Laura Nunez Gonzalez

Magnetic Resonance (MR) technology is in continuous development. Part of the ongoing research is focused on quantitative measurements of the tissue intrinsic parameters: proton density (PD), T1, and T2. Quantitative measurements could overcome inter-session and inter-system variability and facilitate longitudinal studies. However, until the recent development of fast multi-parametric sequences, such as MR Fingerprinting (MRF) or QRAPMASTER, the applicability of quantitative measurements was limited due to the long acquisition time needed.

The aim of this thesis is to gain insight into quantitative MR techniques and bring them closer to the clinical routine. As an initial step, we evaluated the accuracy and the repeatability of QRAPMASTER and MRF-vFA. This is explored in chapter 2 using one phantom and 5 volunteers. These techniques showed good repeatability in phantom and in-vivo experiments, but QRAPMASTER was more accurate than MRF-vFA. However, an advantage of MRF-vFA is the rapid acquisition time (70s, 4.77 times faster than QRAPMASTER). More investigation is needed to reduce the bias of MRF-vFA and improve resolution but its rapid acquisition time could position MRF-vFA as a good candidate at least as a pre-exploration sequence.

The positive results obtained in phantom and healthy volunteers, positioned us for the next step: to evaluate these techniques in patients. The goal of this second step is to identify differences in the quantitative values between healthy and tumoral tissue. For this purpose, we focused on patients with brain tumors, acquiring images using a customized version of QRAPMASTER (MAGiC - chapter 3) and PDQTI ("Accelerated 3D whole-brain T1, T2, and proton density mapping: feasibility for clinical glioma MR imaging" in Appendix-Publications) during the clinical protocol before injecting any contrast-agent.

In clinical routine, contrast media are administered to probe specific tissue properties, such as brain blood barrier leakage. In chapter 3, we evaluated if T1 and T2 values acquired prior to contrast admission could predict the regions that would enhance with contrast agent. Through receiver operating characteristic (ROC) and t-test analysis we showed that it is possible to differentiate the pre-contrast T1 and T2 values between tumoral tissue without T1-enhancement, tumoral tissue with T1-enhancement, and normal white matter. The accuracy of the classification should be improved, but these results encourage further exploration that could lead to substituting some of the conventional sequences to shorten total acquisition time or even the possibility of avoiding contrast-agent injection. The possibility of avoiding contrast-agent injections in MRI could be very relevant, not only because of the elimination of the burden and possible risk for the patients, but also in developing countries where the contrast-agent is not always available.

Although more research is needed with larger cohorts and different pathologies than those used in this thesis, these fast quantitative MR techniques showed good repeatability and sensitivity to disease.

Furthermore, in this thesis, we showed two technical innovations. First, in chapter 4 we showed the feasibility and repeatability of a multi-component analysis in highly undersampled MRF in-vivo data. We successfully obtained white matter, gray matter, CSF, and myelin water fraction maps. Further investigation on the accuracy of the in-vivo segmentation is needed, but the results of the Sparsity Promoting Iterative Joint Non-negative least squares algorithm applied to MRF data (SPIJN-MRF) were promising regarding parameter estimation and tissue-fraction maps segmentation.

The second innovation, presented in chapter 5, was a completely new approach and sequence for fast multi-parametric quantitative MR. We called this new sequence "Multiphase balance non Steady State Free Precession" (MP-b-nSSFP). We described analytically the transient-response of a repetitive sequence and used this to obtain intrinsic (PD, T1, T2) and extrinsic (B0, B1+) parameters with one acquisition. An optimized block of 4 pulses (30o, 175o, 30o, 175o) with phases 0o, 90o, 90o and 0o was repeatedly applied to complete a total of 100 pulses. The signal acquired during this train of pulses was used to estimate the intrinsic and extrinsic parameters. The results were accurate in phantom but it showed substantial underestimation of the T1 values for the in-vivo experiment. This could be due to different effects, such as magnetization transfer. To further validate this sequence, more experiments assessing the accuracy and the reproducibility should be carried out. Also, the sequence is still under investigation and the next steps are towards accelerated 3D acquisition.

In conclusion, the development of fast multi-parametric quantitative techniques improved the practical applicability of quantitative measurements, and with further investigation, they could be an improved tool over the conventional weighted MR images used in clinical imaging. This improvement in clinical routine could be the reduction of acquisition time of some standard scanning sessions by replacing several sequences with just one fast multi-parametric quantitative sequence, the possibility of depicting more substructures, or the amelioration of postprocessing pipelines, for example, avoiding registration (since all the images could be acquired with one single sequence).