Accelerated T₂ Mapping Based On Bloch Signal-Model With Fixed Rank and Sparsity Constraints

Background: Quantification of T₂ values is valuable for a wide range of research applications. The Multi-Echo-Spin-Echo (MESE) protocol offers significantly shorter scan times than the Single Spin Echo protocol, at the cost of strong contamination from stimulated echoes, limitation which the Echo Modulation Curve (EMC) can efficiently overcome. Further acceleration can be achieved by undersampling the k-space domain and using designated reconstruction algorithms. In this work we use fixed rank and sparsity constraints along with EMC to achieve higher acceleration with accurate quantitative T₂ mapping results.

Methods: Brain and calf datasets were scanned from healthy volunteers using standard MESE protocol. Data were retrospectively undersampled by acceleration factors of 2 to 6 and reconstructed by solving the following optimization problem: . T₂ maps where generated using the EMC algorithm. Relative errors were calculated with respect to the fully sampled results and analyzed by means of their mean ± standard deviation (SD). Our results were compared to those of the widely used GRAPPA reconstruction method.

Results: Our results at acceleration factors 2 to 6 are: 2.6±3.5, 3.3±7.6, 4.1±9.2, 4.8±15.1, 5.0±15.3 respectively for the brain dataset, and 4.0±11.3, 5.0±15.3, 6.2±19.3, 6.6±19.8, 7.6±16.9 respectively for the calf dataset. For comparison, GRAPPA’s errors are: 1.9±3.9, 3.5±4.8, 7.7±9.4, 18.9±25.1, 29.3±31.1 respectively for the brain dataset, and 5.3±23.1, 13.7±35.4, 28.6±51.2, 38.6±65.5, 45.6±70.4 respectively for the calf dataset.

Conclusion: Our method achieves accurate and precise T₂ maps for acceleration factors of 2 to 6, while GRAPPA fails in this task for acceleration factors of 4 to 6.