Cardiac magnetic resonance T1 and extracellular volume mapping with motion correction and co-registration based on fast elastic image registration

Fig. 1. Workflow of T1 and extracellular volume (ECV) mapping combined with motion correction and co-registration using the proposed approach. MoCo motion correction, Co-Reg co-registration. For details, see text

Fig. 2. Motion correction for T1 mapping with quality improvement. Short-axis mid native (a) and basal post-contrast (b) ventricular T1 (top row) and fitting error (bottom row) maps of one patient with dilated cardiomyopathy (DCM) before (left column) and after (right column) motion correction. Myocardial structure showed better alignment compared with obvious fitting error at the septal (a) and septal-to-anterior (b) subendocardial segments without motion correction (white arrow)

Fig. 3. Motion correction for T1 mapping without quality improvement (from poor to poor). Short-axis mid ventricular native T1 (top row) and fitting-error (middle row) maps of one patient with ischemic heart disease (IHD) before (left column) and after (right column) motion correction. Selected modified look-locker inversion recovery (MOLLI) images (4 out of 8) showed both great motion (dotted lines) and extensive infarct, which was consistent with the late gadolinium enhancement (LGE) image (arrowheads). Two out of 82 cases remained of poor quality with motion correction and were inadequate for contouring and were thus excluded for quantitative analysis. See text for details

Fig. 4. Image co-registration for extracellular volume (ECV) mapping with (a) and without (b) quality improvement. Short-axis basal ventricular ECV (top row) and native and post-contrast T1 overlay (second row) maps before (left column) and after (right column) image co-registration (Co-Reg) in two patients (a, b) with ischemic heart disease. The individual native and post-contrast T1 maps are shown side-by-side (last row), respectively. In (a), obvious misalignment seen in the T1 overlap map (white arrows) and individual native and post-contrast maps (dotted lines) was corrected with co-registration that yielded a more homogenous ECV map. In (b), image co-registration did not improve interpreter-assessed quality of the ECV map. Distinct motion status (dotted lines) with myocardial infarct involving the anterior and anteroseptal walls was appreciated in greater relief in the post-contrast compared with native T1 map (last row)

Fig. 5. Image co-registration for extracellular volume (ECV) mapping with quality maintained. Short-axis mid ventricular T1 maps (top row), local deformation field (LDF) map (bottom left), and volume change map (bottom right) in two patients (a, b) with ischemic heart disease. For co-registration, the native T1 map (top left) served as reference and the post-contrast T1 map (top right) as target; the final map is displayed (top middle). In both cases, only small and smooth deformation were found in both the LDF checkerboard by slightly deviated vertical or horizontal lines (arrows) and local volume change (LVC) map (≤ 20%). See text for details

Fig. 6. Intra- (a) and inter-observer (b) reproducibility of extracellular volume fraction (ECV) measurements without and with motion correction and co-registration. Bland–Altman plots of measured ECV values at the mid- and basal levels (n = 82 cases), with 95% limits of agreement (LOA) superimposed on the chart for without motion correction or co-registration (w/o MoCo, solid lines), with motion correction only (w/MoCo, dashed line), and with both motion correction and co-registration (w/MoCo and CoReg, dash–dotted line). Improved agreement was found using the proposed method, as shown by the decreased differences and tighter 95% LOA