The course takes place over 2 days
On day 1, we provide a basic introduction to machine learning for imaging, ranging from classification to regression. We introduce basic building blocks to build your first neural network, in both real-valued and complex-valued domains. Once the neural networks are defined, we will talk about training procedures, loss functions, and database selection. The presented building blocks are then linked to MR image reconstruction. We have a quick recap of the basics in MRI reconstruction, and we will then see the basic building blocks in action. Recent developments and applications on machine learning in MR reconstruction are highlighted. Hands-on examples are provided, including guided tutorials, and tasks that are carried out as homework.
On day 2, we present the solutions to the homework. We end our course with presentations of invited speakers, talking about efficient learning strategies and practical implementations of machine learning applications in MRI.
Registration Fees
Ticket Type | Fee |
ESMRMB Member | € 25 |
ESMRMB Junior/Radiographer Member | € 15 |
Non-Member | € 50 |
NOTE! REGISTRATION CLOSES JUL 20:00 (CEST)!
ESMRMB Members can benefit from discounted member rates. Please ensure an active and paid ESMRMB 2022 membership before registering. If your membership is pending some documents/proofs for activation might be missing.
Link to Online Course
The link to join will be send to you after online registration before the course beginning.
Terms and Conditions
ESMRMB 2022 Education Terms & Conditions
ISMRM endorsed
This educational event is endorsed by ISMRM – International Society for Magnetic Resonance in Medicine
Educational Levels & Learning Objectives
This course is intended for MR physicists, basic scientists, clinicians and PhD students who already have a working knowledge of the basic principles of MRI and who wish to gain knowledge and improve their understanding of Machine Learning in MR Imaging.
Participants of this course will learn:
- Learn the basics of machine learning
- Learn how to build and train a neural network
- Link machine learning to MR reconstruction
- Learn how to process complex-valued data
- Learn about recent machine learning applications for MR reconstruction
Course Organiser
Kerstin Hammernik, Technical University of Munich/DE
Thomas Küstner, Tübingen/DE
Faculty
Thomas Küstner
Kerstin Hammernik
Jonathan Tamir
Moritz Blumenthal
Efrat Shimron
Preliminary Programme
Day 1
Basics in Machine Learning
Thomas Küstner – 30 min
- Classification / Segmentation / Regression
- Types of Learning
- Training Procedure
- Loss Function
- Basic Building Blocks
Break & Whiteboard (15 min)
Going complex! Linking machine learning to MR reconstruction
Kerstin Hammernik – 30 min
- Recap: Basics in MRI Reconstruction
- Parallel Imaging
- Compressed Sensing, Low-rank
- Complex-valued Processing
- Complex-valued building blocks
- Data consistency layer
Break & Whiteboard (15 min)
Machine learning MR reconstruction in action
Kerstin Hammernik – 30 min
- Types of Learning for MRI Reconstruction
- Denoising
- PnP
- Unrolled Optimization
- k-space
- Loss functions
- Perceptual
- GAN
Break & Whiteboard (15 min)
Applications of ML MRI Reconstruction
Thomas Küstner – 30 min
Closing remarks, handout of homework
15 min
Day 2
Presentation of solutions to homework
Talk “Deep Learning with BART” by Moritz Blumenthal (University of Göttingen, DE) – 30min
Talk “Memory Efficient Learning “ by Jonathan Tamir (University of Texas at Austin, USA) – 30min
Talk “Inverse Data Crimes” by Efrat Shimron (UC Berkeley, USA) – 30 min
Q&A / panel discussion / conclusion – 10 min