We propose a deep learning (DL) model and a hyperparameter optimization strategy to reconstruct T1 and T2 maps acquired with the magnetic resonance fingerprinting (MRF) methodology. We applied two different MRF sequence routines to acquire images of ex vivo rat brain phantoms using a 7‐T preclinical scanner. Subsequently, the DL model was trained using experimental data, completely excluding the use of any theoretical MRI signal simulator. The best combination of the DL parameters was implemented by an automatic hyperparameter optimization strategy, whose key aspect is to include all the parameters to the fit, allowing the simultaneous optimization of the neural network architecture, the structure of the DL model, and the supervised learning algorithm. By comparing the reconstruction performances of the DL technique with those achieved from the traditional dictionary‐based method on an independent dataset, the DL approach was shown to reduce the mean percentage relative error by a factor of 3 for T1 and by a factor of 2 for T2, and to improve the computational time by at least a factor of 37. Furthermore, the proposed DL method enables maintaining comparable reconstruction performance, even with a lower number of MRF images and a reduced k‐space sampling percentage, with respect to the dictionary‐based method. Our results suggest that the proposed DL methodology may offer an improvement in reconstruction accuracy, as well as speeding up MRF for preclinical, and in prospective clinical, investigations.
10 November 2014 in Physical Chemistry Chemical Physics
In pyruvic acid containing 15 mM trityl below 4 K 13 C polarization and spin–lattice relaxation rates are proportional to the spin–lattice relaxation rate of electrons, suggesting an efficient thermal mixing scenario.
NMR and magnetization measurements in , recently proposed as a strongly correlated metal, are presented. Two different low-frequency dynamics are evidenced. The first one, probed by nuclei, gives rise to a slowly relaxing magnetization at low temperature and is associated with the freezing of spins. This dynamics is similar to the one observed in pristine and originates from Li-depleted chain segments. The second one, evidenced by the spin-lattice relaxation rate, is associated with the hopping of the electrons along Li-rich chains. The characteristic correlation times for the two dynamics are derived, and the role of the disorder is briefly discussed.
1H nuclear spin–lattice relaxation and Dynamic Nuclear Polarization (DNP) have been studied in amorphous samples of trehalose sugar doped with TEMPO radicals by means of mechanical milling, in the 1.6–4.2 K temperature range. The radical concentration was varied between 0.34 and 0.81%. The highest polarization of 15% at 1.6 K, observed in the sample with concentration 0.50%, is of the same order of magnitude of that reported in standard frozen solutions with TEMPO. The temperature and concentration dependence of the spin–lattice relaxation rate 1/T1, dominated by the coupling with the electron spins, were found to follow power laws with an exponent close to 3 in all samples. The observed proportionality between 1/T1 and the polarization rate 1/Tpol, with a coefficient related to the electron polarization, is consistent with the presence of Thermal Mixing (TM) and a good contact between the nuclear and the electron spins. At high electron concentration additional relaxation channels causing a decrease in the nuclear polarization must be considered. These results provide further support for a more extensive use of amorphous DNP-ready samples, obtained by means of comilling, in dissolution DNP experiments and possibly for in vivo metabolic imaging.