{"id":8130,"date":"2026-05-11T12:08:43","date_gmt":"2026-05-11T06:38:43","guid":{"rendered":"https:\/\/www.aimlsystems.org\/2026\/?page_id=8130"},"modified":"2026-09-24T10:18:04","modified_gmt":"2026-09-24T04:48:04","slug":"workshop-bi-and-cv","status":"publish","type":"page","link":"https:\/\/www.aimlsystems.org\/2026\/workshop-bi-and-cv\/","title":{"rendered":"Workshop – Biomedical Imaging and Computer Vision"},"content":{"rendered":"

[et_pb_section fb_built=”1″ fullwidth=”on” _builder_version=”4.25.1″ _module_preset=”default” custom_margin=”-90px||||false|false” global_colors_info=”{}”][et_pb_fullwidth_code _builder_version=”4.25.1″ _module_preset=”default” hover_enabled=”0″ global_colors_info=”{}” sticky_enabled=”0″]<\/p>\n

<\/p>\n
<\/div>\n

<\/p>\n

<\/div>\n

<\/p>\n

<\/p>\n
<\/p>\n
HALF DAY WORKSHOP<\/span><\/i> INVITED TALKS<\/span><\/div>\n

<\/p>\n

Workshop – BICV<\/h1>\n

<\/p>\n

Biomedical Imaging and Computer Vision<\/h2>\n

Connecting advances in computer vision with real clinical needs<\/pee><\/p>\n

<\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
<\/p>\n
<\/p>\n

<\/i> Workshop Description<\/h3>\n

<\/p>\n

The value of biomedical imaging as a rich, multi-scale source of human health biomarkers is increasingly recognised, and it can be complemented by other medical data, such as clinical records and physiological signals. These heterogeneous data sources provide complementary views of health and are pivotal across different domains, ranging from the analysis of cellular microarchitecture for cancer detection to organ-level imaging and language-based clinical assessments. Recent advances in representation learning have allowed for significant progress in extracting meaningful representations from such complex data, supporting applications ranging from early disease detection, to personalized treatment planning. However, several fundamental challenges remain that limit clinical translation. In particular, the sheer volume and heterogeneity of biomedical data make analysis inherently burdensome and non-trivial, especially when multiple modalities must be jointly considered. Furthermore, additional challenges such as limited annotations and the need for interpretability remain critical, particularly in high-stakes applications such as clinical decision making.<\/pee>Importantly, challenges associated with this data do not arise at the level of data analysis only, as they are closely linked to the data acquisition process itself as well. The quality and informativeness of biomedical data are indeed strongly related to how they are acquired. For this reason, a complementary line of research in biomedical imaging focuses on developing novel physical principles and acquisition paradigms to improve data quality at source. The aim is to enable more efficient and cost-effective imaging techniques that enhance contrast mechanisms and facilitate the discovery of novel biomarkers, ultimately easing downstream learning and interpretation. In light of the aforementioned challenges, this workshop will focus on novel approaches to biomedical applications, highlighting how challenges in data acquisition and downstream analysis can be addressed in different settings. Specifically, the workshop aims to encourage discussion on recent developments and solutions in medical imaging across modalities and application domains, with an emphasis on how representation learning and on acquisition strategies can be used and tailored to better support downstream analysis.<\/pee><\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n

<\/i> Official Schedule<\/h3>\n

<\/p>\n

<\/p>\n
<\/p>\n
Time<\/div>\n

<\/p>\n

Event<\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
13:30 \u2013 13:35<\/div>\n

<\/p>\n

Intro<\/strong><\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
13:35 \u2013 14:10<\/div>\n

<\/p>\n

Keynote: Chiara Coletti<\/strong> University of Oxford<\/span> “Can We Image How the Brain Learns? From MRI Physics to Non-invasive Mapping of Human Neurochemistry”<\/span><\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
14:10 \u2013 14:35<\/div>\n

<\/p>\n

Speaker 1: Gianluca De Franceschi<\/strong> Politecnico di Milano<\/span> “Learning and interpreting brain representations with foundation models”<\/span><\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
14:35 \u2013 15:00<\/div>\n

<\/p>\n

Speaker 2: Igor Balaz<\/strong> University of Novi Sad<\/span> “Hierarchical Constraint Inference for Minimal-Data Visual Representation Learning”<\/span><\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
15:00 \u2013 15:30<\/div>\n

<\/p>\n

<\/i> Coffee Break<\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
15:30 \u2013 15:35<\/div>\n

<\/p>\n

Intro e ripresa<\/strong><\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
15:35 \u2013 16:00<\/div>\n

<\/p>\n

Speaker 3: Eleonora Maggioni<\/strong> Politecnico di Milano<\/span> “AI applications to multimodal neuroimaging data for personalised psychiatry”<\/span><\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
16:00 \u2013 16:25<\/div>\n

<\/p>\n

Speaker 4: Federica Buccino<\/strong> Politecnico di Milano<\/span> “From pixels to biomarkers: can AI transform high-resolution biomedical imaging?”<\/span><\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
16:25 \u2013 16:50<\/div>\n

<\/p>\n

Speaker 5: Lara Cavinato<\/strong> Politecnico di Milano<\/span> “Artificial Intelligence in Nuclear Medicine: From Image Reconstruction to Personalized Theranostics”<\/span><\/div>\n

<\/div>\n

<\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n

<\/i> Speakers & Abstracts<\/h3>\n

<\/p>\n

<\/p>\n
<\/p>\n
<\/i><\/div>\n

<\/p>\n

<\/p>\n

Keynote: Chiara Coletti<\/h4>\n

University of Oxford<\/span> <\/p>\n

“Can We Image How the Brain Learns? From MRI Physics to Non-invasive Mapping of Human Neurochemistry”<\/h5>\n

<\/div>\n

<\/div>\n

<\/p>\n

Magnetic resonance imaging allows us to look inside the living human body without ionizing radiation, but conventional MRI primarily reveals anatomy and function rather than the underlying chemistry. Magnetic resonance spectroscopy (MRS) goes one step further, allowing us to non-invasive detect molecules involved in brain metabolism and neurotransmission. However, extending MRS to image neurochemistry across entire brain circuits remains technically challenging, particularly at ultra-high magnetic fields.<\/pee> In this talk, I will introduce the basic physical principles behind MRI and magnetic resonance spectroscopy (MRS), from nuclear spins and radiofrequency excitation to spatial encoding and chemical shift. I will then show how these principles lead to some of the major engineering challenges in ultra-high-field spectroscopic imaging. Using my current research at the University of Oxford as a case study, I will discuss the development of new radiofrequency and parallel-transmission methods for 7 Tesla MRSI, with the goal of simultaneously measuring neurochemical changes across interconnected brain regions during motor learning.<\/pee> Finally, I will briefly discuss where artificial intelligence and data-driven methods may complement this type of physics-based research, from synthetic data and analysis to pulse-design optimization. More broadly, I will argue that computational tools are most powerful when they are applied to a well-defined scientific problem grounded in an understanding of the underlying physics.<\/pee> <\/p>\n
Bio:<\/strong> Chiara Coletti is a postdoctoral researcher in MR spectroscopy at the Oxford Centre for Integrative Neuroimaging, University of Oxford, and a Research Associate at Jesus College. She holds a PhD in Imaging Physics from TU Delft, where her research focused on advanced radiofrequency pulse design for robust quantitative MRI, particularly for cardiac applications. Her work has led to multiple publications, international conference awards, invited lectures, and methods now being explored for clinical translation. She is currently extending her expertise to ultra-high-field brain MR spectroscopy and spectroscopic imaging. She was recently awarded a prestigious Marie Sk\u0142odowska-Curie Postdoctoral Fellowship to develop next-generation methods for imaging neurochemical dynamics across brain circuits at 7T MRI. <\/div>\n

<\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
<\/p>\n

Gianluca De Franceschi Politecnico di Milano<\/span><\/h4>\n

<\/p>\n

“Learning and interpreting brain representations with foundation models”<\/h5>\n

<\/div>\n

<\/p>\n

Foundation models are increasingly being applied to neuroimaging, offering rich representations that can transfer across tasks and datasets. However, models developed for natural images or language do not directly account for the three-dimensional organization and specific semantics of brain imaging. Understanding how these representations can be adapted to neuroimaging, and what information emerges in their embedding spaces, is therefore an important challenge.<\/pee> This talk will present recent work exploring these questions from two complementary perspectives. First, we investigate how a frozen 2D vision foundation model, DINOv3, can be adapted to 3D structural MRI through lightweight spatial learning. Weak positional encoding selectively improves the representation of subcortical anatomy and the anatomical organization of the embedding space, while stronger spatial constraints can disrupt the information inherited from the pretrained model. In parallel, we explore how representations extracted from brain MRI can interact with language models. Concept-based analyses of visual tokens reveal emerging alignment between imaging representations and task-relevant semantic concepts, providing a way to inspect how information is reorganized in a shared visual\u2013language space. Together, these studies illustrate how probing anatomical and semantic organization can help move beyond predictive performance toward more interpretable foundation models for neuroimaging.<\/pee> <\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
<\/p>\n

Igor Balaz University of Novi Sad<\/span><\/h4>\n

<\/p>\n

“Hierarchical Constraint Inference for Minimal-Data Visual Representation Learning”<\/h5>\n

<\/div>\n

<\/p>\n

Visual representation learning in biomedical imaging faces two compounding challenges that scale-based approaches address only partially: annotation scarcity and interpretability demands. Current methods, including self-supervised and contrastive learning, significantly reduce annotation requirements but still rely on large unlabeled corpora to converge on generalizable feature spaces, and produce representations that remain structurally opaque by design.<\/pee> We propose that these limitations reflect an architectural choice of representing visual knowledge as statistical regularities over continuous parameter spaces rather than as explicit structural descriptions. As an alternative, we present a hierarchical constraint inference framework in which visual representations are constructed as symbolic automata encoding primitive structural patterns, and higher-order automata (HOA) encoding their compositional relationships. In this approach, rather than learning from data distributions, the system infers constraints directly from pattern structure, constructing complete representations from minimal examples without a training phase. The framework produces fully interpretable, hierarchically organized representations stored as inspectable graph structures. Primitive patterns are encoded as elementary automata capturing spatial and relational constraints. Composite patterns are encoded as HOAs that explicitly reference their component automata, creating a multi-level hierarchy that reflects the natural compositional organization of visual structure. Similarity detection clusters structurally related automata, enabling generalization across variations.<\/pee> We present proof-of-concept results demonstrating complete hierarchical representation formation from minimal examples, with quantifiable graph metrics tracking structural generalization across multiple levels of abstraction. We discuss how the core principle of constraint inferences over compositional hierarchies translates to pixel-domain biomedical imagery, where annotation is scarce, structural regularity is strong, and interpretability is clinically required, and identify open research questions at the intersection of symbolic structural learning and biomedical computer vision.<\/pee> <\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
<\/p>\n

Eleonora Maggioni Politecnico di Milano<\/span><\/h4>\n

<\/p>\n

“AI applications to multimodal neuroimaging data for personalised psychiatry”<\/h5>\n

<\/div>\n

<\/p>\n

Psychiatric disorders are highly complex and heterogeneous conditions that likely originate from cumulative, small-scale changes across genetic and physiological systems. The significant variability in biological and clinical profiles between individuals has complicated efforts to develop reliable disease biomarkers, suggesting the need to transition from rigid diagnostic categories to personalised clinical profiling.<\/pee> Recent advances in open science, neuroimaging and AI offer a unique opportunity to implement this paradigm shift. One promising approach is normative modelling, which characterises the range of relevant variation in a reference population and quantifies deviations from this normative range, offering a personalised index of atypicality.<\/pee> This lecture will discuss the potential and risks of using AI to study and manage complex psychiatric disorders, focusing particularly on normative modelling and anomaly detection frameworks as powerful means of capturing patients’ unique neurobiological profiles. We will present recent studies that integrate multi-site clinical and neuroimaging data to advance the search for biomarkers of complex diseases such as major depressive disorder and bipolar disorder. This will promote the refinement of AI-powered diagnostics and prognostics in psychiatry.<\/pee> <\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
<\/p>\n

Federica Buccino Politecnico di Milano<\/span><\/h4>\n

<\/p>\n

“From pixels to biomarkers: can AI transform high-resolution biomedical imaging?”<\/h5>\n

<\/div>\n

<\/p>\n

Recent advances in high-resolution imaging and artificial intelligence are transforming the way biomedical tissues can be characterized across multiple spatial scales. Here, we present an integrated framework combining synchrotron phase-contrast micro-computed tomography with AI-driven image analysis to investigate tissue microarchitecture with unprecedented detail. Using human trabecular bone as a case study, we demonstrate how convolutional neural networks and large-scale morphometric analysis enable the automated quantification of millions of osteocyte lacunae from sub-micron-resolution datasets. The approach reveals disease-associated microstructural remodeling signatures in osteoporosis and post-COVID-19 bone that are not captured by conventional densitometric measures alone. Beyond bone research, the presented workflow highlights the broader potential of combining high-resolution imaging and AI for biomarker discovery, mechanobiology, and precision medicine applications in biomedical engineering.<\/pee> <\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
<\/p>\n

Lara Cavinato Politecnico di Milano<\/span><\/h4>\n

<\/p>\n

“Artificial Intelligence in Nuclear Medicine: From Image Reconstruction to Personalized Theranostics”<\/h5>\n

<\/div>\n

<\/p>\n

Artificial Intelligence is rapidly transforming nuclear medicine across the entire clinical workflow. This talk provides an overview of current applications and emerging directions, from image acquisition and reconstruction, including dose reduction, denoising, faster scans, and CT-free PET, to image processing and quantitative analysis, with applications in segmentation, harmonization, and synthetic image generation. We will also discuss the role of AI in diagnosis and clinical decision support, as well as its growing potential in theranostics and personalized medicine, including patient-specific dosimetry, treatment planning, and digital twins. Through representative examples, the talk will highlight both the opportunities and the challenges of translating AI into clinical practice, outlining the transition toward more integrated, quantitative, and personalized nuclear medicine.<\/pee> <\/div>\n

<\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n

<\/i> Workshop Organizers<\/h3>\n

<\/p>\n

<\/p>\n
<\/p>\n
<\/p>\n
<\/i><\/div>\n

<\/p>\n

<\/p>\n

Virginia Tasso<\/h4>\n

Politecnico di Milano<\/pee><\/i> virginia.tasso@polimi.it<\/a><\/div>\n

<\/div>\n

<\/p>\n

Virginia Tasso received her B.Sc. and M.Sc. degrees in Biomedical Engineering from Politecnico di Milano in 2023 and 2025, respectively. In 2025 she also earned her M.Sc. in Biomedical Engineering at University of Illinois Chicago. Currently, she is a Ph.D. student in Information Technology at NECSTLab Politecnico di Milano. Her research focuses mainly on medical image processing and analysis for computer-assisted diagnosis, with the goal of supporting clinicians in improving diagnostic workflows and enabling more efficient patient management.<\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
<\/p>\n
<\/i><\/div>\n

<\/p>\n

<\/p>\n

Eleonora D\u2019Arnese<\/h4>\n

University of Edinburgh<\/pee><\/i> eleonora.darnese@ed.ac.uk<\/a><\/div>\n

<\/div>\n

<\/p>\n

Eleonora D\u2019Arnese is currently Lecturer in Biomedical AI at the School of Informatics of the University of Edinburgh. She received her Ph.D. in Information Technology from Politecnico di Milano in 2023. She also obtained her B.Sc. and M.Sc. degrees in Biomedical Engineering from Politecnico di Milano in 2016 and 2018, respectively, and an M.Sc. degree in BioEngineering from University of Illinois Chicago in 2018. She previously worked as a Postdoctoral Researcher at Politecnico di Milano and at Universit\u00e0 Vita-Salute San Raffaele. Her research focuses on medical image analysis and computer vision, with particular interest in image registration, multi-modal image analysis, radiomics, and AI methods for biomedical applications.<\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
<\/p>\n
<\/i><\/div>\n

<\/p>\n

<\/p>\n

Marco Santambrogio<\/h4>\n

Politecnico di Milano<\/pee><\/i> marco.santambrogio@polimi.it<\/a><\/div>\n

<\/div>\n

<\/p>\n

Marco D. Santambrogio is Associate Professor at the Department of Electronics, Information and Bioengineering (DEIB) of Politecnico di Milano and Adjunct Professor at the College of Engineering of University of Illinois Chicago. He received his M.Sc. degree in Computer Engineering and his Ph.D. in Information Technology from Politecnico di Milano in 2004 and 2008, respectively, and an M.Sc. degree in Computer Science from University of Illinois Chicago in 2004. He was a Research Affiliate at the MIT Computer Science and Artificial Intelligence Laboratory from 2010 to 2015 and founded both the DRESD research group and the NECSTLab at Politecnico di Milano. His research focuses on computer architectures, reconfigurable systems, hardware\/software co-design, embedded systems, and operating systems. He has participated in several national and European research projects, including Scalopes, hArtes, SAVE, FASTER, and EXTRA, serving in roles ranging from researcher and task leader to work-package leader and principal investigator. He is a Senior Member of IEEE and ACM.<\/div>\n

<\/div>\n

<\/div>\n

<\/div>\n

<\/div>\n

<\/p>\n

<\/p>\n
<\/p>\n

Important Dates<\/h3>\n

<\/p>\n