Figure 3

Appreciation of the quality of SR reconstruction depending on the weight \(\lambda\) that controls the strength of the TV regularization. The potential of our framework FaBiAN for optimizing the reconstruction quality through parameter fine-tuning in the presence of motion is illustrated at two GA: 26 and 30 weeks. Two representative clinical cases are provided for comparison. The results for three values of \(\lambda\) are presented. For \(\lambda = 0.1\), the SR reconstruction looks blurry with poor tissue contrast. Using \(\lambda = 3\) improves the contrast but the images look noisy. For \(\lambda = 0.75\), the SR reconstruction is sharp with a contrast between different brain tissues similar to that observed in the 3D isotropic ground truth. Clinical cases from which the simulated HASTE images are derived highlight the accuracy of a SR reconstruction for this intermediate value of \(\lambda\), especially with regards to the definition of the corpus callosum and the delineation of the cortex.