Theory-based single-point T1 mapping for quantitative analysis of first-pass cardiac perfusion MRI: a validation study

Introduction: Quantitative analysis of first-pass contrast-enhanced cardiac perfusion MRI requires a conversion of the T1-weighted signal-time curve to contrast agent (Gd-DTPA) concentration-time curve. For this purpose, a theory-based single-point T1 measurement method has been proposed and validated in phantoms at 1.5T [1-2]. The specific choice of the k-space ordering of the TurboFLASH readout can modulate the effect of B1 variation, particularly at 3T, on the accuracy of the T1 measurement [3]. In this work, we i) investigated the sensitivity to B1 variation of the single-point T1 mapping method, depending on its k-space acquisition order (linear or centric) and ii) validated in vivo the single-point T1 mapping method in the left ventricle (LV) of the cavity and myocardium during contrast enhancement. Methods: A saturation-recovery TurboFLASH sequence was implemented on a 3T whole-body MR scanner (Siemens; Tim Trio). Images were acquired with the following parameters: FOV = 320mm × 262mm, slice thickness = 8mm, matrix = 144 × 94, TE/TR = 1.24/2.4 ms, flip angle = 10o, TSENSE with acceleration factor 2, and effective saturation pulse [4]. The saturation-recovery time delay (TD) was 10/50 ms and total image acquisition time = 136/176 ms for the linear and centric k-space trajectories, respectively. The effective longitudinal magnetization at the center of k-space was calculated as a function of T1, using the Bloch equation, as previously described [1]. A proton density-weighted image was acquired in the first heartbeat, with flip angle = 5o and without the saturation pulse, in order to normalize the image signal. A theoretical relationship was thus obtained between the normalized signal and T1. TDs were chosen so that linear and centric sequences had a similar signal-T1 theoretical relationship. Note that for the pulse sequence parameters used in this study, the normalized signal for T1>1000ms is less than 5% of equilibrium magnetization, indicating that the signal-to-noise ratio (SNR) at pre-contrast is inadequate for the single-point T1 method. Therefore, pre-contrast T1 measurements were performed using a multi-point T1 mapping sequence to ensure accurate T1 measurement. Reference T1 measurements were performed at nominal B1 calibration with a multi-point saturation recovery TurboFLASH sequence with variable TD and a centric k-space trajectory. A least square linear regression was used to fit the 10 points on the saturation recovery curve. First, to investigate the effect of B1 variation, 6 dilutions of Gd-DTPA (Magnevist) in water (T1 [48–1100] ms) were imaged for each k-space trajectory (linear or centric) with the excitation flip angle ranging from 40 to 120 % of nominal 10°. Second, the accuracy of the single-point T1 measurements using the optimal centric k-space trajectory, as determined by the phantom experiment, was assessed in vivo in 4 healthy volunteers (35±12 years old). A basal short-axis plane of the heart was imaged at 9 time points with the single-point and reference multi-point acquisitions: pre-contrast, 5, 10, 15, and 20 min post first injection (0.1mmol/kg) of Gd-DTPA, and 5, 10, 15, and 20 min post a second injection of GdDTPA. Reference T1s were fitted from 7 data points: no saturation and TDs from 100-600 ms (100ms steps). Contours for the myocardium and left ventricular (LV) cavity were drawn manually (Fig.2.A). Measured T1s were converted to Gd-DTPA-concentrations ([Gd]) with the following equation: [Gd] = (1/T1-1/T1) / kGd, assuming fast water exchange condition [5] and T1 relaxivity (kGd) of 3.8L/mmol/s [6-7], and T1 measurement with the multipoint fit to ensure its accuracy. Results: Compared with a linear k-space trajectory, a centric k-space trajectory was found to be less sensitive to clinically relevant B1 variation at 3T in phantoms (Fig.1). Consequently, the centric k-space trajectory was validated in vivo. A good linear correlation was found between the reference and single-point T1 measurements in the LV cavity and wall (Fig.2.B, R=0.97, p<0.001), as well as between the corresponding GdDTPA concentrations ([Gd]) (Fig.2.C, R=0.98, p<0.001). Bland-Altman analyses showed an increasing error in the single-point measurements for longer T1 values, likely due to the correspondingly lower SNR. The underestimation observed for long T1 is also likely related to apparent signal increase due to Rician noise bias. Conclusion: We have validated the single-point T1 mapping sequence against the multi-point T1 mapping sequence in vivo. Compared with a linear k-space trajectory, disadvantages of a centric k-space trajectory are that it requires a longer TD to achieve adequate SNR, and it yields slight edge enhancement, due to high-pass filtering effects in the kspace. Comparative advantages of the centric k-space trajectory are that it is relatively insensitive to B1 inhomogeneities, inflow effects, and refocusing of residual transversal magnetization from imperfect radio-frequency spoiling. Therefore, the modeling of the signal is simpler for the centric k-space ordering than for the linear k-space ordering. The single-point T1 mapping pulse sequence with centric k-space trajectory is thus promising for quantitative analysis of first-pass cardiac perfusion MRI.