The Ritz Adjoint Method for MRI Pulse Design

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging Physics

Background:

• High-field MRI faces challenges with magnetic field inhomogeneities and patient-specific variations.
• Current methods for designing tailored excitation pulses are computationally intensive and slow for real-time customization.

Purpose of the Study:

• To develop a faster method for designing subject-specific radiofrequency (RF) and gradient waveforms for high-field MRI.
• To improve the efficiency and feasibility of real-time, customized pulse optimization in MRI.

Main Methods:

• Representing RF and gradient waveforms using a global Chebyshev polynomial basis to reduce optimization variables.
• Utilizing the adjoint method for efficient computation of derivatives with respect to basis coefficients.
• Implementing GPU acceleration for derivative calculations and enforcing system/safety constraints.

Main Results:

• Achieved a 5- to 10-fold speedup in subject-specific pulse optimization for non-selective excitations.
• Demonstrated comparable speed gains for slice-selective pulse designs.
• Enabled real-time, subject-specific optimization for advanced MRI pulse types.

Conclusions:

• A global waveform basis approach significantly accelerates tailored excitation pulse design in high-field MRI.
• This method enhances the practicality of real-time, subject-specific MRI pulse optimization.
• The findings pave the way for more advanced and personalized MRI applications.