Investigating a residual neural network to generated customizable RF pulses for MRI
Seger Nelson1 and Rebecca Emily Feldman2,3
1Computer Science, Mathematics, Physics, and Statistics, University of British Columbia, Kelowna, BC, Canada, 2Computer Science, Mathematics, Physics, and Statistics, University of British Columbia Okanagan, Kelowna, BC, Canada, 3Biomedical Engineering and Imaging Institute, Icahn School of Medicine at Mount Sinai, New York, NY, United States
We investigate a method for the design of radio frequency (RF) pulses for custom simultaneous multi slice spatial profiles. We trained a variation of residual neural networks on a dataset of 96,000 generated adiabatic “power independent of the number of slices” (PINS) RF pulses, varying parameters such as gradient duty cycle, number of pulse lobes, quadratic phase strength, and transition and filter bandwidth. We compared the direct-design RF pulses to those generated by our model.
This abstract and the presentation materials are available to members only;
a login is required.