Realtime design of radio-frequency pulses and gradient pulses in magnetic resonance imaging

A computer program comprising machine executable instructions stored on a non-transitory computer readable medium for execution by a computational system controlling a medical system, wherein the computer program further comprises:

• a convolutional neural network configured to receive as input a complex array encoding a selection of at least one excitation field of view, wherein the convolutional neural network is configured to output a radio frequency waveform and multiple spatially selective gradient pulse waveforms, wherein the convolutional neural network is a multi-task convolutional neural network with a first output for the radio frequency waveform and a separate output for each of the multiple spatially selective gradient pulse waveforms, wherein execution of the machine executable instructions causes the computational system to:
• receive the selection of the at least one excitation field of view;
• receive initial pulse sequence commands configured for controlling a magnetic resonance imaging system to acquire k-space data descriptive of subject;
• encode the complex array using the at least one excitation field of view;
• receive the radio frequency waveform and the multiple spatially selective gradient pulse waveforms in response to inputting the complex array into the convolutional neural network;
• and
• construct modified pulse sequence commands by modifying the initial pulse sequence commands with the radio frequency waveform and the multiple spatially selective gradient pulse waveforms such that the pulse sequence commands are configured to control the magnetic resonance imaging system to acquire the k-space data from the at least one excitation field of view.

A method of medical imaging using a convolutional neural network for executution by a computational system, wherein the convolutional neural network is configured to receive as input a complex array encoding a selection of at least one excitation field of view, wherein the convolutional neural network is configured to output a radio frequency waveform and multiple spatially selective gradient pulse waveforms, wherein the convolutional neural network is a multi-task convolutional neural network with a first output for the radio frequency waveform and a separate output for each of the multiple spatially selective gradient pulse waveforms, wherein the method comprises:

• receiving, by the computational system, the selection of the at least one excitation field of view;
• receiving, by the computational system, initial pulse sequence commands configured for controlling a magnetic resonance imaging (MRI) system to acquire k-space data descriptive of subject-;
• encoding, by the computational system, the complex array using the at least one excitation field of view;
• receiving, by the computational system, the radio frequency waveform and the multiple spatially selective gradient pulse waveforms in response to inputting the complex array into the convolutional neural network;
• and
• constructing, by the computational system, modified pulse sequence commands by modifying the initial pulse sequence commands with the radio frequency waveform and the multiple spatially selective gradient pulse waveforms such that the pulse sequence commands are configured to control the magnetic resonance imaging system to acquire the k-space data from the at least one excitation field of view.

A method of training a convolutional neural network by a computational system, wherein the method comprises repeatedly:

• generating a training radio frequency waveform and multiple training spatially selective gradient pulse waveforms using a selective excitation pulse design algorithm, by the computational system;
• calculating a modeled excitation field of view, by the computational system, by inputting the training radio frequency waveform and training multiple spatially selective gradient pulse waveforms into a magnetic resonance imaging signal model;
• receiving a forward propagated radio frequency waveform and multiple forward propagated spatially selective training gradient pulse waveforms, by the computational system, by inputting the modeled excitation field of view into the convolutional neural network;
• updating parameters of the convolutional neural network, by the computational system, by performing back propagation with the training radio frequency waveform and the forward propagated radio frequency waveform and with matching pairs of the multiple training spatially selective gradient pulse waveforms and the multiple forward propagated spatially selective training gradient pulse waveforms.