GenMRI

Conventional MRI acquisitions are carefully engineered for reconstruction using a Fourier transform. Those older techniques are either slow or provide only qualitative images. Newer approaches allow complex dynamical behavior of the magnetization during the measurement process, which then requires modeling of the MR signal for image reconstruction to extract quantitative parameters. Despite such advances, state-of-the-art pulse sequences are still affected by arbitrary restrictions imposed by conventional designs and consist of hand-designed building blocks that are not necessarily optimal.

In this project, our aim is to develop a generalized framework for pulse sequence design and image reconstruction that provides the flexibility needed to support techniques based on optimal control with end-to-end optimization of image acquisition and reconstruction. The developed framework will make it possible to apply those techniques in MRI experiments. This requires concepts for describing generalized pulse sequences in a mathematically precise and convenient form useful for numerical optimization. Fully exploiting the ideas developed in this SFB also requires implementation of interfaces to the measurement device. In addition to these basic tasks, we then want to understand to what extent system imperfections and experimental factors might prevent exploiting the full potential of this framework, and therefore need to be taken into account in the mathematical work. From image reconstruction, we will extend our generalized Bloch model reconstruction so that it can directly be used with generalized pulse sequences.