Moritz Zaiss, FAU
We present an open-source, hands-on MR sequence programming and simulation course based on Pulseq. Pulseq (https://pulseq.github.io/) is a well-suited tool to teach students MR sequence programming, as it is simple, standardized, and vendor agnostic. Sequences coded with Pulseq fulfill physical constraints and can be executed on any scanner if an interpreter is available. This makes it possible to conclude a course in a rewarding way, allowing students to measure and reconstruct their own imaging sequences by executing them on a real MRI scanner.
However, teaching MR programming effectively requires immediate feedback to the students from the beginning, starting with analyzing the signal of the first self-written MR sequence. Getting quick feedback for the first experiments, MRI sequence prototypes, or unconventional ideas requires a universal Bloch simulation. Our course tool (https://github.com/mzaiss/MRTwin_pulseq) provides a fast and accurate Phase Distribution Graph Bloch simulation. It can read .seq files directly and calculate accurate signals of a virtual phantom within seconds, which are then added to the Pulseq sequence scheme. This enables a fast feedback loop directly on the students’ PCs while prototyping a sequence or a reconstruction.
In this setup, simulation and measurement are fully interchangeable. This makes it possible to investigate image artifacts, image contrast weighting, or reconstruction problems early on by using the simulation before measuring the final MR sequence on a real system. The efficient simulation also allows to generate synthetic MRI data tailored to the sequence/contrast/artifact of interest.
Heidelberg – Ünal Akünal, Markus Bujotzek. Peter Neher
Uniklinik Köln - Astha Jaiswal
SATORI – Bianca Lassen-Schmidt
MINT Medical GmbH – Sebastian Schaefer
ImFusion – Julia Rackerseder
TuDa - John Kalkhof
UME - Kim Moon
RACOON is developing a federated infrastructure to make multicenter clinical and radiological data sets available for research projects in the field of data analysis and method development. RACOON enables the scaling of medical research projects from a single university hospital up to the national level. This interactive tutorial is designed to give you a comprehensive understanding of the RACOON infrastructure and its many applications.
We will start with a detailed overview of the existing system. Here we will explain and demonstrate the architecture and the various components of RACOON in detail. In the next part of the tutorial, we will focus on how RACOON can be used for scientific projects. We will present and discuss different use cases, such as data annotation, running custom or existing AI algorithms, (federated) training of models, as well as analyzing results.
It is very important to us that this tutorial is as interactive as possible. We therefore welcome your questions and feedback, especially on how you could use RACOON for your own projects. We will do our best to address your concerns and help you get the most out of RACOON. Finally, we would like to give you the opportunity to explore the system for yourself.
Yixing Huang, Uniklinik Erlangen
In this tutorial, our aim is to build a state-of-the-art flat-panel CT reconstruction software. We will begin by developing the basics of CT reconstruction collectively. Each participant will create a basic CT reconstruction pipeline capable of reconstructing flat-panel CT images. Our focus will be on three primary types of reconstruction: parallel-beam, fan-beam, and cone-beam, each with its unique attributes and challenges.
Parallel-beam reconstruction utilizes X-rays emitted in parallel lines, a method where the intricacy lies in addressing the limited data problem that can lead to image artifacts. In contrast, fan-beam reconstruction employs a fan-shaped beam, offering a quicker scanning process but requiring precise correction for the varying distances between the X-ray source and the detector elements to avoid distortion.
The most complex among these, cone-beam reconstruction, involves a cone-shaped X-ray beam and a flat-panel detector. This method, prevalent in modern CT scanners, especially for dental and extremity imaging, introduces more sophisticated distortions known as cone-beam artifacts. These require advanced algorithms for correction, emphasizing the importance of detailed understanding and implementation of these algorithms in the course.
An essential component of our course will be the exploration of hardware acceleration, particularly using GPUs. The computationally intensive nature of CT reconstruction algorithms, especially iterative methods, necessitates the use of hardware acceleration. GPUs can significantly expedite the reconstruction process, enabling real-time applications and optimizing performance, a crucial factor in clinical settings where time and accuracy are paramount.
Through this tutorial, participants will not only learn the technical aspects of CT reconstruction but also understand the practical challenges and considerations in implementing these techniques with flat-panel detectors. This comprehensive approach is designed to equip learners with the knowledge and skills needed to advance the field of medical imaging, fostering the development of more efficient, high-quality imaging techniques.