Fast deep learning reconstruction techniques for preclinical magnetic resonance fingerprinting.

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.

1H-NMR Relaxation of Ferrite Core-Shell Nanoparticles: Evaluation of the Coating Effect.

We investigated the effect of different organic coatings on the 1H-NMR relaxation properties of ultra-small iron-oxide-based magnetic nanoparticles. The first set of nanoparticles, with a magnetic core diameter ds1 = 4.4 ± 0.7 nm, was coated with polyacrylic acid (PAA) and dimercaptosuccinic acid (DMSA), while the second set, ds2 = 8.9 ± 0.9 nm, was coated with aminopropylphosphonic acid (APPA) and DMSA. At fixed core diameters but different coatings, magnetization measurements revealed a similar behavior as a function of temperature and field. On the other hand, the 1H-NMR longitudinal r1 nuclear relaxivity in the frequency range ν = 10 kHz ÷ 300 MHz displayed, for the smallest particles (diameter ds1), an intensity and a frequency behavior dependent on the kind of coating, thus indicating different electronic spin dynamics. Conversely, no differences were found in the r1 relaxivity of the biggest particles (ds2) when the coating was changed. It is concluded that, when the surface to volume ratio, i.e., the surface to bulk spins ratio, increases (smallest nanoparticles), the spin dynamics change significantly, possibly due to the contribution of surface spin dynamics/topology.

Fast, interleaved, Look-Locker-based T1 mapping with a variable averaging approach: Towards temperature mapping at low magnetic field.

Proton resonance frequency shift (PRFS) is currently the gold standard method for magnetic resonance thermometry. However, the linearity between the temperature-dependent phase accumulation and the static magnetic field B0 confines its use to rather high-field scanners. Applications such as thermal therapies could naturally benefit from lower field MRI settings through leveraging increased accessibility, a lower physical and economical footprint, and further consideration of the technical challenges associated with the integration of heating systems into conventional clinical scanners. T 1 -based thermometry has been proposed as an alternative to the gold standard; however, because of longer acquisition times, it has found clinical use solely with adipose tissue where PRFS fails. At low field, the enhanced T 1 dispersion, combined with reduced relaxation times, make T 1 mapping an appealing candidate. Here, an interleaved Look-Locker-based T 1 mapping sequence was proposed for temperature quantification at 0.1 T. A variable averaging scheme was introduced, to maximize the signal-to-noise ratio throughout T 1 recovery. In calibrated samples, an average T 1 accuracy of 85% ± 4% was achieved in 10 min, compared with the 77% ± 7% obtained using a standard averaging scheme. Temperature maps between 29.0 and 41.7°C were eventually reconstructed, with a precision of 3.0 ± 1.1°C and an accuracy of 1.5 ± 1.0°C. Accounting for longer thermal treatments and less strict temperature constraints, applications such as MR-guided mild hyperthermia treatments at low field could be envisioned.

A method for T1 and T2 relaxation times validation and harmonization as a support to MRI mapping.

We present a proof-of-concept study focusing on a method for the intra- and inter-center validation and harmonization of data obtained from MRI T1 and T2 maps. The method is based on a set of MnCl2 samples that provide in-scan ground-truth reference values regardless of the details of the MRI protocol. The relaxation times of MnCl2 aqueous solutions were first measured by means of an NMR laboratory relaxometer, as a function of concentration and temperature. The obtained T1 and T2 values, once renormalized at the scanner temperature, were used as reference values for the MRI mapping measurements of the MnCl2 relaxation times. By using different clinical MRI scanners and sequences, we found a good agreement for standard and turbo sequences (limits of agreement: 5% for IR, SE, IR-TSE; 10% for TSE), while an under-estimation and an over-estimation were found respectively for MOLLI and T2-prep TrueFISP, as already reported in the literature. The linearity of the relaxation rates with the concentration predicted by the Solomon-Bloembergen-Morgan theory was observed for every dataset at all temperatures, except for T2-prep TrueFISP maps results. Some preliminary results of an in vivo experiment are also presented.

Longitudinal and transverse NMR relaxivities of Ln(III)-DOTA complexes: A comprehensive investigation.

Longitudinal and transverse 1H nuclear magnetic resonance relaxivities of Ln(III)-DOTA complexes (with Ln = Gd, Tb, Dy, Er; DOTA = 1,4,7,10-tetraazacyclododecane-N,N',N″,N‴-tetraacetic acid) and Mn(II) aqueous solutions were measured in a wide range of frequencies, 10 kHz to 700 MHz. The experimental data were interpreted by means of models derived from the Solomon-Bloembergen-Morgan theory. The data analysis was performed assuming the orbital angular momentum L = 0 for Gd-DOTA and the aqua ion [Mn(H2O)6]2+ and L ≠ 0 for Dy-, Tb-, and Er-DOTA. A refined estimation of the zero-field-splitting barrier Δ and of the modulation correlation time τv was obtained for [Mn(H2O)6]2+ by extending the fitting of nuclear magnetic relaxation dispersion profiles to the low-field regime. The Gd-DOTA fitting parameters resulted in good agreement with the literature, and the fit of transverse relaxivity data confirmed the negligibility of the scalar interaction in the nuclear relaxation mechanism. Larger transverse relaxivities of Dy-DOTA and Tb-DOTA (∼10 mM-1 s-1) with respect to Er-DOTA (∼1 mM-1 s-1) were observed at 16 T. Such higher values are suggested to be due to a shorter residence time τm that is possibly linked to the fluctuations of the hyperfine interaction and the different shape of the magnetic anisotropy. The possible employment of Dy-DOTA, Tb-DOTA, and Er-DOTA as negative magnetic resonance imaging contrast agents for high-field applications was envisaged by collecting spin-echo images at 7 T. Particularly in Dy- and Tb-derivatives, the transverse relaxivity at 16 T is of the order of the Gd-one at 1.5 T.

Nanosized T1 MRI Contrast Agent Based on a Polyamidoamine as Multidentate Gd Ligand.

A linear polyamidoamine (PAA) named BAC-EDDS, containing metal chelating repeat units composed of two tert-amines and four carboxylic groups, has been prepared by the aza-Michael polyaddition of ethylendiaminodisuccinic (EDDS) with 2,2-bis(acrylamido)acetic acid (BAC). It was characterized by size exclusion chromatography (SEC), FTIR, UV-Vis and NMR spectroscopies. The pKa values of the ionizable groups of the repeat unit were estimated by potentiometric titration, using a purposely synthesized molecular ligand (Agly-EDDS) mimicking the structure of the BAC-EDDS repeat unit. Dynamic light scattering (DLS) and ζ-potential analyses revealed the propensity of BAC-EDDS to form stable nanoaggregates with a diameter of approximately 150 nm at pH 5 and a net negative charge at physiological pH, in line with an isoelectric point <2. BAC-EDDS stably chelated Gd (III) ions with a molar ratio of 0.5:1 Gd (III)/repeat unit. The stability constant of the molecular model Gd-Agly-EDDS (log K = 17.43) was determined as well, by simulating the potentiometric titration through the use of Hyperquad software. In order to comprehend the efficiency of Gd-BAC-EDDS in contrasting magnetic resonance images, the nuclear longitudinal (r1) and transverse (r2) relaxivities as a function of the externally applied static magnetic field were investigated and compared to the ones of commercial contrast agents. Furthermore, a model derived from the Solomon-Bloembergen-Morgan theory for the field dependence of the NMR relaxivity curves was applied and allowed us to evaluate the rotational correlation time of the complex (τ = 0.66 ns). This relatively high value is due to the dimensions of Gd-BAC-EDDS, and the associated rotational motion causes a peak in the longitudinal relaxivity at ca. 75 MHz, which is close to the frequencies used in clinics. The good performances of Gd-BAC-EDDS as a contrast agent were also confirmed through in vitro magnetic resonance imaging experiments with a 0.2 T magnetic field.