty which could enhance human performance by maintaining a balance between motivation and learning. According to the concept of flow, in order to preserve motivation during training, task-difficulty should match participants’ skill levels and should avoid extremes (i.e. exercise through tasks that are not too easy nor too difficult) [14]. Our team developed ROBiGAME [16], a serious game implemented on an end-effector rehabilitation robot using a DDA mechanism, detailed below.
Combining motor and cognitive rehabilitation after stroke using robotic devices
In research and clinical practice, motor and cognitive impairments are usually addressed separately by different therapists. Additionally, most robotic devices are designed to solely target motor rehabilitation (i.e., not additionally including cognitive exercises) [17].
It has been suggested that combined cognitive-motor rehabilitation after stroke could lead to better improvements in motor function when compared with time-matched conventional approaches [18]. Although cognitive training constitutes an essential part of adult stroke rehabilitation, a recent systematic review underlined that cognitive exercises are insufficiently incorporated into robotic devices for combined rehabilitation in the stroke population [19]. Another systematic review identified only one study for robot-assisted cognitive training after stroke [20]. This review also highlighted that one of the main challenges of robotic rehabilitation for cognitive training remains personalisation of task-difficulty using RAT systems [20]. Indeed, participants’ impairment severity and functional deficits vary widely after stroke, leading to differences concerning rehabilitation needs and objectives.
In this proof of concept, we study the DDA mechanism of ROBiGAME, a novel robotic gamified approach, that allows combined upper limb motor and cognitive rehabilitation for attentional impairments following stroke.
Objectives and hypothesis
The primary objective of this proof of concept was to clinically validate ROBiGAME’s DDA mechanism. We evaluated whether parameters determining task-difficulty during gameplay adapted to individually match participants’ degree of motor impairments and/or hemispatial neglect following stroke. We hypothesized that ROBiGAME’s DDA mechanism would lead to a different level of difficulty corresponding to each participant’s degree of impairment.
Secondary objectives examined whether characteristics during gameplay (i.e., number of targets, number and position of visual distractors presented on screen, etc.), defining task difficulty, would rapidly adapt to reach an individualized level of difficulty.
More at link.
