Current cardiac T1 mapping techniques in use clinically are generally limited to single-shot 2D images acquired in a breath hold with ECG gating which implies the need for a regular heart rhythm and reliable breath-holding; both of which are potential causes for reduced accuracy and reproducibility of the T1 maps in clinical practice.^1-4 Heart rate variability or poor ECG triggering has been identified as a major source of error and cause for reduced reproducibility of myocardial T1 maps in the widely used MOLLI T1 mapping technique.^4,5 The resolution of other inversion recovery and saturation recovery 2D single-shot techniques is limited by the acquisition window, especially for subjects with relatively high heart rates; higher resolution requires segmentation with multiple breath holds and potential image mis-registration. To mitigate the dependence of T1 mapping on heart rate and breath-holds, we propose an ungated, free-breathing, continuous inversion recovery approach using low-rank tensors^6,7 modeling the image as partially separable in space, cardiac phase, respiratory phase, and inversion time in order to reduce sampling requirements.All imaging was performed on a 3T Siemens Verio scanner. The proposed sequence uses an ungated, free-breathing, 2D continuous modified golden angle radial acquisition scheme (odd readouts were incremented by the golden angle, even readouts were a 0° navigator readout used for cardiac/respiratory binning and subspace estimation) with 180° inversion pulses every 2.5 seconds, 5° flip angle, echo spacing 3.6ms, resolution 1.7x1.7x8mm³ in a mid-ventricular slice with acquisition time: 58 seconds. The data was reconstructed using an explicit tensor subspace constraint⁷ estimated from the navigator data and from a dictionary of signal curves generated from the Bloch equations to obtain 345 inversion time (TI) images for 15 cardiac phases and 5 respiratory phases. Pixel-wise T1 maps were computed by nonlinear least-squares regression of the resulting TI images from cardiac bins for systole and diastole from a respiratory bin representing end-expiration. Additionally, a MOLLI 5(3)3 T1 map was acquired with the same resolution in diastole and systole in an end-expiration breath hold in a single mid-ventricular slice. T1 was measured by drawing a region-of-interest (ROI) in the septal region on the T1 maps. For the MOLLI 5(3)3 the blood pool T1 was measured from the T1* map which experiences no Look-Locker correction.The scan time for the proposed ungated, free-breathing method was exactly 58 seconds while the MOLLI method required two separate breath hold scans for systole and diastole with a wait time between breath holds. T1 maps from the proposed ungated, free-breathing method and MOLLI 5(3)3 from two healthy subjects (1 female, age 26; 1 male, age 53) are shown in the figure. The T1 values for the septal myocardium from the proposed method and MOLLI 5(3)3 for both subjects are shown in the table. Diastolic native T1 values are higher than systolic values consistent with published data.⁸ Sequence differences (FLASH vs SSFP) may potentially explain differences in myocardial T1 values. Blood pool T1 values for the proposed method are higher due to blood inflow effects with a slice selective readout.We have shown a proof-of-concept continuous IR T1 mapping technique from which T1 maps can be obtained in different cardiac and respiratory phases without ECG gating or breath holds. The proposed method shows promise as a fast T1 mapping technique with no dependence on heart rate or breath holds.