The third (in-person and online) edition of Data4SmartHealth aims to promote knowledge and awareness of artificial intelligence-based methodologies and techniques for analyzing clinical data and building health informatics applications. The workshop is dedicated to physicians and other healthcare professionals, staff of ICT companies working in the field of medical informatics, scholars who wish to deepen their knowledge on the topic, and any other interested people. Attendees of the event will receive an overview of some of the key issues and most promising techniques in the field through four tutorials given by experts in e-health, artificial intelligence and data analytics. In continuity with previous editions, the goal of the workshop is also to consolidate the development of a community, centered on the NOI techpark in Bolzano, committed to the dissemination of the use of digital technologies in healthcare and the joint implementation of research and development projects.
An event organized by
in cooperation with
The video recording of the event is available here.
8:20 - 8:50 Registration
8:50 - 9:00 Welcome
9:00 - 9:50 Tutorial 1
9:55 - 10:45 Tutorial 2
10:45 - 11:15 Coffee break @Noisteria
11:15 -12:05 Tutorial 3
12:10 -13:00 Tutorial 4
13:00 - 13:10 Closing
Prof. Fabio Stella, Università degli Studi di Milano Bicocca
Towards smarter health care: can Artificial Intelligence help?
Artificial intelligence, machine learning and deep learning are introduced together with their capabilities and main limitations. The ladder of causation, i.e., the hierarchy of artificial intelligence tasks proposed by Judea Pearl is introduced and discussed. The three rungs of the ladder, namely, seeing, intervening and counterfactuals, are exploited to clarify which research questions can be answered depending on the type of the available data and on the available domain/expert knowledge. We show that prediction is not the same as decision making, and that knowing the story behind the data is fundamental to make effective and explainable decisions. This is achieved by presenting and discussing elementary examples where ignoring the story behind the data directly brings to non-sense conclusions. We conclude by illustrating some examples where causal Bayesian networks have been developed to help clinicians and physicians to reason under uncertainty and to make decisions.
Speaker's short bio: Fabio Stella is an associate professor at the University of Milan-Bicocca, Department of Informatics, Systems and Communication, where he leads the model and algorithms for data and text mining research lab. His main research interests are Bayesian networks and causal networks for health care, biology and finance. He has been the Principal Investigator of several research projects funded by private companies in the health care sector. He published more than 100 papers, and has been awarded the 10% best reviewers at NeurIPS in 2020 and the outstanding reviewer award at NeurIPS, AISTATS and ICML in 2022.
A bird-eye-view of process mining (with a focus on healthcare)
Process mining is a discipline at the intersection of data science, business process management and computing, whose ultimate goal is to support organizations in the continuous improvement of their processes based on actual facts rather than fiction. This is obtained through the analysis of data produced during the execution of such processes. Of particular relevance are event data, tracing which process instances have been executed and which relevant events occurred therein. Each event typically comes with key information pointing to reference process instance, the related activity in the process, the execution timestamp, together with additional relevant attributes (such as the resource who generated the event). Process mining relates such event data to process (and other organizational) models, making it possible to discover process models from data (thus describing what actually happened, and not what should happen), to check conformance of actual and expected behaviors, and to enhance models with extracted information, highlighting frequent paths, inefficiencies, and bottlenecks.
Process mining techniques are of particular relevance in highly complex, unstructured, and unpredictable organizational contexts, such as that of healthcare. The goal of this tutorial is to provide a bird-eye view of process mining and illustrate the main process mining functionalities and techniques. To do so, we will use an event log from containing events of sepsis cases from a hospital.
Speakers' short bio: Fabrizio Maria Maggi in an associate professor at the Faculty of Computer Science, Free University of Bozen-Bolzano. His areas of research span process modeling and process analysis. He has coauthored articles on process mining, automated revision of business process models, declarative business process modeling, monitoring of business constraints at runtime, service oriented architectures, service choreographies, and service composition.
Marco Montali is a full professor and vice-dean for studies at the Faculty of Computer Science, Free University of Bozen-Bolzano. He develops novel, foundational and applied techniques grounded in artificial intelligence, logics, and formal methods, to create intelligent agents and information systems that combine processes and data, with a particular interest in the combination of model-driven and data-driven techniques. He is co-author of more than 230 papers, many of which in top-tier conferences and journals, and recipient of 10 best paper awards and 2 test-of-time awards.
Dr. Paolo Avesani, Fondazione Bruno Kessler
AI in Medicine: a case study in clinical neuroscience
The recent innovation in brain neuroimaging is going to have a disruptive impact in the diagnosis and treatment of brain disorders. The novel techniques allow the characterization of functional and structural brain networks at the individual level. The common neurosurgical practices are going to be revised accordingly. This new scenario poses new challenges both for clinicians and artificial intelligence. While a data driven approach to neuroimaging is opening new opportunities, we have to not neglect potential pitfalls and computational biases. We present the experience of adoption of advanced computational methods in the Clinical Neuroscience Department at S. Chiara Hospital in Trentino.
Speaker's short bio: Paolo Avesani is the head of the Neuroinformatics Lab (NILab) raised as a joint initiative of the Center for Digital Health of Fondazione Bruno Kessler and the Center for Mind/Brain Sciences of the University of Trento. His research interests include the data-driven computational methods for the characterization of functional and structural brain connectivity.
Dr. Ivan Donadello, Free University of Bozen-Bolzano
Computational Persuasion for Behavior Change in Healthcare
Virtual coaches (VCs) are computer programs (mobile phone apps or chatbots) that simulate conversations with people by mimicking a human being. Their aim is to support and help users in performing a particular task. They have become an increasingly relevant resource in healthcare for the management of chronic diseases and for promoting behavioral changes in the self-management of individuals. For example, a growing number of studies report their effectiveness in reducing perceived stress and abnormal eating behavior, improving symptoms of anxiety, depression and in coping with diabetes mellitus. In this tutorial, we will present a promising field of Artificial Intelligence for developing VCs: the Computational Persuasion. This provides VCs of smart (and tailored to users) dialogue strategies to persuade users at adhering medical prescriptions and advice. We will see the main techniques underlying Computational Persuasion as well as some important results in Behavior Change for Healthcare.
Speaker's short bio: Ivan Donadello is a researcher at the Faculty of Computer Science, Free University of Bozen-Bolzano. His current research interest mainly focuses on virtual agents (or softbots) able to recognize the health state of a person and give suggestions to improve it. His expertise encompasses the fields of Knowledge Representation (ontologies and Fuzzy Logic), Machine/Deep Learning, Computer Vision, eHealth and Explainable Artificial Intelligence.
Floriano Zini is an academic technologist at Smart Data Factory, the technology transfer laboratory of the Faculty of Computer Science at the Free University of Bozen-Bolzano. Over the past 10 years, his research activity has focused on the design, development and evaluation of mobile and web-based information systems for personalized health care. Personalization is achieved through the application of Human Computer Interaction and Artificial Intelligence methodologies and algorithms, particularly Machine Learning. Floriano holds a Ph.D. in computer science and has authored about 60 scientific papers on e-health, personalized information systems, grid computing, multiagent systems and evolutionary computing. According to Google Scholar, his h-index is 21.