Deep Learning Optimization of Acquisition-Parameter and Slice-Ordering Schedule for Fast Quantitative Multi-Slice CEST

Motivation: To obtain optimized schedules for a multi-slice CEST sequence Goal(s): To develop a deep learning schedule optimization approach for joint continuous and discrete scan parameters optimization. Approach: A joint deep learning framework was developed and used to generate optimized 10-slice acquisition schedules. The optimized schedules were compared in simulations to a random 10-slice schedule and an optimized 1-slice schedule. The optimized 10-slice and 1-slice schedules were also compared in vivo in a healthy subject. Results: The proposed approach yielded acquisition schedules that provided tissue maps with lower error than a random schedule or an optimized 1-slice schedule. Impact: This work proposes a novel deep learning framework that enables simultaneous optimization of continuous and discrete scan parameters for a rapid quantitative multi-slice CEST sequence. The proposed method provides acquisition schedules that enable quantitative CEST imaging with lower error.