Accuracy of a mitral valve segmentation method using J-splines for real-time 3D echocardiography data

Area of Science:

• Biomedical Engineering
• Cardiovascular Imaging
• Computational Anatomy

Background:

• Patient-specific models of the heart's mitral valve (MV) are crucial for surgical planning.
• 3D echocardiography (3DE) offers high-resolution MV leaflet geometry but its accuracy beyond closure and in various dysfunctions is unquantified.

Purpose of the Study:

• To quantitatively assess the accuracy of a 3DE-based J-spline segmentation methodology for mitral valve leaflets.
• To evaluate accuracy across different temporal frames (closure, peak closure, opening) and valvular dysfunctions (control, prolapsed, billowing).
• To establish an experimental platform for validating MV segmentation techniques for surgical planning.

Main Methods:

• A 3DE-based J-spline segmentation methodology was applied to porcine MV models in a pulsatile simulator.
• Ground-truth 4D leaflet coordinates were used for accuracy assessment during various dynamic phases and simulated dysfunctions.
• The methodology was also applied to a patient dataset to evaluate clinical relevance.

Main Results:

• Mean distance errors between segmented and ground-truth MV leaflets were 0.40 ± 0.32 mm (control), 0.52 ± 0.51 mm (flail), and 0.74 ± 0.69 mm (billowing).
• 95% of distance errors were below 1.64 mm across all models and temporal frames.
• Segmentation successfully identified a regurgitant orifice and post-operative improvements in a patient case.

Conclusions:

• The 3DE-based J-spline segmentation methodology demonstrates high accuracy for mitral valve leaflet modeling.
• The validated methodology is suitable for assessing MV geometry beyond systolic closure and in various valvular dysfunctions.
• This work supports the development of advanced MV segmentation tools for patient-specific surgical planning.