Banner
Home      Log In      Contacts      FAQs      INSTICC Portal
 

Keynote Lectures

Wirewalking over Two Medical AI Chasms: Results and Open Problems in Making "Valid AI" Also Useful in Medical Practice
Federico Cabitza, Università degli Studi di Milano-Bicocca, Italy

On Trust and Trustworthiness of Interpretable AI Methods for Decision Support in Radiology
Katja Bühler, VRVis, Austria

Value-based Healthcare: Contributions from Biomedical Engineering
Ana Rita Londral, Universidade NOVA de Lisboa, Portugal



 

Wirewalking over Two Medical AI Chasms: Results and Open Problems in Making "Valid AI" Also Useful in Medical Practice

Federico Cabitza
Università degli Studi di Milano-Bicocca
Italy
 

Brief Bio
Federico Cabitza, MEng, PhD, is an Associate Professor at the University of Milano-Bicocca (Milan, Italy) where he teaches Human-Computer Interaction and where he coordinates the research activities of the MUDI Lab (Modelling Uncertainty, Decisions and Interaction). Since 2016 he has also had a research appointment with the IRCCS Orthopaedics Institute Galeazzi in Milano (Italy) and more recently with San Raffaele Hospital. He is associate editor of the International Journal of Medical Informatics (IJMEDI, ISSN: 1386-5056) and stable member of the editorial boards of the MAKE - Machine Learning and Knowledge Extraction journal (ISSN: 2504-4990), the CRCL Cross-Disciplinary Research in Computational Law (ISSN 2736-4321), and the Journal of Medical Artificial Intelligence (ISSN 2617-2496). He has co-chaired International workshops (on Data Visualization in Healthcare and knowledge IT artifacts), conference tracks (on Socio-technical design), conference programs (for the Italian Chapter of AIS and BIOSTEC Healthinf in 2020) and special issues on impacted Journals (i.e., the CSCW journal, Program, and the Health Informatics Journal). He is the author of more than 140 research publications to date, in international conference proceedings, edited books and scientific journals, including the JAMA, the CSCW Journal, Computers in Biology and Medicine, Behaviour and Information Technology, International Journal of Human Computer Studies, Computers in Human Behavior, International Journal of Approximate Reasoning, and the Journal of Visual Languages and Computing. His current research interests regard the design and evaluation of interactive systems and decision support based on machine learning techniques in the Healthcare domain.


Abstract
Achieving a pragmatic, or even an ecological validation (Cabitza and Zeitoun, 2019) of medical AI systems that nevertheless exhibit very high (statistical) accuracy has been observed to be more complicated than initially expected (Coiera et al. 2018): in fact, most of the challenges that make technically sound systems perform poorly in real-world settings lie in the so called “last mile of implementation” (Coiera, 2019). This evocative concept expresses the semantic difference between developing medical machine learning (or medical AI) and the mere application of machine learning techniques to medical data. Moreover, we will make the point that the space bewtween machine learning development and clinical practice, is not a flat and regular path, but rather presents two chasms: the chasm of human trust, and the chasm of machine experience. The former one requires to focus on usability and explainability, while the latter ones requires data governance and to focus on data work, including practice of “data awareness” and “data hygiene”. I will discuss these notions, and report about some researches I personally conducted while trying to bridge the above chasms with mixed fortunes: what we recognize as still open problems are exciting opportunities to look at a seemingly established field from a fresh perspective (the interactionist perspective) and develop solutions that focus on the utility of the technology rather than following the mirage of accuracy.



 

 

On Trust and Trustworthiness of Interpretable AI Methods for Decision Support in Radiology

Katja Bühler
VRVis
Austria
 

Brief Bio
Katja Bühler is Scientific Director of the VRVis Research Center in Vienna, Austria. She completed her studies in Mathematics at KIT, Karlsruhe, Germany and holds a PhD in Computer Science from TU Wien, Austria. In 2003 she became head of the Biomedical Image Informatics Group at VRVis. She is member of the management board of Austrian Bioimaging and associate editor of Computers and Graphics, the Visual Computer and Frontiers in Bioinformatics. Katja’s scientific roots are in reliable computing, numerics and geometry processing. Today's focus of her research is on the development of highly efficient methods to provide access to information encoded in biomedical images with the aim to accelerate image-based decision making. For this purpose, she is fusing expertise in image processing, machine and deep learning, high performance computing, data mining, visualization and human computer interaction to novel visual computing solutions for medicine and life science. Research results of her group received numerous scientific awards and resulted in several patents. The software emerged from the groups research is helping radiologists, radiotherapists and surgeons to cope with multi-modal images and diagnostic tasks in their daily clinical routine. The e-science platform Brain* addresses the urgent need to manage and exploit image intense data collections and accelerates multi-omics and image-based research of neuroscientists.


Abstract
AI is widely recognized as disruptive technology and rising expectations to have a high impact on the digital transformation of the healthcare sector within the next years. This not only includes novel and powerful opportunities for optimization and automation of administrative and routine tasks, but also the support of clinical activities, including high-risk applications like diagnostics and clinical-decision support. For the latter, technical trustworthiness of the system is a basic requirement for its application in clinical routine. However, the technical reliability might not necessarily be sufficient to generate trust in the system on the user´s side (as trust is a complex psychological phenomenon and its definition highly depends on tasks and environments). Taking image-based diagnostics in radiology as a rapidly evolving area of application of AI methods I will first reflect on the different aspects of trust in this context, and oppose it with recent developments on strategies to increase explainability and interpretability of AI. Based on a selection of our recent results I will highlight potential pitfalls and the urgent need for domain awareness in the design of trustworthy interpretability methods and conclude with open challenges for future research.



 

 

Value-based Healthcare: Contributions from Biomedical Engineering

Ana Rita Londral
Universidade NOVA de Lisboa
Portugal
 

Brief Bio

Ana graduated in Electrical and Computer Engineering, in Instituto Superior Técnico, and holds a PhD in Biomedical Sciences(Neurosciences) from the Faculty of Medicine, University of Lisbon.
She has experience in closely working in R&D activities with both companies and academia in the development of digital health-related systems and devices. From her solid engineering background, the PhD research studies in a faculty of medicine gave her important skills to enable her to apply innovative technology to further enhance the healthcare digital transformation and promote health and social-related quality of life. She prioritizes collaborative and multidisciplinary networks, by validating tools and methods in real-world pilots/studies with the full engagement of practitioners, patients and, in a broader approach, citizens.
She lectured Assistive Technologies, Machine Learning and Programming courses, as invited professor in Universidade de Aveiro and Instituto Politécnico de Setúbal. Moved by her interest in assistive technologies, and their application to the context of neurodegenerative conditions, she developed research with international teams in the field of accessibility, HCI, assistive communication, speech and biosignals processing.

In 2018, she had co-leadership with NOVA University of Lisbon, in designing a collaborative laboratory that was approved and certified by the Portuguese National Funding Agency for Science, Research and Technology (FCT). She now is the executive and scientific director of the Value for Health CoLAB for developing interdisciplinary research in eHealth and Value-based healthcare, in close collaboration with NOVA, Fraunhofer Portugal, CUF and Vodafone Portugal. 


Abstract
Healthcare systems are struggling with sustainability and accessibility to populations, worldwide. As resources are finite, the concern with identifying treatments that bring the higher value to patients and to society is rising a global movement of Value-based Healthcare (VBHC). Measuring outcomes and costs to achieve those outcomes is a starting point to measure value in Health. Remote patient monitoring technologies and data science are very relevant to capture outcomes that are relevant to patients and to make healthcare more accessible and sustainable, as also more accountable.I will present the topic of VBHC and introduce to opportunities for biomedical engineers to contribute for the development of technologies of high-value in healthcare. Two case-studies will be presented related to the contexts of cardiac surgery and emergency hospital services.



 



 


 



 


footer