Automated quantification of mitral valve geometry on multi-slice computed tomography in patients with dilated cardiomyopathy: Implications for transcatheter mitral valve replacement

Objectives
The primary aim of this study was to quantify the dimensions and geometry of the mitral valve complex in patients with dilated cardiomyopathy and significant mitral regurgitation. The secondary aim was to evaluate the validity of an automated segmentation algorithm for assessment of the mitral valve compared to manual assessment on computed tomography.

Background Transcatheter mitral valve replacement (TMVR) is an evolving technique which relies heavily on the lengthy evaluation of cardiac computed tomography (CT) datasets. Limited data is available on the dimensions and geometry of the mitral valve in pathological states throughout the cardiac cycle, which may have implications for TMVR device design, screening of suitable candidates and annular sizing prior to TMVR.

Methods A retrospective study of 15 of patients with dilated cardiomyopathy who had undergone full multiphase ECG gated cardiac CT. A comprehensive evaluation of mitral valve geometry was performed at 10 phases of the cardiac cycle using the recommended D-shaped mitral valve annulus (MA) segmentation model using manual and automated CT interpretation platforms. Mitral annular dimensions and geometries were compared between manual and automated methods.

Results Mitral valve dimensions in patients with dilated cardiomyopathy were similar to previously reported values (MAarea Diastole: 12.22 ± 1.90 cm2), with dynamic changes in size and geometry between systole and diastole of up to 5%. The distance from the centre of the MA to the left ventricular apex demonstrated moderate agreement between automated and manual methods (ρc = 0.90) with other measurements demonstrating poor agreement between the two methods (ρc = 0.75–0.86).

Conclusions
Variability of mitral valve annulus measurements are small during the cardiac cycle. Novel automated algorithms to determine cardiac cycle variations in mitral valve geometry may offer improved segmentation accuracy as well as improved CT interpretation times.