3764

View Abstract / View Presentation

Learning how to Clean Fingerprints -- Deep Learning based Separated Artefact Reduction and Regression for MR Fingerprinting

Yiling Xu¹, Elisabeth Hoppe¹, Peter Speier², Thomas Kluge², Mathias Nittka², Gregor Körzdörfer², and Andreas Maier¹
¹Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany, ²Magnetic Resonance, Siemens Healthcare, Erlangen, Germany

We propose a deep learning two-step approach for MRF reconstruction. 1) Artefacts in the fingerprints are reduced with a U-net model. 2) The cleaned fingerprints are used as inputs for a regression network. Performance of our method is shown superior to a direct regression without the first step.

Figure 3: Quantitative maps from test scan

Row 1: Ground truth maps Row 2: Reconstructed maps without prior artefact reduction and relative error maps Row 3: Reconstructed maps from cleaned fingerprints and relative error maps A clear improvement in the reconstruction quality can be observed using a prior DL-based artefact reduction approach.

Figure 1: U-net architecture for the artefact reduction

As input for our U-net-based artefact reduction we use the main 50 complex-valued components of SVD-compressed fingerprints for each slice, resulting in 100 input channels for real and imaginary parts. The network outputs are the cleaned fingerprints of the same size as the input. For the training, the Mean-Squared-Error of the compressed signals from the dictionary matches and the network outputs is used.