Quantitative MR Image Reconstruction using Parameter-Specific Dictionary Learning with Adaptive Dictionary-Size and Sparsity-Level Choice

Andreas Kofler (PTB) published together with Kirsten Kerkering (PTB), Laura Göschel (Charité-Unviersitätsmedizin Berlin), Ariane Fillmer (PTB) and Christoph Kolbitsch (PTB) an article in IEEE Transactions on Biomedical Engineering.

The article describes a new method for obtaining quantitative MR parameters using adaptive dictionary learning and sparse coding. Instead of applying dictionary learning for reconstructing the qualitative images first and then performing a fit to obtain the quantitative parameters, the approach directly uses the dictionary-based regularization on the sought quantitative parameters. By doing so, the reconstruction is accelerated by a factor of approximately seven. In addition, different dictionaries for each of the different parameters are used and hyper-parameters for the dictionary and the sparse coding area adaptively adjusted depending on the considered data.

In the article the method is to a T1-parameter mapping in the brain and compared to several other methods using sparsity-based regularization.

The accepted article is available at ieee.org/document/10209588.

Contact:

Andreas Kofler, E-Mail: 📧 andreas.kofler(at)ptb.de