Detailed Program
Monday 7th April 2025
Content: Registration for the Spring School will take place in the foyer of the DZHI building. Please make sure to register before the start of the first session.
Details of each session will be announced later
Content: Bridging Domain Expertise with AI for Next-Generation MRI: Known
Operators, Low-Field Applications, Contrast Discovery, and Sequence
Programming
Recent advances in deep learning have enabled the seamless integration
of known operators—such as Fourier transforms and physical forward
models—into neural architectures, greatly enhancing the fidelity and
interpretability of medical image reconstruction. In low-field MRI,
variational networks leverage these operator constraints to stabilise
the iterative reconstruction process, compensating for reduced
signal-to-noise ratios while preserving essential anatomical
information. Furthermore, the paradigm of differentiable MRI
reconstruction opens new horizons in sequence design by allowing the
joint optimization of sampling trajectories in k-space with respect to
specific signal characteristics. Approaches like MRZero permit
model-driven discovery of scanning protocols, reducing the need for
labor-intensive experimentation. Going one step further, the GPT4MR
initiative harnesses generative large language models (e.g., OpenAI o1
or xAI’s Grok) to reason about MRI physics and pulse sequences, offering
intuitive programming support for sequence development. The synergy of
operator-based deep networks, differentiable reconstruction pipelines,
and next-generation language models sets a new direction for innovation
in MRI, promising enhanced image quality, reduced acquisition times, and
unprecedented adaptability in clinical and research settings.
Coffee Break
Details of each session will be announced later
Content: Lecture: Solving Differential Equations with Neural Networks
This session provides a comprehensive introduction to differential equations and explores
how neural networks, particularly Physics-Informed Neural Networks (PINNs),
can be used as a powerful tool for solving them.
Designed for both beginners and those with some prior exposure to machine
learning or differential equations, the lecture combines theory with hands-on
practice using modern tools.
Access the materials on GitHub
Lunch Break
Details of each session will be announced later
Content: Tutorial: Solving Differential Equations with Neural Networks
This session provides a practical introduction to differential equations and explores
how neural networks, particularly Physics-Informed Neural Networks (PINNs),
can be used as a powerful tool for solving them.
Designed for both beginners and those with some prior exposure to machine
learning or differential equations, the lecture combines theory with hands-on
practice using modern tools.
Access the materials on GitHub
Exercise Sheet:
Get the exercise sheet here
Coffee Break
Details of each session will be announced later
Details of each session will be announced later
Details of each session will be announced later
Tuesday 8th April 2025
Content: In this session we will cover the basics of MRI physics, including the principles of magnetic resonance, the role of magnetic fields, and the fundamentals of image formation.
Coffee Break
Details of each session will be announced later
Content: This session will provide an introduction to the basics of numerical simulations, including the principles of numerical methods, their applications in MRI, and the importance of simulations in advancing MRI technology.
Lunch Break
Details of each session will be announced later
Content: This session will provide a hands-on introduction to numerical simulations, including practical exercises and examples of how numerical methods are applied in MRI.
Coffee Break
Details of each session will be announced later
Content: Artificial intelligence (AI) is transforming cardiology, from automated image analysis to predictive risk modelling and AI-driven decision support. This keynote will explore the current landscape of AI applications in cardiovascular medicine, focusing on its role in cardiac imaging, diagnosis, risk stratification, and personalized therapy. While AI has demonstrated remarkable capabilities—enhancing diagnostic accuracy, improving workflow efficiency, and enabling early disease detection—its real-world deployment faces critical challenges, including bias, interpretability, regulatory constraints, and clinical integration. This talk will bridge the gap between AI research and clinical practice, providing insights into successful applications, ongoing limitations, and the future of AI in cardiovascular medicine. The session will conclude with a discussion on open questions and future research directions to foster collaboration between AI scientists and clinicians.
Wednesday 9th April 2025
Content: This session will provide an overview of the differences between 1.5T, 3T, and 7T MRI systems, including their advantages and disadvantages, safety considerations, and other important issues related to their use in clinical practice and research.
Coffee Break
Details of each session will be announced later
Content: This session will provide an overview of the MAGNET4Cardiac7T project, including its objectives, methodology, and expected outcomes. The session will give you a insight into the used models and the project's approach to simulating 7T MRI using deep learning techniques.
Lunch Break
Details of each session will be announced later
Content: The hackathon is an opportunity for participants to work on a specific project or problem related to the topics covered in the Spring School.
Participants will be divided into teams and will have the chance to collaborate, share ideas, and develop solutions.
The hackathon will be supervised to answer questions and provide guidance.
Participants are encouraged to bring their own laptops and any materials they may need for the hackathon.
To get started with the hackathon visit the GitHub Repository.
Information about the evaluation and the evaluations script can be found in the Hackathon Evaluation Repository.
Coffee Break
Details of each session will be announced later
Content: The hackathon is an opportunity for participants to work on a specific project or problem related to the topics covered in the Spring School.
Participants will be divided into teams and will have the chance to collaborate, share ideas, and develop solutions.
The hackathon will be supervised to answer questions and provide guidance.
Participants are encouraged to bring their own laptops and any materials they may need for the hackathon.
To get started with the hackathon visit the GitHub Repository.
Information about the evaluation and the evaluations script can be found in the Hackathon Evaluation Repository.
Thursday 10th April 2025
Content: The presentation focuses on physics-based machine learning for cardiac MRI reconstruction. We'll begin with a brief overview of cardiac MRI, outlining its potential and current challenges. We will then define physics-based machine learning in MR reconstruction, contrasting it with alternative approaches of transforming undersampled acquisitions into high-quality images. A key component of physics-based methods is ensuring data consistency, and we will discuss the critical role of transfer functions in modeling gradient behavior and accurately determining k-space trajectories. We will then explore diffusion probabilistic models, a powerful generative approach that has shown significant promise for physics-based MR reconstruction, and demonstrate its potential for high-quality cardiac MRI.
Coffee Break
Details of each session will be announced later
Content: The hackathon is an opportunity for participants to work on a specific project or problem related to the topics covered in the Spring School.
Participants will be divided into teams and will have the chance to collaborate, share ideas, and develop solutions.
The hackathon will be supervised to answer questions and provide guidance.
Participants are encouraged to bring their own laptops and any materials they may need for the hackathon.
To get started with the hackathon visit the GitHub Repository.
Information about the evaluation and the evaluations script can be found in the Hackathon Evaluation Repository.
Lunch Break
Details of each session will be announced later
Content: Fast. Sharp. Informed. - A New Era in MRI
Coffee Break
Details of each session will be announced later
Content: The hackathon is an opportunity for participants to work on a specific project or problem related to the topics covered in the Spring School.
Participants will be divided into teams and will have the chance to collaborate, share ideas, and develop solutions.
The hackathon will be supervised to answer questions and provide guidance.
Participants are encouraged to bring their own laptops and any materials they may need for the hackathon.
To get started with the hackathon visit the GitHub Repository.
Information about the evaluation and the evaluations script can be found in the Hackathon Evaluation Repository.
Details of each session will be announced later
Friday 11th April 2025
Details of each session will be announced later
Coffee Break
Details of each session will be announced later
Content: In 2016, machine learning techniques have been first introduced to solve the inverse problem of MR image generation from accelerated acquisitions (1,2,3). Since then, the field has grown substantially, and a wide range of machine learning methods have been developed, applied to a wide range of imaging applications and already rolled out as clinical products by all major scanner manufacturers. In this lecture, I will start with the background of an artificial intelligence process to generate MR images from the acquired measurement data. In particular, I will discuss physics informed architectures that map iterative algorithms onto neural networks. I will discuss their performance for a range of clinical applications (4,5,6) as well as ongoing challenges related to data availability (7), generalizability and validation of the results. I will also include a discussion of the lessons learnt from the recent fastMRI image reconstruction challenges organized jointly with Facebook AI research (8).
References:
1. Hammernik et al. Learning a variational model for compressed sensing MRI reconstruction. Proc. ISMRM p33 (2016).
2. Hammernik et al. Learning a Variational Network for Reconstruction of Accelerated MRI Data. MRM, 79:3055-3071 (2018).
3. Knoll et al. Deep Learning Methods for Parallel Magnetic Resonance Image Reconstruction. IEEE Signal Processing Magazine, 37:1:128-40 (2020).
4. Johnson et al. Deep learning reconstruction enables highly accelerated biparametric MR imaging of the prostate. JMRI 56: 184-195 (2022).
5. Johnson et al. Deep Learning Reconstruction Enables Prospectively Accelerated Clinical Knee MRI. Radiology 307:e220425 (2023).
6. Vornehm et al. CineVN: Variational network reconstruction for rapid functional cardiac cine MRI. Magnetic Resonance in Medicine 93:138-150 (2025)
7. Knoll et al. fastMRI: a publicly available raw k-space and DICOM dataset for accelerated MR image reconstruction using machine learning. Radiology Artificial Intelligence (2:2020).
8. Knoll et al. Advancing machine learning for MR image reconstruction with an open competition: Overview of the 2019 fastMRI challenge. MRM 84 (6), 3054-3070 (2020).
Lunch Break
Details of each session will be announced later
Details of each session will be announced later