In the last few years, the world witnessed an exponential increase in digital wellness wearables that can collect data in real-time about physical, chemical, and physiological properties of the body, to assess wellbeing and to contribute to the early detection of a variety of health conditions.
Already 10 years ago, our group developed and made available to the community at large, the BITalino platform – a hardware device and associated software providing a window to the inner-self through biosignal monitoring, including Electrocardiography (ECG) and Electro-Dermal Activity (EDA), just to name a few. Its modular and flexible design had the dual purpose of being ready to be used without requiring technical (hardware or software) skills, but also to be split and integrated into the design of novel systems.
Our vision encompassed both the design of wearables (e.g. chest bands and wrist bands for continuous biosignal monitoring) and disruptive innovation, putting forward the concept of invisibles, i.e., the integration of smart sensor technology into everyday objects.
Our early work included the development of prototypes such as keyboards, armrests, automotive steering wheels, but more recently smart sensors have been applied more widely, even in unexpected scenarios, such as toilet seats.
Due to the prevalence of CardioVascular Disorders (CVDs), ECG is one of the body signals that rises more interest in the research community. This has also motivated some of the flagship research of our group over the last decade. Focusing on a single modality (ECG), by combining signal acquisition, signal processing, machine learning (ML), and artificial intelligence (AI) techniques, we are unveiling its potential in diverse application domains, namely in health conditions, emotion assessment, and biometric identity recognition.
Despite the intra-subject variability and associated challenges in data analysis, the ECG enables person identification based on inter-subject variability associated with intrinsic unique physiological characteristics of each person, which is reflected in the morphology and dynamics of this signal modality. In fact, this is the basis of a patented and award-winning technological portfolio created by our group.
Emotions also lead to changes in ECG dynamics, as well in other biosignals, namely the EDA. Exploring multi-modality, the ongoing work of the group addresses the real-world assessment of emotional responses to film-based elicitation materials and other performing arts. By exploring unsupervised and supervised learning methods, we aim at characterizing emotional responses both under individual and collective elicitation settings, to evaluate, unveil, and model the impact of group interaction in emotional responses.
Concerning health, in addition to cardiac-related disorders, ECG has proven to be highly relevant in the detection and prediction of epileptic crises. Major contributions of ongoing work in this area are putting forward holistic perspectives, integrating more modalities and additional sources of information, where invisibles, wearables, and smartphones play an essential role working synergistically as a constellation of monitoring devices. By exploring ML & AI techniques, application to physiological data, contextual information, and behavioral patterns, continuously monitored in both clinical and ambulatory settings, we aim to better predict, and possibly prevent, epileptic crisis events, through personalized care.
The wealth of information gathered poses challenges in terms of data analysis, but it also raises privacy-related vulnerability concerns. In our work, innovation and scientific achievements put emphasis on privacy by designing solutions that go beyond data anonymization, secure data protection, or restricted access. These include the development of privacy-preserving non-invertible data transformations already at the data acquisition device level, as well as beyond state-of-the-art ML & AI techniques exploring novel data representations based on higher-order dissimilarity measures and/or embedding techniques.
The diversity and complexity of the topics addressed by our group build on the collaborative work of highly committed interdisciplinary and multidisciplinary teams, involving engineers and clinicians, in close collaboration with national and international institutions. Together, research and development are driven to question and provide solutions to real-world problems that may empower society towards better-personalized health, self-awareness, and wellbeing.
Ana Fred is a member of the Pattern and Image Analysis Lisbon (PIA-Lx) research group within IT.
Image credits: IT
Visit her profile: