(P124) COMPARISON OF POST-PROCESSING TECHNIQUES FOR OPTIMIZING THE QUALITY OF CARDIAC MAGNETIC RESONANCE T1 MAPPING IMAGES

Background: Myocardial T1 mapping is well established as a quantitative, non-invasive tool for myocardial tissue characterization (1). Incidence of abnormal myocardium are common in patients with cardiomyopathies but may additionally stem from anatomical misalignment due to motion (2). The latter influence is not truly representative of abnormal myocardium, thus may result in falsely elevated T1 values and impair the image quality (IQ) of T1 mapping images. Two techniques: MOdified Look-Locker Inversion recovery with Motion Correction (MOLLI-MOCO) and MOLLI-MOCO with deep learning (MOLLI-MOCO-DL), have been proposed to target these limitations (3,4).

METHODS AND RESULTS: Eighteen patients were scanned on a 3T SIGNA Premier (GE Healthcare, Milwaukee, USA). Three slices were acquired in the short-axis orientation using the MOLLI T1 mapping sequence. Images were further processed using both MOLLI-MOCO and MOLLI-MOCO-DL. The global and segmental myocardial T1 values were measured using cvi42 (Circle Cardiovascular Imaging Inc, Calgary, Canada) for each MOLLI, MOLLI-MOCO, and MOLLI-MOCO-DL using a blinded, cross-sectional analysis. An experienced reader also evaluated IQ using a 4-point Likert scale where 1 = non-diagnostic IQ, 2 = diagnostic IQ with many artefacts, 3 = few artefacts, and 4 = perfect IQ. Normal myocardial T1 values, obtained from a cohort of healthy volunteers, were used to classify segments of the myocardium as increased (greater than 2 SD), normal (within 2 SD) or decreased (lower than 2 SD) (5). The Fleiss Kappa statistic was used to assess the similarity between the classification of the three methods. All 18 patients were successfully scanned, and their maps analyzed. An example of the T1 maps obtained using each method for one subject is shown in Figure 1. The MOLLI-MOCO and MOLLI-MOCO-DL techniques resulted in a higher IQ compared with MOLLI (Figure 2, A). All three methods had a reasonable agreement in the classification of myocardial tissue with a Fleiss Kappa score of 0.78. MOLLI-MOCO-DL and MOLLI-MOCO techniques classified a greater number of segments as normal compared with MOLLI, indicating that they may suffer from fewer false positives than conventional MOLLI (Figure 2, B).

Conclusion: T1 MOLLI-MOCO and MOLLI-MOCO-DL improves IQ and classifies fewer segments as abnormal compared to MOLLI. Higher IQ is important for image analysis and for the robustness of T1 mapping evaluation. Further investigation into advanced motion compensation techniques is necessary to determine the robustness of their measured T1 values and their diagnostic accuracy.