Accurate T1 Quantification using the Variable Flip Angle Method

Sammanfattning

MRI is widely used qualitatively to visualize contrast between tissues and pathologies. Although essential in diagnostics, qualitative MRI has limitations, which motivates the need for quantitative MRI (qMRI). Quantitative MRI converts conventional images into parameter maps, such as T1 relaxation time maps, which are used for evaluation of changes in tissue and for monitoring treatment response. Several methods have been developed for generating T1 maps. One commonly used approach in many studies is the Variable flip angle (VFA) method, but its accuracy is affected by several factors. In this method, multiple images are acquired at different flip angles while the repetition time (TR) is kept constant. Specifically, a series of spoiled gradient-recalled echo (SPGR) images are acquired using a set of various flip angles, from which T1 maps are subsequently derived. This thesis aims to investigate potential sources of error that affect the accuracy of VFA-based T₁ mapping, by systematically evaluating how acquisition parameters and analysis choices influence T1 estimation accuracy. To investigate this, a calibrated MR phantom (Mini Hybrid Phantom model 137, CaliberMRI) was scanned on two GE Healthcare scanners, Signa PET-MR and Signa Premier, using various measurement protocols. The vendor implemented SPGR sequence and a custom SPGR sequence developed at Karolinska Hospital were examined. Acquisition parameters such as flip angles, repetition time, spoiling parameter, and B1 maps were varied in order to assess their influence. The acquired data were analyzed using nonlinear least-squares fitting to estimate T1, and the bias of T1 estimations was determined by comparing estimates to reference values provided with the phantom. The results indicate that T1 estimation accuracy depends on repetition time (TR), scanner type, and the applied B1 correction. The accuracy shows less dependence on the gradient spoiling area. At both TR = 5.8 ms and TR = 10 ms, both scanners show large errors for the Region Of Interests (ROIs) with long T1 s, and smaller errors for ROIs with short T1. However, at TR = 5.8 ms, the T1 estimates show less bias and a narrower error spread around the zero-error reference line. B1 correction further reduces the error distribution, particularly when using low flip angle B1 maps. In conclusion, T1 estimation accuracy showed strong dependence on repetition time, increasing the TR reduced the accuracy of T1 estimates. Furthermore, investigation of spoiling parameters using the custom SPGR sequence showed that the gradient spoiling strength had only a minor effect on T1 estimation accuracy. Finally, applying B1 correction reduced error dispersion, and the most accurate T1 estimates were achieved when using low flip angle B1 maps.