m-Health Solutions and Data Analysis to Detect and Predict Risky and Adverse Health Conditions in Older Adults

In this thesis, we investigated the design, development, and experimental evaluation of two m-health systems, based on the integration of IoT and wearable sensors with personal mobile devices, to monitor stress and nutrition in frail older adults. In addition, they have been empowered with data-driven AI methodologies for data preparation, analysis, and inference aimed at providing decision support to early detect and manage adverse and risky conditions in each domain. Specifically, the novel solutions presented in this thesis include a smart Decision Support System (DSS) for online and high-resolution physiological stress detection during motor-cognitive training, aimed at treatment personalisation, and a smart DSS for semi-continuous and automatic malnutrition assessment, in order to enhance traditional clinical screening tools. Performance analysis is focused on prediction accuracy as top-priority system reliability measure, by exploiting both data collected by real users through pilot studies, but also including public datasets. In addition, the proposed m-health solutions underwent a further evaluation in terms of technical reliability, User Acceptance (UA), and Quality of Experience (QoE), which are essential to assess long-term applicability and usability by both older adults and their care givers.