A VFA scheme was empirically designed with Bloch equation simulations to minimize two-parameter fit errors with SASHA data, consisting of scaling the prescribed flip angle for the first 45 pulses by sin(x) for π/90 < × < π/2. The first 5 data acquisitions in the pulse train were discarded, matching the number of dummy pulses with linear catalyzation in the standard SASHA sequence. SASHA, SASHA-VFA, and MOLLI T
1 imaging was performed on 4 healthy volunteers (Siemens Aera 1.5T) on a mid-ventricular short-axis slice with typical bSSFP imaging readout parameters: 1.01/2.44 ms TE/TR, 8 mm slice thickness, 112 × 192 matrix size, 270 × 360 mm
2 field of view, rate 2 GRAPPA with 24 in-place ACS reference lines, 78% phase resolution, and 7/8 partial Fourier for a total imaging duration of ~175 ms. SASHA datasets were acquired with 9 images having equally spaced TIs from 165-780 ms following BIR-4 saturation, plus a non-saturated image. Standard SASHA was acquired with 5 (dummy) linear catalyzation pulses and SASHA-VFA was acquired with sinusoidal scaling described above, both with a target flip angle of 70°. MOLLI data was acquired with a 5-(3)-3 configuration, 120 ms TI start, 80 ms TI increment, 35° flip angle, and a tan/tanh adiabatic inversion pulse [
3]. T
1 pixel map were generated and the mean and standard deviation calculated for an ROI enclosing the entire LV myocardium.