Author + information
- Dee Dee Wang1,
- Wouter Houthoofd2,
- Lotte Huysmans3,
- William O'Neill1,
- Julie Maes3,
- Clare Ward4 and
- Janelle Schrot4
Left ventricular outflow tract (LVOT) obstruction is a life-threatening complication after transcatheter mitral valve replacement (TMVR), occurring in ∼10% of patients. While CT assessment is used to predict LVOT obstruction, traditional 2D neo-LVOT prediction methods carry little operator confidence and have high operator interobserver variability. This may result in inappropriate inclusion and potential exclusion of patients for TMVR therapies. Traditional TMVR planning tools utilize 2D generated centerlines to predict neo-LVOT post-TMVR. To date, there is no study looking at inter-observer variability between neo-LVOT prediction using the traditional 2D based center-line approach versus novel 3D computer-aided-design semi-automatic prediction of neo-LVOT obstruction. The objective of this work is to compare interrater reliability of the 3D method to the traditional 2D method.
3D models of the heart were segmented from 24 pre-TMVR CT scans. A virtual transcatheter aortic valve model was positioned in the mitral annulus, simulating a valve-in-MAC TMVR procedure. Two modelers independently measured the neo-LVOT surface area in Mimics (Materialise Leuven, Belgium), using 2 methods. In the 3D semi-automatic method, the modeler defined an initial plane in the LVOT. The software then surveyed the neo-LVOT space to identify the smallest neo-LVOT area. In the 2D manual method, the modeler drew a centerline through the LVOT into the ascending aorta on CT images. A neo-LVOT plane was created perpendicular to the centerline, at the bottom of the device. In both methods, the neo-LVOT area of the selected plane was calculated in a slice of the 3D model. Intraclass Correlation Coefficient (ICC) and absolute difference (paired t-test) of neo-LVOT areas between observers were calculated for both methods.
Higher interrater agreement was found for the 3D (ICC: 0.99) vs 2D method (ICC: 0.87). There was a significantly higher average difference in area between modelers for the 2D method (63 ±93mm2) compared to the 3D method (0.38±0.90mm2) (p=0.003).
With the 3D semi-automatic tool, 2 modelers estimated a more reliable neo-LVOT area than with the 2D centerline method.
STRUCTURAL: Valvular Disease: Mitral