Periodic Reporting for period 1 - REPAIRS (Relearning Perception Action In Rehabilitation from a Systems perspective)
Periodo di rendicontazione: 2021-01-01 al 2022-12-31
The first main objective of REPAIRS entails developing fundamental knowledge on learning new perception-action couplings. Since it is generally assumed that variability in practice is helpful in learning a new perception-action skill, several of the fellows examine how variability in training conditions can contribute to learning a new skill. For instance, we are studying how a vibrotactile device can help people with a visual impairment pick up an object. We have established that variability in object size and object location helps blindfolded people to learn to use the signals from the vibrotactile device to guide their grasping actions. Moreover, we have found that participants have an intrinsic variability in their joint angle coordination pattern that differs among individuals. Intrinsic variability, and differences therein, will be exploited in upcoming studies.
In line with the first objective, several experimental setups have been developed to study coordination in joint actions. A ‘doubles pong’ task has been implemented to study how two individuals coordinate their actions to jointly achieve a goal. The focus will be on the information that guides the actions of the players and how communication between players contributes to the goal achievement. The first steps have been made in implementing an artificial agent in the doubles pong task so that it can be applied in rehabilitation. In a different type of setup developed by other fellows, two people had to move a surface to role a ball into a target. These projects focus on how the dynamics of the action system contribute to the dynamics of goal achievement.
The second main objective of REPAIRS entails developing knowledge on requirements for translating knowledge on perception-action learning to rehabilitation practice. Therefore, fellows prepare experiments with stroke patients, people with visual impairment and people with ASD. In addition, we exploit machine learning techniques to develop a model to predict patient-specific, evidence-based treatment pathways. We are currently implementing Generative Adversarial Networks to estimate individualized treatment effects. We are also developing a rehabilitation training module in which inertial sensor technology is used in exergames. Finally, we focus on normative practices and their coordination in interdisciplinary projects to find ways of improving interdisciplinary coordination. Revealing tension between practices in different disciplines (scientific, clinic, industry), we formulate requirements for translating scientific results in rehabilitation practice.
To widely share the developed knowledge, we developed toolboxes on theories, methodologies and analyses that are open for public use and can be found on our website. For instance, we give background on the Uncontrolled Manifold Method, present a tutorial on how to explain dynamic touch from an Ecological Psychology perspective, and we give a perspective on knowledge translation and coordination of practices when people of different disciplines work together. In addition, the fellows were active in communicating their work through presentations at local, national and international scientific meetings. We have active social media accounts where instant updates are given, and the team is presented.
Our findings on intrinsic variability in arm joint movements and variability of practice in training, and the calibration methods for sensor-based joint measurements, will be translated to computer games for rehabilitation of stroke patients that potentially improve their functional abilities.
The ideas of emergent joint actions can be applied in treatment of people with autistic spectrum disorder to improve their social interaction and communication skills.
The knowledge on variability in training in using vibrotactile devices can have impact on the rehabilitation of visually impaired people. To increase the potential impact, we are currently conducting an interview study with visually impaired individuals.
Currently, we use machine-learning to develop a patient-specific, evidence-based treatment pathway for patients with low back pain and knee arthroplasty, which might affect the patient trajectory in the clinic.
The guidelines of translating knowledge between different domains might have a lasting impact on how research into rehabilitation practice is set up and implemented.