Leader: Egidio De Benedetto (UNINA); Other collaborator(s):
The objective of the task are: 1) Analysis of technological solutions related to risk-health prevention. 2) Obtain and digitize information about patient journey necessary for prevention and assistance. 3) Use results of task 3.1 to identify smart solution to be developed. 4) Develop Machine Learning and Artificial Intelligence algorithms able to offer user-friendly solutions, recognizing and learning patients’ habits as well as making the caregiver’s intervention immediate and effective. 5) Develop of a smart platform with a dashboard for caregivers and patients to manage and keep track of elderly people in different contexts. 6) Implement technological policies to foster personalized prevention considering data privacy and security.
Brief description of the activities and of the intermediate results
In the technological context of brain-computer interfaces, a paradigm that has also been widely studied and investigated recently has been identified for neurodegenerative diseases, namely that of motor imagination. It is an active paradigm, that is, it is based on voluntary modulation of brain potentials and consists of imagining a movement without actually performing it.
In the first months of activity, a machine learning algorithm based on the "Filter-Bank Common Spatial Pattern" (FBCSP), which is widely used for motor imagination classification, was developed. In order to test this algorithm, initial data were collected on healthy subjects to prove the validity of a suitable setup for neuromotor exercise.
In the next phase, the algorithm will be refined by exploiting advanced artificial intelligence techniques, and a first prototype exergame for rehabilitation will be developed. For this, additional data will have to be collected, especially from patients with neurodegenerative diseases.
Main policy, industrial and scientific implications
The goals to be achieved by pursuing these activities in the AGE-IT project are:
Brief description of the activities and of the intermediate results
Between April and June, an upgrade of the machine learning system was completed with the integration of an asynchronous pipeline. This implementation improved the flexibility of the user interface, allowing users to freely choose what to imagine and when to do it. The developed interface uses an avatar moving in a simulated environment, collecting objects (coins) along a road divided into three lanes. Users can imagine moving their right or left hand to allow the avatar to move right or left respectively or relax while keeping the avatar's position unchanged.
Main policy, industrial and scientific implications
From a strategic point of view, this system could promote more natural, customisable and accessible interactions, particularly for users with disabilities, marking an advance in human-machine interfaces. On an industrial level, the innovative use of avatars and imaginative interaction technologies opens up new perspectives, particularly in the fields of gaming, neurological rehabilitation and psychological well-being. Scientifically, the project introduces new opportunities for research into asynchronous machine learning systems and the understanding of human cognitive potential. The analysis of imagery brain responses offers significant contributions to the understanding of brain-computer interfaces (BCI), with both therapeutic and technological implications. The system could improve attention, stimulate neuroplasticity and slow cognitive decline in patients with neurodegenerative diseases, facilitating new rehabilitation strategies.
Brief description of the activities and of the intermediate results
In July-September, the developed machine learning systems were tested on three patients suffering from Parkinson's, dementia and autism, respectively. These tests evaluated the effectiveness of two types of interfaces: one synchronous (guided by directions) and one asynchronous (free choice of imagined movement). However, all patients experienced difficulties in using the asynchronous interface, as they were unable to control the avatar correctly. As a result of these problems, they switched to the use of a synchronous interface, in which a ball moves left or right according to the users' imagined movement. The patient with dementia was excluded after the second use, since, during a preliminary test with executed movement, she frequently confused the limbs to be moved, despite the instructions received.
Main policy, industrial and scientific implications
In terms of policy, the adaptation of interfaces according to the specific needs of patients highlights the importance of developing more flexible and customisable technological solutions, favouring inclusivity for those with cognitive and motor disorders. From an industrial point of view, the switch to a different interface, which has improved interaction in autistic and parkinsonian patients, shows that the technology industry can benefit from the continuous optimisation of these systems to better meet clinical and rehabilitation needs. On a scientific level, the difficulties encountered by dementia patients and the analysis of cognitive and motor responses indicate that the brain-computer interface (BCI) needs further improvement. Before starting a large-scale experimental campaign, the interface needs to be made more engaging and robust to better adapt to different patient groups, thus improving therapeutic efficacy and user experience.
Brief description of the activities and of the intermediate results
Between October and December, the system was tested on a large sample of healthy subjects in order to retrieve a baseline performance, while highlighting further bottlenecks to tackle. The healthy subjects were able to play to the two levels of the proposed game, with increasing velocity and hence difficulty. Moreover, the second level speed was adapted in real time to the gaming performance. Among the results, this adaptability appeared interesting even for the ensuing patients’ scenario. Even the engagement, the mental effort, and the overall user experience were monitored by means of self-assessment questionnaires. The classification accuracy with the wearable and portable EEG acquisition device, which will be crucial for telerehabilitation and home care of future patients, resulted about 55 % on three-classes, with improvements foreseen as the training continues.
Main policy, industrial and scientific implications
The first adaptivity feature introduces with the second level of the serious game highlighted the need to invest on that for a proper treatment of patients with neurodegenerative diseases. Further studies will also investigate assistive movement while monitoring the degree of engagement of the patient, so that the degree of assistance can be refined on the specific patients’ conditions. From an industrial point of view, having a multi-level game with adaptive features helps to bring part of the treatment at home and to improve its efficacy by a personalized approach, which is not too difficult nor too easy to prevent disengagement mechanisms. On a scientific level, a first journal paper was submitted on the topic, and the need to focus on specific neurodegenerative diseases involving motor impairment arose while preparing this drafting and reasoning on the results. Therefore, future developments will particularly focus on Parkinson’s disease and/or multiple sclerosis while also evaluating proper effectors in response to the motor imagery tasks.
Preparation of a manuscript on the machine learning system aforementioned