CNRS Laboratoire des Signaux et Systèmes,
CentraleSupélec,
Université Paris-Saclay,
France
Dr. Giuseppe Valenzise is a CNRS researcher at the Laboratoire des Signaux et Systèmes, CentraleSupélec, Université Paris-Saclay, France, where he leads the Multimedia and Networking team. He is Editor-in-Chief of the Springer EURASIP Journal on Image and Video Processing. He received his PhD from Politecnico di Milano, Italy. His research covers image and video processing, with a focus on compression (traditional and learning-based), light field and point cloud coding, quality assessment, high dynamic range imaging, and machine learning for visual analysis. He has co-authored over 100 publications in these areas and received the EURASIP Early Career Award in 2018. Giuseppe was General Chair of ICME 2025 and regularly serves on organizing committees of flagship conferences such as ICIP. He has been Associate Editor for IEEE TCSVT, IEEE TIP, and Signal Processing: Image Communication. He is currently Chair of the IEEE SPS Multimedia Signal Processing Technical Committee and a member of the IEEE CAS Multimedia Systems and Applications Technical Committee.
AI-based compression of visual content: current trends, challenges and opportunities
In recent years, deep representation learning and generative models have led to significant advances in visual information compression, from traditional 2D images and videos to emerging immersive formats such as light fields, point clouds, and 3D Gaussian splats. In this talk, I will present an overview of some recent approaches for coding both 2D and 3D content. In particular, I will review learning-based schemes for 3D point cloud compression and discuss generative face video coding (GFVC) for ultra-low-bitrate video conferencing. I will conclude by outlining open problems and challenges in the field, with the goal of stimulating discussion.
Polytechnic University of Coimbra,
Portuguese Oncology Institute of Porto,
Portugal
Inês Domingues holds a PhD in Electrical and Computer Engineering (FEUP, 2015), a MSc in Electrical and Telecommunications Engineering (UA, 2008), and a BSc in Applied Mathematics (FCUP, 2004). She is currently an Assistant Professor at the Polytechnic University of Coimbra and a researcher at the Portuguese Oncology Institute of Porto (CI-IPOP) and at the RISE – Health Research Network.
Her research focuses on artificial intelligence for medical image analysis, with contributions in deep learning, radiomics, multimodal data fusion, and explainable AI, particularly in oncology. She is the Principal Investigator of the TIPTOP project and has been involved in several national and international research projects, including ESTIMA. Inês has authored over 100 scientific publications and serves on the editorial boards of journals such as IEEE Access, Artificial Intelligence in Oncology, and BMC AI. She has received multiple awards, including four IEEE Access Outstanding Associate Editor recognitions.
Inês is actively involved in the scientific community as a reviewer, editor, and member of organizing and program committees for conferences such as MICCAI, ISBI, and IbPRIA. She is Past-President of the Portuguese Pattern Recognition Association (APRP) and a member of the IAPR Governing Board.
AI for Oncology: Challenges, Pitfalls, and Promises
The application of Artificial Intelligence (AI) to Medicine has demonstrated considerable potential, while also presenting significant challenges in clinical translation. This talk will address common issues associated with medical data—such as scarcity, multimodality, imbalance, and model opacity—and outlines practical mitigation strategies, including data fusion techniques, synthetic data generation using diffusion models, and explainable AI. The discussion is illustrated through the clinical case of the ESTIMA project, which focuses on treatment response in patients with esophageal cancer.
University of Porto,
INESC TEC,
Portugal
Francesco Renna received the Laurea Specialistica degree in telecommunication engineering and the Ph.D. degree in information engineering, both from the University of Padova, Padova, Italy, in 2006 and 2011, respectively. Between 2007 and 2019, he held Visiting Researcher and Postdoctoral appointments with Infineon Technology AG, Princeton University, Georgia Institute of Technology (Lorraine Campus), Supelec, University of Porto, Duke University, University College London, and University of Cambridge. From 2019 to 2022, he has been an Assistant Researcher at Instituto de Telecomunicações, Porto. Since 2022, he is a researcher with INESC TEC. Since 2023, he has been an Assistant Professor at the University of Porto, Portugal. His research interests include high-dimensional information processing and biomedical signal and image processing, with focus on the use of artificial intelligence solutions for cardiovascular disease screening in underprivileged scenarios.
Dr. Renna was the recipient of a Marie Sklodowska-Curie Individual Fellowship from the European Commission and a Research Contract within the Scientific Employment Stimulus program from the Portuguese Foundation for Science and Technology.
Multimodal digital auscultation and AI for low-cost cardiovascular disease screening and monitoring
Several regions of developing countries face difficulties in diagnosing and treating both congenital and acquired heart conditions. This is mainly due to the lack of infrastructure and cardiology specialists in geographically large areas and difficulty in accessing health services. A non-invasive assessment of the mechanical function of the heart, performed at point-of-care settings, can provide early information regarding congenital and acquired heart diseases. In particular, cardiac auscultation and the analysis of the phonocardiogram (PCG) can unveil fundamental clinical information regarding heart malfunctioning by detecting abnormal sound waves in the PCG signal. In this talk, we will explore recent advances in the analysis of cardiac auscultation data. In particular, we will review the transformative impact carried out by deep learning architectures in extracting key clinical findings from heart sounds. Recent new directions in multimodal methods to assess the electromechanical activity of the heart will be also presented.