Author + information
- Patricia Lopes1,
- Roel Wirix-Speetjens1,
- Jan Sijbers2,
- Jos Vander Sloten3,
- Johan Bosmans2,
- Paul Van Herck4 and
- Janelle Schrot5
The accurate delineation of the mitral annulus is crucial for an adequate transcatheter mitral valve replacement planning. However, its manual annotation is prone to inconsistencies and interoperator variability associated with variations in the operators’ expertise, high inhomogeneity in the patient population, and differences between protocols. In an attempt to address these challenges, this study investigated the suitability of an automated method for the mitral annulus identification and quantification compared with a manual approach.
A statistical shape model was generated from 50 4D CT scans of human subjects without cardiovascular disease. The statistical shape model was subsequently fitted to 10 additional cases for automatic identification of the native saddle-shaped mitral annulus, for both the systolic and the diastolic phases. Concurrently, the mitral annulus was manually indicated by 2 operators. Anteroposterior (AP) and commissure-commissure (CC) diameters, projected perimeter, and projected area were calculated to evaluate the interoperator variability and the differences between the average values obtained for the manual approach and those for the automatic method.
Table 1 summarizes the differences between the measurements resulting from the manually and automatically generated mitral annulus.
|Measurement||Average Operator||Automatic||Operator 1 − Operator 2||Operators − Automatic|
|AP diameter, mm||29.9 ± 1.6||26.7 ± 2.7||−1.1 ± 0.6||3.3 ± 2.1|
|CC diameter, mm||35.3 ± 3.3||35.6 ± 2.1||−0.1 ± 1.1||−0.3 ± 2.3|
|Projected perimeter, mm||107.9 ± 4.5||108.6 ± 3.2||1.6 ± 1.7||−0.6 ± 3.6|
|Projected area, mm2||870.1 ± 55.7||884.9 ± 76.6||−13.6 ± 179.8||−27.4 ± 78.4|
The measurements resulting from the automatically predicted annulus are in good correspondence with those resulting from the manual delineations. The largest differences were observed in the aortic-mitral curtain, which is reflected in the measurements, as the differences for the AP diameters are larger than those for the CC diameters. Even though the number of patients is still limited, this ongoing study suggests that automation of mitral valve quantification is possible and its performance is in line with that of a manual approach, potentially contributing to more consistent pre–transcatheter mitral valve replacement planning.
STRUCTURAL: Valvular Disease: Mitral