DTU Compute

𝐀 𝐯𝐢𝐬𝐢𝐭 𝐟𝐫𝐨𝐦 𝐚 𝐩𝐢𝐨𝐧𝐞𝐞𝐫 𝐢𝐧 𝐌𝐞𝐝𝐢𝐜𝐚𝐥 𝐈𝐦𝐚𝐠𝐞 𝐂𝐨𝐦𝐩𝐮𝐭𝐢𝐧𝐠 – Earlier in March, DTU Compute welcomed Professor 𝐒𝐭𝐞𝐩𝐡𝐞𝐧 𝐌. 𝐏𝐢𝐳𝐞𝐫 from the University of North Carolina – UNC Department of Computer Science. Pizer wrote what is considered the world’s first dissertation in medical image computing and has a long‑standing history of collaboration across computer science, mathematics, statistics, and healthcare. His reason for visiting us was to share the methods he has spent three-quarters of a lifetime developing and to meet researchers who can take them further. Across hospitals, medical image analysis is gradually gaining ground as a support tool for clinicians and radiologists in their work with X‑ray images, ultrasound, and MRI and CT scans. Startups within the field are emerging, and the research area has a strong presence at DTU Compute, where almost 40 researchers in the Visual Computing section work within the field. DTU is also part of TUH - Technical University Hospital of Greater Copenhagen, where there is extensive collaboration on image analysis. Pizer has seen imaging technologies evolve dramatically. Each modality measures different physical properties, he noted, and provides different challenges to extract shape properties. Naturally, the conversation turned to AI during the visit to DTU Compute. The AI models, he said, perform extremely well on some fractions of objectives, but poorly on others: "And they offer little ability to tell you when or why they fail. You cannot understand the process inside them. One of the main areas I see people at DTU working on is how to extract from the behaviour of these systems what they do well and what they fail at. Without that understanding, the adoption rate among competent physicians will be slow.” At DTU Compute, Professor Aasa Feragen said, it has been a great pleasure and professionally very interesting to host Stephen Pizer: “Stephen Pizer has supported me greatly as a young researcher, and it was a pleasure to see him here at DTU - Technical University of Denmark for the first time. From a professional perspective, it was also extremely interesting to discuss the applicability of open-source tools and the direction in which medical image analysis is heading.” You can find a link to the full story in the comments below. #ai #medicalimagecomputing #visualcomputing #healthtech

Do you find it difficult to get started using AI in your SME? Then keep reading! 𝐁𝐮𝐢𝐥𝐝 𝐀𝐈 𝐬𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐬 𝐭𝐡𝐚𝐭 𝐦𝐚𝐭𝐭𝐞𝐫 – Join us for two hands-on workshops! The AI Denmark Explorer Workshop is 𝐚 𝐟𝐫𝐞𝐞, 𝐩𝐫𝐚𝐜𝐭𝐢𝐜𝐞-𝐨𝐫𝐢𝐞𝐧𝐭𝐞𝐝 𝐩𝐫𝐨𝐠𝐫𝐚𝐦𝐦𝐞. No coding background is required. 📅 11 May – 𝐅𝐫𝐨𝐦 𝐈𝐝𝐞𝐚 𝐭𝐨 𝐏𝐫𝐨𝐭𝐨𝐭𝐲𝐩𝐞 Create AI products that scale with your business. 📅 13 May – 𝐅𝐫𝐨𝐦 𝐏𝐫𝐨𝐭𝐨𝐭𝐲𝐩𝐞 𝐭𝐨 𝐏𝐫𝐨𝐝𝐮𝐜𝐭 Scale your AI solution without losing control over complexity. The workshops are designed to be suitable for: 🔹 SMEs (size 3-500 employees) and aimed at business‑oriented professionals who want to turn ideas into workable AI prototypes 🔹 business-oriented professionals with ideas to test 🔹 and professionals who want to understand AI as a practical tool. Register by 1 May at the latest. Find the links in the comments 👇 Instructor: Nicki Skafte Detlefsen DTU Compute | DTU - Technical University of Denmark, #AIBoost, and AI Denmark #ai #SME #aiboost

The EU AI Act is no longer a future problem — it's here, and the hard questions are no longer about policy. They're about implementation. How do you actually validate a high-risk AI system for compliance? How do you measure fairness, defend against adversarial threats, and document your choices in a way that holds up to audit? And if you're asking yourself whether it's easier to simply scale down your system to sidestep the stricter requirements — that question deserves a proper answer too. These are the questions we're gathering around at the YODA Workshop — Operationalizing the EU AI Act. 📅 7 May 2026 📍 Copenhagen University, North Campus YODA (Yearning to Operationalize Democratic AI) is a practitioner-first initiative bridging academic research and real-world AI governance. This half-day workshop brings together ML engineers, compliance experts, AI architects, risk managers, and civil society voices to exchange concrete case studies — not theory. You'll leave with: → Peer-validated approaches to EU AI Act compliance → An honest picture of when compliance is feasible — and when descoping is the right call → Connections to researchers, practitioners, and regulatory specialists working on the same problems → Input into an anonymized report shared with the wider community → Access to a long-term network for ongoing technical and policy dialogue We're looking for professionals working on the front lines of high-risk and medium-risk AI in health, finance, public administration, and beyond. If you're building or overseeing AI systems and want to move beyond policy debate into technical practice — this is your room. 🔗 Register at yoda-workshop.dk Pioneer Centre for AI (P1) Camilla Vang Arffmann Institut for Matematiske Fag, Københavns Universitet Datalogisk Institut, Københavns Universitet - DIKU Boris Düdder Olga Kokoulina Rasmus Stig Beck Jensen Sneha Das DTU Compute har #EUAIAct #AIGovernance #ResponsibleAI #AICompliance #MachineLearning #YODA This is an AI generated and human adjusted post!

Jeg elsker at arbejde på DTU - Technical University of Denmark. Det er der mange grunde til. Mest skyldes det selvfølgelig hverdagen og de mennesker jeg ser hver dag på 'Cognitive Systems' sektionen, som jeg føler mig ufatteligt heldig at være endt iblandt (ikke mindst vores frygtløse leder Lars Kai Hansen). Mere generelt er DTU Compute bare et fedt sted at være. Et sted hvor det føles som om der er medvind for ens ideer og initiativer. Og så er det fedt på grund af de studerende, som på en eller anden måde formår at imponere og inspirere en gammel hund som mig til at lære nye tricks hvert eneste år. Og for at det ikke skal være løgn, når kilderne til arbejdsglæden hele vejen til toppen. Tak Christine Nellemann for kommentaren nedenfor! Det er fedt at arbejde et sted hvor det er en selvfølge, at etikken en essentiel del af den vilde teknologiske udvikling vi er igang med lige nu.

🌟 𝐖𝐞𝐥𝐜𝐨𝐦𝐞 Theresa Frehr, 𝐏𝐡𝐃 𝐬𝐭𝐮𝐝𝐞𝐧𝐭 𝐢𝐧 𝐨𝐮𝐫 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞 𝐒𝐲𝐬𝐭𝐞𝐦𝐬 𝐬𝐞𝐜𝐭𝐢𝐨𝐧 𝐚𝐭 𝐭𝐡𝐞 𝐃𝐞𝐩𝐚𝐫𝐭𝐦𝐞𝐧𝐭 𝐨𝐟 𝐀𝐩𝐩𝐥𝐢𝐞𝐝 𝐌𝐚𝐭𝐡𝐞𝐦𝐚𝐭𝐢𝐜𝐬 𝐚𝐧𝐝 𝐂𝐨𝐦𝐩𝐮𝐭𝐞𝐫 𝐒𝐜𝐢𝐞𝐧𝐜𝐞! 🌟   Theresa’s PhD project is titled “𝐅𝐞𝐝𝐞𝐫𝐚𝐭𝐞𝐝 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐟𝐨𝐫 𝐏𝐞𝐫𝐬𝐨𝐧𝐚𝐥𝐢𝐳𝐞𝐝 𝐀𝐮𝐝𝐢𝐨𝐥𝐨𝐠𝐲: 𝐈𝐦𝐩𝐫𝐨𝐯𝐢𝐧𝐠 𝐇𝐞𝐚𝐫𝐢𝐧𝐠 𝐀𝐢𝐝 𝐏𝐞𝐫𝐟𝐨𝐫𝐦𝐚𝐧𝐜𝐞.” 🎧🧠   "In my PhD, I work with federated learning — a distributed learning approach where data remains decentralized, which is especially important for sensitive data that cannot be shared centrally. My goal is to use this method to improve hearing aid performance by tailoring hearing aids to each individual." 🔬✨   If you’re curious to learn more about the project, 👉 https://lnkd.in/eEQN4AQk   We’re excited to follow Theresa’s research journey! 💙📊   DTU - Technical University of Denmark DTU Compute #DTU #DTUCompute #PhD #dkforsk #CognitiveSystems #FederatedLearning #Audiology #HearingAids #MachineLearning #HealthTech

🌟 𝐖𝐞𝐥𝐜𝐨𝐦𝐞 Dinh Quoc Trung Dang, 𝐏𝐡𝐃 𝐬𝐭𝐮𝐝𝐞𝐧𝐭 𝐢𝐧 𝐨𝐮𝐫 𝐕𝐢𝐬𝐮𝐚𝐥 𝐂𝐨𝐦𝐩𝐮𝐭𝐢𝐧𝐠 𝐬𝐞𝐜𝐭𝐢𝐨𝐧 𝐚𝐭 𝐭𝐡𝐞 𝐃𝐞𝐩𝐚𝐫𝐭𝐦𝐞𝐧𝐭 𝐨𝐟 𝐀𝐩𝐩𝐥𝐢𝐞𝐝 𝐌𝐚𝐭𝐡𝐞𝐦𝐚𝐭𝐢𝐜𝐬 𝐚𝐧𝐝 𝐂𝐨𝐦𝐩𝐮𝐭𝐞𝐫 𝐒𝐜𝐢𝐞𝐧𝐜𝐞! 🌟   Trung’s PhD project is called "𝐔𝐧𝐜𝐞𝐫𝐭𝐚𝐢𝐧𝐭𝐲 𝐐𝐮𝐚𝐧𝐭𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐑𝐞𝐬𝐩𝐨𝐧𝐬𝐢𝐛𝐥𝐞 𝐀𝐈 𝐃𝐢𝐚𝐠𝐧𝐨𝐬𝐭𝐢𝐜𝐬".   "In my PhD project, I explore how we can make AI systems more reliable, interpretable, and fair. Even highly accurate models can make critical mistakes — especially in fields like healthcare and law — so we need AI that can express uncertainty, explain its decisions, and avoid bias. By studying how reliability, explainability, and bias interact, we aim to make AI systems more transparent and trustworthy." 🤖✨   We're excited to see where this project leads! 🚀 If you’re curious to learn more about the project, feel free to get in touch via his LinkedIn 👉 Dinh Quoc Trung Dang   DTU - Technical University of Denmark DTU Compute   #DTU #DTUCompute #dkforsk #phd #ResponsibleAI #MachineLearning #UncertaintyQuantification #VisualComputing

𝐀 𝐧𝐞𝐰 𝐬𝐨𝐜𝐢𝐚𝐥 𝐧𝐞𝐭𝐰𝐨𝐫𝐤 𝐦𝐨𝐝𝐞𝐥 𝐩𝐫𝐨𝐯𝐢𝐝𝐞𝐬 𝐧𝐞𝐰 𝐰𝐚𝐲𝐬 𝐨𝐟 𝐮𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝𝐢𝐧𝐠 𝐡𝐨𝐰 𝐫𝐞𝐚𝐥-𝐥𝐢𝐟𝐞, 𝐟𝐚𝐜𝐞-𝐭𝐨-𝐟𝐚𝐜𝐞 𝐬𝐨𝐜𝐢𝐚𝐥 𝐧𝐞𝐭𝐰𝐨𝐫𝐤𝐬 𝐚𝐫𝐞 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝.  The model has been developed by DTU Compute in collaboration with Yale University. The model draws on data from 176 remote villages in Honduras and is relevant to research on the optimal spread of desirable information. The findings have just been published as 𝑒𝑎𝑟𝑙𝑦 𝑎𝑐𝑐𝑒𝑠𝑠 𝑖𝑛 𝑁𝑎𝑡𝑢𝑟𝑒 𝐶𝑜𝑚𝑚𝑢𝑛𝑖𝑐𝑎𝑡𝑖𝑜𝑛𝑠. The model is designed to characterise structure of 𝐦𝐮𝐥𝐭𝐢𝐩𝐥𝐞𝐱 𝐬𝐨𝐜𝐢𝐚𝐥 𝐧𝐞𝐭𝐰𝐨𝐫𝐤𝐬 – networks with several layers of relations between individuals, and where the model takes into account that each person is part of several networks simultaneously. A person may be someone’s friend, but also someone from whom they seek health advice or a financial loan, for instance. "Our developed network model makes it easier to understand how multiplex social networks are organised by characterising the roles of all the individuals within a village. Although it is basic research, such knowledge can potentially have great significance,” says Professor of Machine Learning at DTU Compute, Morten Moerup. It is former DTU Compute PhD student Nikolaos Nakis who is the first author of the study. Morten Mørup and Professor Sune Lehmann have been his supervisors. During his PhD, Nikolaos was on exchange with Professor Nicholas Christakis at the Human Nature Lab at Yale University, where he was introduced to the extensive field study. When Nikolaos returned to Denmark, work began on analysing Christakis’ data using ideas from his PhD project. Today, Nikolaos Nakis is a postdoc at the Human Nature Lab, and the work on developing the network model (𝐌𝐮𝐥𝐭𝐢𝐩𝐥𝐞𝐱 𝐋𝐚𝐭𝐞𝐧𝐭 𝐓𝐫𝐚𝐝𝐞‑𝐨𝐟𝐟 𝐌𝐨𝐝𝐞𝐥 (𝐌𝐋𝐓)) has now been completed. “What excites me about this model is that it helps uncover the hidden roles people play across different domains of life. Rather than looking at relationships in isolation, it shows how different kinds of ties fit together and how people navigate these trade-offs in everyday life,” says Nikolaos Nakis. #socialnetworks #dkforsk DTU - Technical University of Denmark Find the link to the story in the comment. Here you can also read details of the figure below.

AortaExplorer is out! We have just published an open source, fully automated AI framework for comprehensive aortic analysis from contrast-enhanced CT scans. Why does this work matter? Aortic morphology is central to understanding cardiovascular risk, aging, and disease—but manual assessment is slow and hard to scale. With AortaExplorer, we validated automated diameter measurements in over 10,000 CT scans from Copenhagen General Population Study, introduced new metrics like aortic tortuosity, and cut analysis time to just a few minutes per scan.   This work was a joint effort across DTU Compute, Rigshospitalet and Novo Nordisk, with contributions from Rasmus Paulsen, Linnea Hjordt Juul, Klaus Fuglsang Kofoed, Jørgen Tobias Kühl, Michael Huy Cuong Pham, Kristine Aavild Sørensen and myself. I hope the tool will empower others working in large cohort studies and clinical imaging research. The tool is publicly available on GitHub: https://lnkd.in/eSXSDpXe, and don’t hesitate to reach out to me or the other co-authors if you are interested! Find the paper here: https://lnkd.in/ecZna-Ne

𝐉𝐨𝐛 𝐒𝐞𝐫𝐯𝐢𝐜𝐞 | 𝐉𝐨𝐢𝐧 𝐨𝐮𝐫 𝐭𝐞𝐚𝐦𝐬! Would you like to be part of DTU Compute? We currently have a handful of open positions. At DTU Compute, you will become part of an international and interdisciplinary environment where research, teaching, and innovation go hand-in-hand. We believe in a diverse, inclusive workplace with a flexible work-life balance. We look forward to hearing from you! Find the link to the jobs in the comments. 𝑁𝑜𝑡𝑒 𝑡ℎ𝑎𝑡 𝑡ℎ𝑒 𝑑𝑒𝑎𝑑𝑙𝑖𝑛𝑒 𝑖𝑠 𝑐𝑜𝑚𝑖𝑛𝑔 𝑢𝑝 𝑠𝑜𝑜𝑛 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑗𝑜𝑏𝑠! 𝐀𝐬𝐬𝐨𝐜𝐢𝐚𝐭𝐞 𝐏𝐫𝐨𝐟𝐞𝐬𝐬𝐨𝐫 𝐨𝐫 𝐃𝐓𝐔 𝐓𝐞𝐧𝐮𝐫𝐞 𝐓𝐫𝐚𝐜𝐤 𝐀𝐬𝐬𝐢𝐬𝐭𝐚𝐧𝐭 𝐏𝐫𝐨𝐟𝐞𝐬𝐬𝐨𝐫 𝐢𝐧 𝐂𝐨𝐦𝐩𝐮𝐭𝐞𝐫 𝐒𝐜𝐢𝐞𝐧𝐜𝐞 (𝐒𝐨𝐟𝐭𝐰𝐚𝐫𝐞 𝐒𝐲𝐬𝐭𝐞𝐦𝐬 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠), 📣 Deadline 15 March. 𝐏𝐡𝐃 𝐬𝐜𝐡𝐨𝐥𝐚𝐫𝐬𝐡𝐢𝐩 𝐢𝐧 𝐍𝐨𝐧𝐥𝐢𝐧𝐞𝐚𝐫 𝐈𝐧𝐯𝐞𝐫𝐬𝐞 𝐏𝐫𝐨𝐛𝐥𝐞𝐦𝐬 – 𝐓𝐡𝐞𝐨𝐫𝐲, 𝐂𝐨𝐦𝐩𝐮𝐭𝐚𝐭𝐢𝐨𝐧𝐬 𝐚𝐧𝐝 𝐄𝐱𝐩𝐞𝐫𝐢𝐦𝐞𝐧𝐭𝐬, 📣 Deadline 15 March. 𝐏𝐨𝐬𝐭𝐝𝐨𝐜 𝐢𝐧 𝐌𝐚𝐜𝐡𝐢𝐧𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 (𝐌𝐋) 𝐟𝐨𝐫 𝐁𝐢𝐨𝐥𝐨𝐠𝐢𝐜𝐚𝐥 𝐃𝐚𝐭𝐚, Deadline 25 March. 𝐏𝐡𝐃 𝐬𝐜𝐡𝐨𝐥𝐚𝐫𝐬𝐡𝐢𝐩 𝐢𝐧 𝐂𝐫𝐲𝐩𝐭𝐨𝐠𝐫𝐚𝐩𝐡𝐲, Deadline 19 April. 𝐀𝐬𝐬𝐨𝐜𝐢𝐚𝐭𝐞 𝐏𝐫𝐨𝐟𝐞𝐬𝐬𝐨𝐫 𝐨𝐫 𝐃𝐓𝐔 𝐓𝐞𝐧𝐮𝐫𝐞 𝐓𝐫𝐚𝐜𝐤 𝐀𝐬𝐬𝐢𝐬𝐭𝐚𝐧𝐭 𝐏𝐫𝐨𝐟𝐞𝐬𝐬𝐨𝐫 𝐢𝐧 𝐂𝐲𝐛𝐞𝐫𝐬𝐞𝐜𝐮𝐫𝐢𝐭𝐲, Deadline 30 April. #jobservice #dkforsk #datascience #computerscience #cybersecurity