We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

physics.med-ph

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Physics > Medical Physics

Title: Clipped DeepControl: deep neural network two-dimensional pulse design with an amplitude constraint layer

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 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.
Subjects: Medical Physics (physics.med-ph); Machine Learning (cs.LG); Systems and Control (eess.SY)
Cite as: arXiv:2201.08668 [physics.med-ph]
  (or arXiv:2201.08668v1 [physics.med-ph] for this version)

Submission history

From: Mads Vinding [view email]
[v1] Fri, 21 Jan 2022 12:34:36 GMT (3713kb)

Link back to: arXiv, form interface, contact.