Some relevant studies have indicated that approximately 36% of these survivors (i.e. more than 9 million in 2013 only) are left with significant disabilities 5 years after their stroke, and >40% (i.e. more than 10 million) require assistance with activities of daily living.
Despite evidence that participation in formal rehabilitative therapies lessens disability after stroke, less than a third receive inpatient or outpatient therapies. Of those who do access therapies, the frequency of use varies by geographic location and socioeconomic status. In this context, the development of new strategies able to expand the access to rehabilitation to an increased number of stroke patients, also enabling home-based conduction and monitoring, are increasingly necessary both for patients, their families and for the healthcare and social services sustainability. Since many barriers could limit access to continuous physical rehabilitation for these patients, devices that complement or assist in the rehabilitation process can be of great help.
Among different approaches proposed by the scientific community, technological systems based on accelerometers seem to be among the most promising. Accelerometers are small low cost electronic devices, able to measure body parts acceleration on three axes. Many researchers have already highlighted that accelerometers have the capability to provide reliable and objective information on quantity and intensity of patient limbs movements during recovery process.
Wearable devices such as accelerometers allow to monitor exercises and daily activities. Machine learning methodologies have already been applied for modelling and contextualizing accelerometric signals to identify activity types (walking, dressing, eating, washing up, etc.) or to recognize to which rehabilitative exercise these signals are linked to. These techniques allow to estimate the recorded movement quality, providing information useful to identify the context in which movements are performed. Results of these type of studies are promising and they demonstrate that machine learning is a preferred approach for accelerometric data analysis, since able to exceed actual limits that today are hampering commercial product development for real time analysis of movement.
Within this scenario, Camlin-ARC takes its place. ARC is a platform based on wearable inertial sensors and machine learning algorithms, designed to bring the rehabilitation at post-stroke patients' home, following hospital discharge.
