T2 mapping is becoming increasingly important for tissue characterization in patients with myocardial pathologies (e.g. myocardial edema)¹. However, clinical T2 mapping has been typically limited to 2D images acquired during breath-holding (BH) resulting in low resolution images, mis-registration in-between BH or motion during BH. Free-breathing 3D T2 mapping has been proposed using respiratory motion correction by means of conventional diaphragmatic 1D navigation (d1D NAV), but the scan time is unpredictable². 1D self-navigated 3D T2 mapping has been proposed³; this approach has, however, the disadvantage that the 1D self-gating signal also contains contributions from static tissue (e.g. chest wall). Recently, image-based navigation (iNAV)⁴ has emerged as an alternative approach for respiratory motion correction, and its accuracy demonstrated in coronary MR angiography. The iNAV allows for direct respiratory motion estimation of the heart with 100% scan efficiency. The purpose of this study was to implement and evaluate, for the first time, iNAV motion correction for 3D T2 mapping on a 3T clinical scanner.An ECG-triggered pulse sequence with interleaved T2prep acquisitions preceded by an iNAV for motion estimation and a saturation pulse to reset the magnetization for every heartbeat was implemented, as shown in Figure 1. To generate T2 maps, multiple ECG-triggered mid-diastolic images were acquired with varying degrees of T2 weighting using T2prep pulses with three echo times (TE = 0, 26 and 46 ms). Two-dimensional iNAV were acquired for each k-space segment by spatially encoding 8 startup pulses of a balanced steady state free precession (SSFP) image acquisition. With this approach, translational respiratory induced motion was directly measured on the heart and corrected for in foot-head and left-right directions with 100% scan efficiency. A mono-exponential function was fitted on a pixel-by-pixel basis to the series of images with incremental T2prep echo time to generate a T2 map. T2 maps were acquired in 13 healthy subjects (7 females, 6 males, 29 ± 6 years) using a 3T clinical scanner (Achieva, Philips Healthcare, The Netherlands) covering the whole heart with 2mm isotropic resolution. For comparison, a 3D T2 prep-based and d1D NAV corrected and a 2D BH T2 prep-based T2 mapping sequence were acquired in the short-axis view.The mean ± standard deviation (SD) T2 values of the myocardium were 45.7 ± 5.7 ms, using the proposed free-breathing 3D T2-prep approach using the iNAV. The T2 values did not display significant differences (p < 0.05) compared to those obtained with the 3D T2prep d1D NAV corrected (47.1 ± 8.9 ms) and the BH 2D T2prep (46.1 ± 6.3 ms) T2 mapping sequence. The images acquired with the proposed method are shown in Figure 2, including a short-axis 3D T2prep d1D NAV and 2D BH T2prep T2 map. The mean ± SD scan time of the 3D T2prep iNAV was 4:56 ± 1:50 minutes for the healthy subjects while it was 14:20 ± 3:00 minutes for the 3D T2prep d1D NAV T2 mapping sequence.The proposed image navigator based free-breathing high-resolution whole heart 3D T2 mapping approach is feasible and can be performed within less than 5 minutes with similar accuracy to that of the lower-resolution 2D-BH T2 mapping approach. These promising results warrant further investigation in patients with acute myocardial infarction, myocarditis or heart transplantation.