Cardiovascular magnetic resonance determinants of left ventricular noncompaction

Insufficient precision remains in accurately identifying left ventricular noncompaction (LVNC) from the healthy normal morphologic spectrum. We aim to provide a better distinction between normal left ventricular trabeculations and LVNC. We used a previously well-defined cohort of 120 healthy volunteers for normal reference values of the trabecular/compacted ratio derived from a consistent selection of short-axis cardiovascular magnetic resonance images. We performed forward selection of logistic regression models, selecting the best model that was subsequently assessed for discrimination and calibration, validated, and converted into a clinical diagnostic chart to benchmark the boundaries of detection from a cohort of 30 patients considered to have LVNC. We showed that 3 combinations of a maximal end-diastolic trabecular/compacted ratio (≥1 [apex], >1.8 [midcavity]), (>2 [apex], ≥0.6 [midcavity]), or (>0.5 [base], >1.8 [midcavity]) separate the cohorts with the highest accuracy (C statistic [95% confidence interval] of 0.9749 (0.9748 to 0.9751) for the diagnostic chart). Quantitative cardiovascular magnetic resonance also shows that patients considered to have LVNC have a significantly reduced ejection fraction compared with normal volunteers. At midcavity and apical level, it is difficult to identify papillary muscles that are replaced by a dense trabecular meshwork. In conclusion, we developed a new, refined, diagnostic tool for identifying LVNC, based on an a priori assessment of the trabecular architecture in healthy volunteers.