לוח שנה משולב

Zoom Lecture: https://weizmann.zoom.us/j/99058507421


Magnetic Resonance Imaging (MRI) is a superb imaging modality that provides high-quality images of the human body. However, one of its major limitations is the long acquisition time, which hinders the MRI clinical use. The acquisition time can be shortened by acquiring less data; however, this requires suitable methods for accurate image reconstruction from subsampled data, which is acquired with a sub-Nyquist rate.
In this seminar, four novel methods for image reconstruction from subsampled data will be presented. These methods build upon the well-established frameworks of Parallel Imaging (PI) and Compressed Sensing (CS), utilize a-priori knowledge about data sparsity, and address current limitations of PI-CS methods. The first two methods accelerate static MRI scans by introducing the Convolution-based Reconstruction (CORE) framework, which offers a parameter-free non-iterative reconstruction. Experiments with in-vivo 7T brain data demonstrated that these methods perform comparably to the well-established GRAPPA and l1-SPIRiT methods, with the advantage of shorter computation times and reduced need for parameter calibration. The next two developed methods accelerate dynamic MRI scans that provide temperature monitoring in High Intensity Focused Ultrasound (MRgHIFU) thermal ablation treatments. The developed methods enable rapid MR monitoring by reconstructing temperature changes from subsampled data. Validation experiments were performed with in-vivo data from clinical treatments of prostate cancer in humans; these showed that the proposed methods significantly outperform two state-of-the-art methods in the temperature reconstruction task