Help Privacy Policy Disclaimer
  Advanced SearchBrowse




Journal Article

Image denoising for real-time MRI.


Klosowski,  J.
Biomedical NMR Research GmbH, MPI for biophysical chemistry, Max Planck Society;


Frahm,  J.
Biomedical NMR Research GmbH, MPI for biophysical chemistry, Max Planck Society;

External Resource
No external resources are shared
Fulltext (restricted access)
There are currently no full texts shared for your IP range.
Fulltext (public)
There are no public fulltexts stored in PuRe
Supplementary Material (public)

(Supplementary material), 2MB


Klosowski, J., & Frahm, J. (2017). Image denoising for real-time MRI. Magnetic Resonance in Medicine, 77(3), 1340-1352. doi:10.1002/mrm.26205.

Cite as: https://hdl.handle.net/11858/00-001M-0000-002B-8421-B
Purpose To develop an image noise filter suitable for MRI in real time (acquisition and display), which preserves small isolated details and efficiently removes background noise without introducing blur, smearing, or patch artifacts. Theory and Methods The proposed method extends the nonlocal means algorithm to adapt the influence of the original pixel value according to a simple measure for patch regularity. Detail preservation is improved by a compactly supported weighting kernel that closely approximates the commonly used exponential weight, while an oracle step ensures efficient background noise removal. Denoising experiments were conducted on real-time images of healthy subjects reconstructed by regularized nonlinear inversion from radial acquisitions with pronounced undersampling. Results The filter leads to a signal-to-noise ratio (SNR) improvement of at least 60% without noticeable artifacts or loss of detail. The method visually compares to more complex state-of-the-art filters as the block-matching three-dimensional filter and in certain cases better matches the underlying noise model. Acceleration of the computation to more than 100 complex frames per second using graphics processing units is straightforward. Conclusion The sensitivity of nonlocal means to small details can be significantly increased by the simple strategies presented here, which allows partial restoration of SNR in iteratively reconstructed images without introducing a noticeable time delay or image artifacts.