Medical Artificial Intelligence
The Medical Artificial Intelligence research group (MedAI) was established in 2018 with the primary objective of developing and applying Artificial Intelligence (AI) methods to support clinical decision-making across various medical disciplines.
Introduction
The Medical Artificial Intelligence research group (MedAI) was established in 2018 with the primary objective of developing and applying Artificial Intelligence (AI) methods to support clinical decision-making across various medical disciplines.
The group focuses on the analysis of medical images, physiological signals, and health data to extract clinically meaningful information for diagnosis, disease monitoring, and risk assessment. Core application domains include ophthalmology, radiology, and neurophysiology, with particular emphasis on challenging settings such as small datasets, heterogeneous acquisition conditions, and evolving clinical protocols. Typical use cases range from disease grading in inflammatory eye conditions to sleep analysis and seizure prediction from EEG, as well as computer-aided detection and risk modelling using multimodal health data.
Methodologically, MedAI develops machine learning and signal processing approaches tailored to clinical constraints. This includes deep learning and foundation models for multimodal data, combining images, time-series signals, and structured or longitudinal health data. A strong focus is placed on generalisation and robustness, enabling models to transfer across hospitals, devices, and populations. The group also investigates domain adaptation, efficient learning under limited data, and scalable pipelines for real-world deployment.
A central pillar of the group’s research is trustworthy AI in healthcare. This encompasses the evaluation and mitigation of demographic bias, interpretability of model decisions, and the systematic validation of models under clinically realistic conditions. The group contributes methods to assess trade-offs between fairness and performance, and develops tools to ensure that AI systems remain reliable and equitable when deployed in practice.
MedAI follows a reproducible and collaborative research methodology, combining open scientific software, rigorous experimental design, and close interaction with clinical partners. This enables the development of AI systems that are not only technically sound, but also clinically relevant, transparent, and ready for translation into healthcare environments.
Current group members
Alumni
Ongoing projects
Artificial Intelligence (AI) has become a powerful and pervasive technology in recent years, influencing numerous aspects of our daily lives. We encounter AI through recommendation algorithms in online stores, voice-activated smartphone assistants, or the widespread use of technologies like ChatGPT. However, its rapid growth and integration into society raise complex questions and concerns among the public. Public opinions on AI vary widely; while some people are enthusiastic about its potential to revolutionize industries and enable breakthroughs in fields such as medicine, others fear that AI could lead to undesirable outcomes, such as a loss of human control or privacy. These views are often shaped by media narratives and the competing interests of different stakeholders, which play a significant role in influencing both public opinion and policy decisions.
Our team of scientists and communication experts aims to enhance the understanding of AI technologies among the Swiss people, with a particular focus on teenagers and female students, to create a positive societal impact. Building on the foundations of our previous project, NewsOnAI, we will expand beyond traditional media such as newspapers and employ diverse methods, including artistic performances and interactive exhibitions. We will design these activities to be highly interactive, encouraging active participation and dialogue. Activities will include themed theater plays that explore AI’s impact on everyday life, exhibitions where participants can interact with AI tools, and workshops specifically designed for teenagers and female students to discuss AI’s future role in society. Feedback collected will include real-time audience reactions, structured questionnaires, and focus group discussions, which will be analyzed to continuously refine and adapt our engagement strategies.
Our primary audience includes Swiss citizens interested in cultural activities, particularly teenagers who are keen to follow new trends. Additionally, we are committed to addressing gender aspects by designing content and activities that specifically appeal to female students. We aim to inspire and empower young women to take on more prominent roles in shaping the digital world, acknowledging that they have historically been underrepresented in these fields. As societal attention shifts toward greater inclusion, our project will contribute to fostering a more balanced and equitable digital future.
While many individuals in our target groups may lack in-depth technical knowledge of AI, they often encounter new AI products, companies, and social issues through various media channels, including newspapers and science fiction movies. As a result, they may be aware of recent developments but also susceptible to misunderstandings and controversies related to technologies such as ChatGPT, Elon Musk's brain-chip startup, and other emerging AI applications. It is crucial to recognize that media portrayal significantly influences public opinion on AI, both positively and negatively. Media creators, even if they are not experts in AI, often produce content that captures public attention, which high-profile figures, including entrepreneurs, CEOs, and politicians, may leverage to advance their agendas. This can sometimes lead to skewed public perceptions, whether intentionally or unintentionally. Given this landscape, it is essential for AI scientists to collaborate with media creators, providing evidence-based insights to ensure accurate and balanced information is shared with the public. Our project fosters such collaboration, ensuring that both the potential and limitations of AI are clearly communicated. By sharing our findings through diverse media outlets, we aim to reach a broad audience, extending beyond Switzerland. Furthermore, our proactive engagement efforts will foster dynamic, two-way communication between scientists and the public, using interactive methods in exhibitions and theater plays to engage teenagers and female students specifically. Analyzing the feedback from these initiatives will provide invaluable insights into public perspectives on emerging technologies. This understanding will guide scientists in pursuing research directions that effectively address societal concerns, demonstrating the tangible benefits of our project for both scientific advancement and societal well-being. We anticipate that our efforts will have a multiplying social impact over time, promoting informed public discourse and a deeper understanding of AI technologies.
Project DREAM aims to develop a scalable foundation model for
multi-channel EEG signal analysis, spanning from basic polysomnography
to high-density and intracranial recordings. The goal is to create a
unified AI model capable of adapting to various EEG configurations while
maintaining simplicity, interpretability, and computational efficiency.
By leveraging convolutional neural networks and attention mechanisms,
the model will extract robust features for diverse tasks, including
sleep stage classification, epilepsy detection, and cognitive function
assessment. Collaborations with leading hospitals will provide
real-world datasets for validation, ensuring the model’s generalization.
Ultimately, this project seeks to advance AI-driven biomedical signal
processing, enhancing diagnostic and monitoring capabilities in
neurology and sleep medicine.
The algorithmic bias remains one of the key challenges for the wider applicability of Machine Learning (ML) in healthcare. Statistical modeling of natural phenomena has gained traction due to increased representation capacity and data availability. In medicine, particularly, the use of ML models has increased significantly in recent years, especially to support large scale screening, and diagnosis. However impactful, the study of demographic bias of newly developed or already deployed ML solutions in this domain remains largely unaddressed. This is particularly true in the medical imaging domain, where it remains challenging to associate demographic attributes with features. Out of the most recent results, the "impossibility of fairness" establishes some criteria for demographic impartiality cannot be reached simultaneously. Among other factors, the lack of raw data, in particular for intersections of minorities, is one of the greatest issues that remain unaddressed due to their challenging nature. This proposal addresses three important challenges in the domain of ML fairness for medical imaging: (i) Create novel ways to train ML models for medical imaging tasks, that can be automatically adjusted to become more useful (maximize performance), group or individually fair, (ii) Quantify fairness boundaries of ML models and associated development data, and finally, (iii) Build systems whose joint performance with humans in the decision loop is fair towards various individuals and demographic groups. To achieve these goals, we will develop a novel evaluation framework and loss functions that take into account model utility together with all aspects of demographic fairness one may wish to address. A generative framework, trained to isolate tunable demographic features, will provide large-scale data simulation covering minorities and intersections. We will then study fairness (safety) boundaries through a modified learning curve setup, analyzing and quantifying limits in both ML models and training data. Finally, we will study how humans-in-the-decision-loop affect the fairness of hybrid human-AI systems, and address post-deployment utility/fairness tuning by embedding weight coefficients directly into the trained model. The development of methods and tools to detect, mitigate, or remove bias will improve the safety of ML models deployed in healthcare. We expect our work will help define new operational boundaries for the responsible deployment of artificial intelligence tools.
Past projects
Fluorescein angiography (FA) is a unique tool for the analysis of the retinal vasculature in both its morphology and its function: it is the only clinical method that allows to evaluate the function and integrity of the blood-retinal barrier, providing means for the detection and grading of inflammatory diseases affecting blood vessels in the eye (Vasculitis) [1]. FA is an invasive technique involving a non-negligible level of risk for the patient. The interpretation of an angiography is intrinsically challenging, requiring years of clinical experience. The overarching objective of this proposal is to develop and (clinically) validate a system to automatically detect and grade inflammatory eye diseases, with minimal risk, allowing for better patient management and care. This proposal also represents the first attempt to automatically grade angiography images and holds the potential to help doctors in the challenging interpretation of these data, delivering cutting edge machine learning solutions that would support clinicians at our hospitals, and benefit the scientific community at large. To reduce risk for the patient, we will search for novel biomarkers from other ophthalmic imaging modalities, such as fundus images, that would be less invasive, cheaper and safer to acquire, but correlate well with inflammatory diseases. In this pilot project, we propose to establish data, annotations, and develop a prototype grading system (machine learning model), to automatically evaluate inflammatory signs from FA images. A preliminary study on novel biomarkers from alternate modalities will complement this phase of our project. We intend to disseminate our work via scientific articles to be submitted to medical and computational journals. If this pilot is successful, a follow-up project will be submitted.
Collaborative breakthrough research in Artificial Intelligence (AI) requires access to well-tuned systems, specific computing, large datasets, and dedicated storage, often only accessible to a select few. This reality slows down complementary research, as a substantial amount of project time is often dedicated to repeating setups, understanding computation and storage intricacies, data properties, and how to properly transfer systems between institutions, to pursue overall objectives. Examples can be often found in cases encompassing consortia with computational versus non-computational specialists, or containing a virtuous mixture of domain experts (medical doctors, biologists, psychologists) and data scientists. Reproducibility and technology transfer are essential tools in thriving projects, however these concepts are costly to implement and maintain. Deploying AI solutions requires multidisciplinary expertise, the right hardware, and AI specialists to tune and ready tools for collaborative use. In practice, lack of specific expertise slowdowns partner-to-partner communication affecting overall productivity. In projects with industry or government-academia partnerships driven by concrete societal needs, replicating project conclusions with different, private datasets, or allowing partners to infer from pre-trained models using adequate hardware setups, is often avoided because of these barriers. Moreover, the growing scale, complexity and impact of contemporary AI systems such as Large Language Models (LLMs) accelerates the need for accessible infrastructures which can guarantee systematised, transparent and increased collaborative work. We intend to bridge these gaps by building “CollabCloud", a cloud-based research infrastructure to boost collaborative research for current and future projects at the Idiap Research Institute. The main focal points will be boosting the ability of easily exporting researchworkflows, exploring AI models by both computational and non-computational experts, and allowing controlled access to shared storage and computing power for collaborative projects. Beyond these goals, CollabCloud will enable Idiap to participate in developing important topics shaping the future of AI, such as Federated Learning, and cloud-based scientfic networks, which require connectivity and storage capabilities adapted to such purposes. As discussed in the institutional support letter, this vision aligns well with Idiap’s future, its predicted growth (with a current data center reaching the limits of its maximum capacity), and the notion of Cross Research Groups (CRG), that is part of our 2021-2024 Research Program as approved by the Federal State Secretariat for Education, Research, and Innovation (SERI).
The overarching goal of this project is the development of specific and reliable biomarkers for improved patient stratification and tumor classification, and the understanding of the mechanisms responsible for therapy resistance, in particular for drugs in development in Novartis through the integrative analysis of clinical, images of hematoxilin and eosin stained tissue sections, and multiomics data (bulk RNA-sequencing, 10X Genomics single-cell sequencing, Smart-seq 3 full-length single-cell sequencing data, spatial transcriptomics).
We are a group of scientists who joined forces to enable artificial intelligence (AI) technologies to make a greater impact on Swiss society by participating in a public debate over the latest news. We plan to publish research-based opinion articles on mainstream media in Switzerland, collect the readers’ feedback on the published articles, and follow up by organizing live seminars for the audience in the area. We will also post rounded responses on social media to ensure bi-directional dialogue. Our primary target audience is Swiss citizens reading a newspaper. They are mostly non-scientific public and may not have specific knowledge of AI technologies. We often receive news about new products, companies, and social issues related to AI technologies. An example is a controversy in the U.K. over an algorithm used to substitute university-entrance exams in 2020. Another example is an accident in testing an autonomous driving car. Usually, the public’s perception of information is affected by the tones and perspectives taken by the writers. News reporters are not necessarily technical experts and are often incentivized to publish striking news stories to attract more audiences. As some entrepreneurs, celebrity CEOs, and politicians may try to magnify the public hype for their benefit, the conflict of interests can influence media perspectives. Therefore, there is a need for AI scientists to collaborate with journalists to provide evidence-based analysis in public debate whenever appropriate. Without such cooperation, the general audience would miss out on essential information regarding AI. Scientists can help the public better understand AI technological issues by clarifying the potential capabilities and limitations. Not many people would read scientists’ research published in academic journals. However, publishing news articles has a greater impact, as the article is distributed to a much broader audience in Switzerland and worldwide through popular media platforms. In addition, follow-up activities at a local level will ensure the interactive and bi-directional dialogue between scientists and the readers. The analysis of the feedback will allow the project team to understand the public’s perspectives in facing future technologies. In turn, the acquired understanding can guide scientists to pursue an impactful research direction. After this Agora project, the current members will be able to form a new team to engage other scientists to study societal trends, raise their voices, and provide research-based analysis to guide the public about future technologies. The social impact will be multiplied over time.