Autonomous socially assistive robotics in pediatric clinical practice

Resumen

The development of new devices to support neurological recovery is a current challenge for clinical professionals and engineers [Tapus et al. 2007]. Particularly, in the last decade, robotic applications have demonstrated their great potential as novel approaches [Druzbicki et al. 2013]. Socially Assistive Robotics refers to those robots that provide assistance to human beings through social interaction. This technology is particularly interesting in healthcare domains since it is able to elicit more favorable responses to the treatment [Okamura et al. 2010]. All these approaches start from the same hypothesis: the interaction provided by a social robot helps patients to get engaged with the treatment, in addition to automatic data gathering and reporting, helping to relieve the workload of healthcare professionals while reducing the socio-economic costs. However, socially assistive robotics is a novel multidisciplinary research line at an early stage of development and evaluation, and with a large number of challenges ahead. Contributing to the evolution of knowledge requires efforts to understand the foundations of the different disciplines that it encompasses. Under this context, this thesis arises from four foundations: neurorehabilitation, socially assistive robotics, gamification and artificial intelligence. The integration of these fundamentals aims to design a child-robot interaction framework to enhance the pediatric clinical practice. The designed framework is provided with an intelligent system, so that no engineer is required either to control the interaction or to adapt the system. During the development of this thesis the framework has been used and evaluated in two different tasks: pediatric rehabilitation (NAOTherapist) and motion encouragement. Being the first one the central application of the presented work. In NAOTherapist, child-robot sessions are composed of playful immersive activities based on reward and positive reinforcement to improve motivation and, therefore, adherence to treatments. The ultimate goal is to demonstrate the feasibility of this framework in real healthcare settings, so a user-centered prototyping is proposed by involving the user during each development phase. A prototype was initially evaluated with more than 120 of typically developing children, obtaining a generalized high degree of active engagement [Pulido et al. 2017]. After that, three evaluation scenarios exposed the platform to the real practice: a first contact to get closer to the target individual, a long-term experience to determine personalization needs [Pulido et al. 2019], and an intensive intervention to evaluate the motivation and adherence to treatment. About 20 pediatric patients participated in the studies with very promising results. In all cases, the sessions with the robot provided a greater motivation compared to the conventional treatment, getting patients to exceed the objectives marked by the experts. Positive reinforcement and rewarding the patient were fundamental aspects to maintain motivation. The robot autonomy was also a key point, so making the robot taking its own decisions improved the perception of social entity. The interviewed relatives detected functional and self-esteem enhancements in their children, and experts confirmed the system utility and usability for application in pediatrics. Additionally, it is important to note that many of the related SAR platforms for physical rehabilitation have not had experience with patients in clinical settings, and those that had it, were a proof of concept that was not continued. In addition, most of them lack autonomy and do not integrate motion tracking and patient monitoring systems. NAOTherapist platform is a non-contact motor rehabilitation system that has been extensively evaluated and progressively enhanced in different evaluation episodes with very promising results. The first insights reveal the utility and feasibility of the proposed framework for its use in pediatric neurorehabilitation interventions: 1) the system conforms to the clinical guidelines fulfilling the proposed objectives, 2) the robot as a social communication interface guarantees an active engagement improving the experience of the interventions, 3) the integration of immersion mechanisms and rewards encourage the motivation of the patients, improving the adherence to the treatment, and finally, 3) the cognitive architecture has proven to be enough autonomous and robust to face the clinical practice relieving the workload of healthcare professionals. Each of these evaluations allowed the platform to evolve, incorporating functionalities and detecting new future needs. In total, 244 different children (21 of them pediatric patients) interacted with NAOTherapist in a total of 429 sessions executed without significant incidences. Of these 429 sessions, 206 were in clinical settings. Regarding to the rest of the stakeholders, 11 relatives and 20 clinical experts were consulted through interviews and questionnaires. Despite these extensive evaluations, there is still much work to do to achieve the ultimate intended goal: ``\textit{the incorporation of technologies, such as NAOTherapist, in routine therapeutic procedures}''. Although these studies offer an initial experience from different scenarios in the search of new requirements, the results presented here help to establish a solid base to extend this line of research aiming at offering novel tools to healthcare professionals.