The effect of acquired ShMOLLI field of view on myocardial T1 values in UK Biobank

Background: The UK Biobank (UKB) project has acquired CMR images in 100,000 participants across four imaging centres in the UK. One component of the protocol is short-axis cardiac T₁ mapping with the Shortened Modified Look-Locker Inversion recovery (ShMOLLI) sequence[1]. Over the project the ShMOLLI acquisition protocol has remained unchanged to ensure data consistency. Despite this, some parameters do change on a per-participant basis due to the physiological/anatomical differences between participants. This includes the TI times (which are dependent on RR interval), scan planning, and automatic scanner adjustments e.g. transmit power.

The UKB ShMOLLI protocol uses the Siemens DOT engine[2] to plan the ShMOLLI acquisition, to maximize consistency between participants[3]. This includes minimising FOV in the phase-encode (PE) direction, subject to chest size, while maintaining a constant spatial resolution (Figure 1). This minimises the required number of readout lines on a per-participant basis, reducing readout time and cardiac-motion blur. However, this also introduces variability in the degree of readout saturation between participants, which could impact T₁.

In this abstract we use Bloch simulations and phantom experiments to quantify the FOV dependence of T₁ in the UKB cardiac protocol.

Methods: A full Bloch simulation of the ShMOLLI acquisition scheme was implemented to match the ShMOLLI sequence as implemented in the UKB imaging study[3]. Three T₁ values were chosen (1004,1281,1506ms) to match the T₁ measured in the native myocardium tubes of a T1MES[4] phantom with a slow IR sequence. With an interval RR=900 ms (67 bpm), PE direction FOVs of 50-100% readout FOV (% PE FOV) were simulated and a linear fit of T₁ vs FOV calculated for each tube.

Phantom experiments were performed at the UKB Bristol imaging centre using a 1.5T MRI scanner (Magnetom Aera, Siemens Healthineers) with the UKB cardiac protocol. A simulated RR=900ms was used for all experiments.

Results: Plots of the results are shown in Figures 2. The Bloch simulations gave reconstructed T₁ for each tube of 877, 1114, 1306ms with 100% PE FOV. The slopes of the linear fits of T₁ to PE FOV were -0.48, -0.64, -0.77 ms/%FOV.

In the phantom experiments the tubes had T₁ values of 877.6±5.7, 1125.4±9.7, 1323.4±11.1ms (mean±SD respectively with 100% PE FOV. The slopes of the linear fits of T₁ to PE FOV were -0.37, -0.44, -0.57 ms/%FOV.

Conclusion: Acquired field of view in PE direction, and therefore chest size, is a confounding factor in the UKB study ShMOLLI T₁ mapping results. Previous studies have noted a significant correlation between BMI (which is expected to be correlated to chest size) and myocardial T₁ in the UKB dataset[5], suggesting that this effect is seen in the imaging study results. Further experiments in vivo, such as additional % PE FOV measurements in a subset of the UKB repeat imaging cohort, will be necessary to fully quantify this effect and allow correction of myocardial T₁ for PE FOV in the UKB dataset.

Figure 1A: example ShMOLLI baseline images showing how the FOV in the phase encode direction is adjusted as a percentage of FOV in the frequency encode direction to minimize the readout duration required. Figure 1B: Simulated evolution of the longitudinal magnetization during the ShMOLLI acquisition (T1=1281ms) demonstrating the impacts of each FOV shown in Figure 1A.
Figure 2A: Plot of T1 against % PE FOV for each of the three simulated T1MES tubes. Figure 2B: Plot of T1 against % PE FOV for each of the three native T1MES tubes measured with the UKB ShMOLLI protocol.