Nowadays Artificial Intelligence (AI) has become a fundamental component of healthcare applications, both clinical and remote, but the best performing AI systems are often too complex to be self-explaining. Explainable AI (XAI) techniques are defined to unveil the reasoning behind the system’s predictions and decisions, and they become even more critical when dealing with sensitive and personal health data. It is worth noting that XAI has not gathered the same attention across different research areas and data types, especially in healthcare. In particular, many clinical and remote health applications are based on tabular and time series data, respectively, and XAI is not commonly analysed on these data types, while computer vision and Natural Language Processing (NLP) are the reference applications. To provide an overview of XAI methods that are most suitable for tabular and time series data in the healthcare domain, this paper provides a review of the literature in the last 5 years, illustrating the type of generated explanations and the efforts provided to evaluate their relevance and quality. Specifically, we identify clinical validation, consistency assessment, objective and standardised quality evaluation, and human-centered quality assessment as key features to ensure effective explanations for the end users. Finally, we highlight the main research challenges in the field as well as the limitations of existing XAI methods.

Personalised training of motor and cognitive abilities is fundamental to help older people maintain a good quality of life, especially in case of frailty conditions. However, the training activity can increase the stress level, especially in persons affected by a chronic stress condition. Wearable technologies and m-health solutions can support the person, the medical specialist, and long-term care facilities to efficiently implement personalised therapy solutions by monitoring the stress level of each subject during the motor and cognitive training. In this paper we present a comprehensive work on this topic, starting from a pilot study involving a group of frail older adults suffering from Mild Cognitive Impairment (MCI) who actively participated in cognitive and motor rehabilitation sessions equipped with wearable physiological sensors and a mobile application for physiological monitoring. We analyse the collected data to investigate the stress response of frail older subjects during the therapy, and how the cognitive training is positively affected by physical exercise. Then, we evaluated a stress detection system based on several machine learning algorithms in order to highlight their performances on the real dataset we collected. However, stress detection algorithms generally provide only the identification of a stressful/non stressful event, which is not sufficient to personalise the therapy. Therefore, we propose a mobile system architecture for online stress monitoring able to infer the stress level during a session. The obtained result is then used as input for a Decision Support System (DSS) in order to support the medical user in the definition of a personalised therapy for frail older adults.

The extremely pervasive nature of mobile technologies, together with the user's need to continuously interact with her personal devices and to be always connected, strengthen the user-centric approach to design and develop new communication and computing solutions. Nowadays users not only represent the final utilizers of the technology, but they actively contribute to its evolution by assuming different roles: they act as humans, by sharing contents and experiences through social networks, and as virtual sensors, by moving freely in the environment with their sensing devices. Smart cities represent an important reference scenario for the active participation of users through mobile technologies. It involves multiple application domains and defines different levels of user engagement. Participatory sensing, opportunistic sensing, and mobile social networks (MSNs) currently represent some of the most promising people-centric paradigms. In addition, their integration can further improve the user involvement through new services and applications. In this article we present SmartCitizen app, an MSN application designed in the framework of a smart city project to stimulate the active participation of citizens in generating and sharing useful contents related to the quality of life in their city. The app has been developed on top of a context- and social-aware middleware platform (CAMEO) able to integrate the main features of people-centric computing paradigms, lightening the app developer's effort. Existing middleware platforms generally focus on a single people-centric paradigm, exporting a limited set of features to mobile applications. CAMEO overcomes these limitations and, through Smart- Citizen, we highlight the advantages of implementing this type of mobile application in a smart city scenario. Experimental results shown in this article can also represent the technical guidelines for the development of heterogeneous people-centric mobile applications embracing different application domains.

WiFi Direct introduces new opportunities to deploy real opportunistic networks through users' mobile devices. However, its original specification does not take into account all the parameters that can emerge from an opportunistic network scenario, not only in terms of technical requirements (e.g., available resources and connectivities) but also of users characteristics and profiles, which can heavily influence the system's performances and devices' interactions. In this work we investigate the feasibility of creating opportunistic networks on top of WiFi Direct framework by analyzing the protocol's performances in real scenarios with a variable number of mobile devices. Experimental results show the times required to form a group of variable size and the best configurations to support opportunistic networking operations and upper layer applications.