WideHealth Podcast Series

This podcast is part of the "WideHealth Seminars". This project (widehealth.eu) has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 952279  

Speaker: Milene Teixeira 

Title: Automating the Generation of Dialogue Managers for Healthcare   

Abstract: Health dialogue systems are required to respect some special requirements such as predictability and reliability. Given the complexities of the health domain, these systems frequently rely on knowledge-based techniques. However, the automated generation of reliable policies is a challenging task and it remains an open problem. This talk will first present the challenges of current techniques for dialogue management of health dialogues. Then, I will present an approach that integrates semantic awareness and AI planning which was proposed with the aim of simplifying and automating the generation of health dialogue managers. Finally, I will discuss some of the results obtained from a living lab that was conducted in the context of the WideHealth project.  

Short Bio: Milene Santos Teixeira is a Ph.D. candidate in Computer Science at the University of Trento – Italy. Her current research focuses on the integration of AI Planning and information management techniques to address health dialogues. In 2018, she concluded her master’s degree in Computer Science at the Federal University of Santa Maria, having conducted part of her research at Brock University. Milene has also collaborated with the LASIGE group (University of Lisbon) in the context of the European project WideHealth.

This podcast is part of the "WideHealth Seminars". This project (widehealth.eu) has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 952279

Speaker: Stefan Konigorski 

Title:  StudyU: A platform for conducting digital N-of-1 trials that link personalized medicine and population health research   

Abstract: Traditionally, effect estimates of health interventions have been obtained from studies of large groups of individuals. However, the derived average effects do not allow meaningful insights on whether an intervention will help a given individual – which is at the center of personalized medicine. We have developed the StudyU platform (arxiv.org/abs/2012.14201) which allows evaluating the effectiveness of health interventions on an individual level by digitally designing, publishing, and conducting so-called N-of-1 trials. In N-of-1 trials, every participant compares different health interventions of interest over time. The data generated from N-of-1 trials are hence single time series, usually within complex causal graphs, and the goal is to test interpretable effects of the interventions. The power of N-of-1 trials can be further enhanced by including sensor data to measure health outcomes. In this talk, I will introduce N-of-1 trials and the StudyU platform, present some of our work on the statistical methods for the analysis and discuss how the StudyU platform might be helpful in bridging individual-level and population-level studies by aggregating multiple N-of-1 trials.  

Short Bio: Stefan Konigorski, PhD, is a Senior Researcher in the Digital Health & Machine Learning chair at the Hasso Plattner Institute in Potsdam Germany, where he leads the Health Intervention Analytics lab. He is also Adjunct Assistant Professor in the Genetics and Genomic Sciences Department at the Icahn School of Medicine at Mount Sinai in New York. He develops statistical and machine learning methods to derive causal effects from complex observational and experimental studies, with a specific research focus on investigating personalized health trajectories and digital health interventions by using N-of-1 trials and adaptive trials.

This podcaste is part of the "WideHealth Seminars". This project (widehealth.eu) has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 952279

Speaker: Walter Maetzler 

Title:  Digital biomarkers for chronic diseases: Lessons learned   

Abstract: In recent years, many -wearable- digital devices have conquered the consumer and fitness market, and the medical and health industry also expected an enormous development boost from this advance. However, the results currently available on the detection of disease, its progression and therapy through such digital devices are rather disappointing. The regulatory bodies as well as many clinicians argue that this is mainly due to the fact that the development always starts from the technological, but not from the clinical, or even better, patient level. In this webinar, a large EU research project, IDEA-FAST, will be used as an example to show how informed digital and device-agnostic biomarkers can be developed for quality-of-life-relevant symptoms in various chronic diseases.  

Short Bio: Walter Maetzler is full professor for neurogeriatrics and deputy director of the neurology department of the University Hospital in Kiel, Germany. His main clinical interest is on Parkinson’s disease and other disorders associated with functionally relevant movement and cognitive disabilities. He leads a research group focusing on the analysis and validation of mobile sensor technology in supervised (“lab- or clinic-based”) and unsupervised (“home-based”) assessments. He is involved as principal investigator, chief clinical investigator and workpackage leader in multiple international projects investigating the potential of mobile sensor technology to improve our understanding of disease progression and treatment response in Parkinson’s disease. Examples at a European level are IDEA-FAST, Mobilise-D, Fair-Park II and Keep Control. Currently, he serves as the co-chair of the Technology task force of the Movement Disorders Society.

This podcast is part of the "WideHealth Seminars". This project (widehealth.eu) has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 952279  

Speaker: Kyle Montague  

Title:  Democratising Healthcare Technologies: Wearables to cue for drooling in Parkinson’s Disease  

Abstract: Digital technologies are rapidly transforming the healthcare landscape. Artificial Intelligence and Machine Learning are helping researchers to discover more and more about diseases – leading to new breakthroughs in treatments and cures. In recent years we seen a surge in the use of wearable technologies to track and monitor our physical activity and psychological measurements of daily life, giving healthcare professionals a greater understanding patient of symptoms and behaviours. With much of this innovation making its way into mainstream consumer devices, there is an opportunity for a new generation of self-management technologies that not only support interactions with healthcare professionals and researchers but enable individuals to take greater control of their health. The democratisation of healthcare technologies would not only allow an individual access to their health information, but it also seeks to provide the means by which they leverage that information to enrich and enhance their lives. Providing people living with Parkinson’s the know-how and resources to transform their ideas and desires into interventions and tools is key to enabling new breakthroughs. In this talk, I will discuss an ongoing project where we are designing and developing CueBand, a wearable device specifically for people living with Parkinson’s. CueBand is an open and customisable technology to support symptoms of Parkinson’s, such as cueing for decreased automatic swallowing, while also providing research grade data collection. Together with the Parkinson’s community we want to develop CueBand to create an entirely democratised infrastructure to transform the future of healthcare technologies.   

Short bio:  Kyle Montague is an Associate Professor in Computer and Information Sciences at Northumbria University, where he co-leads the Northumbria Social Computing (NorSC) research group. His research expertise is in Human-Computer Interaction and Digital Civics, with much of his work exploring novel applications and configurations of digital technologies to tackle societal challenges across a broad range of health and social care topics through both largescale approaches and small embedded participatory work with communities.

This podcast is part of the "WideHealth Seminars". This project (widehealth.eu) has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 952279  

Speaker: Cátia Pesquita 

Title: Knowledge Science for trust in AI-based biomedical and clinical applications   

Abstract: 

Biomedical and clinical applications of artificial intelligence are increasingly popular in the scientific community. However, concerns about potential bias and the lack of explainability of high-performing machine learning methods such as deep learning are limiting their adoption in practice. In this talk I explain what knowledge science is and why it is key to assess the trustworthiness of biomedical data and AI outcomes. In particular, I discuss three contexts, data, domain and user, and draw on specific examples to illustrate pitfalls and how knowledge science can overcome them.   

Short bio:  

Catia Pesquita is an Assistant Professor in Computer Science at Faculdade de Ciências da Universidade de Lisboa and a Senior Researcher at LASIGE where she leads the Health and Bioinformatics Research Line of Excellence. She has a multidisciplinary background in Biology and Computer Science, and she develops her research at the intersection between the areas of Knowledge Representation and Data Mining, with a focus on biomedical and healthcare applications. She has made internationally recognized contributions, namely in the areas of ontology-based semantic similarity and ontology alignment, winning multiple awards and competitions. She is deeply interested in how human knowledge can be communicated to computers and vice-versa.

Speaker: Mitja Lustrek 

Title: Activity recognition with a few twists: Experiences from SHL Challenges 2019 and 2020   Abstract: Sussex-Huawei Locomotion (SHL) Dataset was recorded by three people carrying four phones in different locations on their bodies for seven months. It is labelled with eight locomotion activities: still, walking, running, biking, car, bus, train and subway. It was used in three machine-learning competitions organized in collaboration with the HASCA workshop at the Ubicomp conference in 2018–20. While the 2018 challenge presented a relatively standard activity-recognition problem, 2019 and 2020 introduced a few twists. In 2019, the goal was to recognize activities with the phone in the hand location, while most of the training data was provided for the other three locations. In 2020, the goal was to recognize activities with the phone in an unknown location when carried by two different persons, while most of the training data was provided for the third person. The talk will explain how the team from Jozef Stefan Institute tackled these twists with cross-location transfer learning, machine learning to identify the unknown phone location, and trying to separate the persons with clustering.   

Short bio:  Mitja Lustrek received his PhD degree from the Faculty of Computer and Information Science of the University of Ljubljana in 2007. He was a postdoc at the Institute for Biostatistics and Informatics in Medicine and Ageing Research in Rostock, Germany in 2010. He has worked at the Department of Intelligent Systems at Jozef Stefan Institute, Ljubljana, Slovenia ever since. He is currently employed there as a senior research associate and the head of the Ambient Intelligence Group. His main research interest is the analysis of sensor and other data related to human health and behavior using machine learning. He has been the principal investigator in a number of international research projects on this topic. He was a member of the teams scoring highly in several computer-science competition, such as the XPrize Pandemic Response Challenge and Tricorder competition, EvAAL competition and Sussex-Huawei Locomotion Challenge 2018-2020. He also served as the chair of the Slovenian Artificial Intelligence Society.

Speaker: Diogo Branco 

Title: DataPark: Reflections from a Longitudinal Deployment of a Digital Platform for PD Monitoring   

Abstract: Designing tools that are meaningful for healthcare environments can be difficult. There is the need to take into consideration the idiosyncrasies of dealing directly with clinicians, and patients and their families. This talk will focus on the importance of doing embedded research together with the ones that will use the software designed. For that, the talk will first introduce Datapark, a web platform for continuous monitoring of Parkinson's Disease. Additionally, the motivations and steps for designing the platform will be explained. Then, the most relevant components will also be highlighted. After that, the talk will focus on the challenges, barriers, and learnings of designing a platform and maintaining long-term collaboration with clinicians.   

Short bio:  Diogo Branco is a PhD student in Computer Science at Faculdade de Ciências da Universidade de Lisboa. His research focused on Human-Computer Interaction, particularly in health. For the last four years, he has been designing, developing, and evaluating applications, and platforms for different healthcare domains (e.g. people with Parkinson’s Disease, parents of young children with food disorders). Diogo is currently collaborating on several projects, such as IDEA-FAST (IMI) and FoodParenting.

Speaker: Hristijan Gjoreski, UKIM

Title: Wearable Computing and its Machine-Learning Applications

Abstract:

Hristijan Gjoreski is Assistant Professor at the Ss. Cyril and Methodius University in Skopje, Macedonia. He finished his PhD at the Jozef Stefan Institute in Slovenia, and was postdoctoral researcher at University of Sussex, UK.  His research experience is in the domains of applied Artificial Intelligence and Machine Learning. He has specialized in development of machine-learning algorithms in the areas of e-health, wearable computing, activity recognition and affective computing.  He has participated more than 12 international projects, and currently is a coordinator of the European Horizon 2020 Twinning project - WideHealth. He has 3 international patent applications, has more than 100 scientific publications, and 1700 citations with h-index of 23.  He has received award "Best Young Scientist" for 2016 from the President of Republic of Macedonia. He established and is organizer of the Data Science Macedonia group, with more than 1000 members. He has won 3 international machine learning competitions for human activity recognition with wearable sensors, which experience will be presented during this talk.

Speaker: Venet Osmani, Fondazione Bruno Kessler

Title: Predicting deterioration of critically ill patients

Abstract:

Venet Osmani, PhD is a senior researcher at Fondazione Bruno Kessler Research Institute. Previously, he was a lecturer at the department of Psychology and Cognitive Science at University of Trento, Italy and a visiting researcher at Georgia Institute of Technology, USA. His earlier research focused primarily on monitoring and analysing human behaviour. Specifically, using personal and environmental sensing applied to healthcare, including predicting depressive and manic episodes of bipolar patients and detecting occupational stress from smartphone sensors. Currently, the focus of his research is on analysis of clinical data (EHR) using machine learning methods to model disease and patient trajectories both for chronic conditions as well as in critical care. In this work he collaborates with some of the leading healthcare institutions in the US, including Cleveland Clinic, Mayo Clinic, MIT, as well as several leading European research institutions. He is an Expert Evaluator for European Commission (Horizon 2020 Programme), UK Medical Research Council (MRC), Swiss National Science Foundation (SNSF) and several other scientific funding institutions.

Speaker: Orhan Konak, Hasso-Plattner Institute

Title:  IMU-Based Trajectory Image Classification for Human Activity Recognition

Abstract: Recent trends in ubiquitous computing have led to a proliferation of studies that focus on human activity recognition (HAR) utilizing inertial sensor data. However, the performances of such approaches are limited by the amount of annotated training data, especially in fields where annotating data is highly time-consuming and requires specialized professionals, such as in healthcare. In image classification, this limitation has been mitigated by powerful oversampling techniques such as data augmentation. In this talk, we will evaluate how transforming inertial sensor data into movement trajectories and further 2D heatmap images can be advantageous for HAR when data are scarce. We will briefly discuss how a performance advantage can be achieved for small datasets, which is usually the case in healthcare. Moreover, movement trajectories provide a visual representation of human activities, which can help researchers to better interpret and analyze motion patterns.

Short bio: Orhan Konak graduated in Computational Engineering – Mathematics at the University of Applied Science Berlin in 2010. After working as a software engineer and forecast manager for eight years, he joined HPI in 2018 as a research assistant/PhD student. His research focuses on human activity recognition, through which classification of activities contributes to lower the documentation time for nurses. He is also very passionate about football.