-
Clipped DeepControl: deep neural network two-dimensional pulse design with an amplitude constraint layer
Authors:
Mads Sloth Vinding,
Torben Ellegaard Lund
Abstract:
Advanced radio-frequency pulse design used in magnetic resonance imaging has recently been demonstrated with deep learning of (convolutional) neural networks and reinforcement learning. For two-dimensionally selective radio-frequency pulses, the (convolutional) neural network pulse prediction time (few milliseconds) was in comparison more than three orders of magnitude faster than the conventional…
▽ More
Advanced radio-frequency pulse design used in magnetic resonance imaging has recently been demonstrated with deep learning of (convolutional) neural networks and reinforcement learning. For two-dimensionally selective radio-frequency pulses, the (convolutional) neural network pulse prediction time (few milliseconds) was in comparison more than three orders of magnitude faster than the conventional optimal control computation. The network pulses were from the supervised training capable of compensating scan-subject dependent inhomogeneities of B0 and B+1 fields. Unfortunately, the network presented with a non-negligible percentage of pulse amplitude overshoots in the test subset, despite the optimal control pulses used in training were fully constrained. Here, we have extended the convolutional neural network with a custom-made clipping layer that completely eliminates the risk of pulse amplitude overshoots, while preserving the ability to compensate the inhomogeneous field conditions.
△ Less
Submitted 21 January, 2022;
originally announced January 2022.
-
Optimal control gradient precision trade-offs: application to fast generation of DeepControl libraries for MRI
Authors:
Mads Sloth Vinding,
David L. Goodwin,
Ilya Kuprov,
Torben Ellegaard Lund
Abstract:
We have recently demonstrated supervised deep learning methods for rapid generation of radiofrequency pulses in magnetic resonance imaging (https://doi.org/10.1002/mrm.27740, https://doi.org/10.1002/mrm.28667). Unlike the previous iterative optimization approaches, deep learning methods generate a pulse using a fixed number of floating-point operations - this is important in MRI, where patient-spe…
▽ More
We have recently demonstrated supervised deep learning methods for rapid generation of radiofrequency pulses in magnetic resonance imaging (https://doi.org/10.1002/mrm.27740, https://doi.org/10.1002/mrm.28667). Unlike the previous iterative optimization approaches, deep learning methods generate a pulse using a fixed number of floating-point operations - this is important in MRI, where patient-specific pulses preferably must be produced in real time. However, deep learning requires vast training libraries, which must be generated using the traditional methods, e.g. iterative quantum optimal control methods. Those methods are usually variations of gradient descent, and the calculation of the fidelity gradient of the performance metric with respect to the pulse waveform can be the most numerically intensive step. In this communication, we explore various ways in which the calculation of fidelity gradients in quantum optimal control theory may be accelerated. Four optimization avenues are explored: truncated commutator series expansions at zeroth and first order, a novel midpoint truncation scheme at first order, and the exact complex-step method. For the spin systems relevant to MRI, the first-order truncation is found to be sufficiently accurate, but also up to five times faster than the machine precision gradient. This makes the generation of training databases for the machine learning methods considerably more realistic.
△ Less
Submitted 2 July, 2021;
originally announced July 2021.
-
DeepControl: 2D RF pulses facilitating $B_1^+$ inhomogeneity and $B_0$ off-resonance compensation in vivo at 7T
Authors:
Mads Sloth Vinding,
Christoph Stefan Aigner,
Sebastian Schmitter,
Torben Ellegaard Lund
Abstract:
Purpose: Rapid 2D RF pulse design with subject specific $B_1^+$ inhomogeneity and $B_0$ off-resonance compensation at 7 T predicted from convolutional neural networks is presented.
Methods: The convolution neural network was trained on half a million single-channel transmit, 2D RF pulses optimized with an optimal control method using artificial 2D targets, $B_1^+$ and $B_0$ maps. Predicted pulse…
▽ More
Purpose: Rapid 2D RF pulse design with subject specific $B_1^+$ inhomogeneity and $B_0$ off-resonance compensation at 7 T predicted from convolutional neural networks is presented.
Methods: The convolution neural network was trained on half a million single-channel transmit, 2D RF pulses optimized with an optimal control method using artificial 2D targets, $B_1^+$ and $B_0$ maps. Predicted pulses were tested in a phantom and in vivo at 7 T with measured $B_1^+$ and $B_0$ maps from a high-resolution GRE sequence.
Results: Pulse prediction by the trained convolutional neural network was done on the fly during the MR session in approximately 9 ms for multiple hand drawn ROIs and the measured $B_1^+$ and $B_0$ maps. Compensation of $B_1^+$ inhomogeneity and $B_0$ off-resonances has been confirmed in the phantom and in vivo experiments. The reconstructed image data agrees well with the simulations using the acquired $B_1^+$ and $B_0$ maps and the 2D RF pulse predicted by the convolutional neural networks is as good as the conventional RF pulse obtained by optimal control.
Conclusion: The proposed convolutional neural network based 2D RF pulse design method predicts 2D RF pulses with an excellent excitation pattern and compensated $B_1^+$ and $B_0$ variations at 7 T. The rapid 2D RF pulse prediction (9 ms) enables subject-specific high-quality 2D RF pulses without the need to run lengthy optimizations.
△ Less
Submitted 25 September, 2020;
originally announced September 2020.
-
Ultra-fast (milliseconds), multi-dimensional RF pulse design with deep learning
Authors:
Mads Sloth Vinding,
Birk Skyum,
Ryan Sangill,
Torben Ellegaard Lund
Abstract:
Purpose: Some advanced RF pulses, like multi-dimensional RF pulses, are often long and require substantial computation time due to a number of constraints and requirements, sometimes hampering clinical use. However, the pulses offer opportunities of reduced-FOV imaging, regional flip-angle homogenization, and localized spectroscopy, e.g., of hyperpolarized metabolites. We propose a novel deep lear…
▽ More
Purpose: Some advanced RF pulses, like multi-dimensional RF pulses, are often long and require substantial computation time due to a number of constraints and requirements, sometimes hampering clinical use. However, the pulses offer opportunities of reduced-FOV imaging, regional flip-angle homogenization, and localized spectroscopy, e.g., of hyperpolarized metabolites. We propose a novel deep learning approach to ultra-fast design multi-dimensional RF pulses with intention of real-time pulse updates. Methods: The proposed neural network considers input maps of the desired excitation region of interest, and outputs a multi-dimensional RF pulse. The training library is retrieved from a large image database, and the target RF pulses trained upon are calculated with a method of choice. Results: A relatively simple neural network is enough to produce reliable 2D spatial-selective RF pulses of comparable performance to the teaching method. For binary regions of interest, the training library does not need to be vast, hence, re-establishment of the training library is not necessarily cumbersome. The predicted, single-channel pulses were tested numerically and experimentally at 3 T. Conclusion: We demonstrate a relatively effortless training of multi-dimensional RF pulses, based on non-MRI related inputs, but working in an MRI setting still. The prediction time of few milliseconds renders real-time updates of advanced RF pulses possible.
△ Less
Submitted 7 January, 2019; v1 submitted 6 November, 2018;
originally announced November 2018.
-
Cryogen-Free dissolution Dynamic Nuclear Polarization polarizer operating at 3.35 T, 6.70 T and 10.1 T
Authors:
Jan Henrik Ardenkjaer-Larsen,
Sean Bowen,
Jan Raagaard Petersen,
Oleksandr Rybalko,
Mads Sloth Vinding,
Marcus Ullisch,
Niels Chr. Nielsen
Abstract:
Purpose: A novel dissolution dynamic nuclear polarization (dDNP) polarizer platform is presented. The polarizer meets a number of key requirements for in vitro, pre-clinical and clinical applications. Method: It uses no liquid cryogens, operates in continuous mode, accommodates a wide range of sample sizes up to and including those required for human studies, and is fully automated. Results: It of…
▽ More
Purpose: A novel dissolution dynamic nuclear polarization (dDNP) polarizer platform is presented. The polarizer meets a number of key requirements for in vitro, pre-clinical and clinical applications. Method: It uses no liquid cryogens, operates in continuous mode, accommodates a wide range of sample sizes up to and including those required for human studies, and is fully automated. Results: It offers a wide operational window both in terms of magnetic field, up to 10.1 T, and temperature, from room temperature down to 1.3 K. The polarizer delivers a 13C liquid state polarization for [1-13C]pyruvate of 70%. The build-up time constant in the solid state is approx. 1200 s (20 min), allowing a sample throughput of at least one sample per hour including sample loading and dissolution. Conclusion: We confirm the previously reported strong field dependence in the range 3.35 to 6.7 T, but see no further increase in polarization when increasing the magnetic field strength to 10.1 T for [1-13C]pyruvate and trityl. Using a custom dry magnet, cold head and recondensing, closed-cycle cooling system, combined with a modular DNP probe, automation and fluid handling systems; we have designed a unique dDNP system with unrivalled flexibility and performance.
△ Less
Submitted 27 August, 2018; v1 submitted 14 August, 2018;
originally announced August 2018.
