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

Download:

Current browse context:

cs.CV

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

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

Electrical Engineering and Systems Science > Image and Video Processing

Title: Analysis of Deep Complex-Valued Convolutional Neural Networks for MRI Reconstruction

Abstract: Many real-world signal sources are complex-valued, having real and imaginary components. However, the vast majority of existing deep learning platforms and network architectures do not support the use of complex-valued data. MRI data is inherently complex-valued, so existing approaches discard the richer algebraic structure of the complex data. In this work, we investigate end-to-end complex-valued convolutional neural networks - specifically, for image reconstruction in lieu of two-channel real-valued networks. We apply this to magnetic resonance imaging reconstruction for the purpose of accelerating scan times and determine the performance of various promising complex-valued activation functions. We find that complex-valued CNNs with complex-valued convolutions provide superior reconstructions compared to real-valued convolutions with the same number of trainable parameters, over a variety of network architectures and datasets.
Subjects: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Medical Physics (physics.med-ph)
Cite as: arXiv:2004.01738 [eess.IV]
  (or arXiv:2004.01738v4 [eess.IV] for this version)

Submission history

From: Elizabeth Cole [view email]
[v1] Fri, 3 Apr 2020 19:00:23 GMT (1332kb)
[v2] Wed, 8 Apr 2020 00:41:29 GMT (1331kb)
[v3] Mon, 11 May 2020 04:02:38 GMT (1332kb)
[v4] Tue, 12 May 2020 01:21:36 GMT (1332kb)

Link back to: arXiv, form interface, contact.