Image-based computational models for TAVI planning: from CT images to implant deployment

Area of Science:

• Cardiovascular Surgery
• Medical Imaging
• Computational Biology

Background:

• Transcatheter aortic valve implantation (TAVI) is a standard treatment for severe aortic valve stenosis in non-operable patients.
• Accurate preoperative planning, particularly selecting the correct implant size and type, is critical for successful TAVI outcomes.
• Current planning relies heavily on clinician experience due to complex patient anatomy and device variability.

Purpose of the Study:

• To develop and validate an integrated computational method for patient-specific TAVI planning.
• To automatically estimate aortic anatomy from CT images and predict TAVI implant deployment.
• To provide a more objective and predictive tool for TAVI procedure planning.

Main Methods:

• Automatic extraction of aortic root, leaflets, and calcifications from CT images using modeling and machine learning.
• Finite element method simulation to compute TAVI implant deployment within patient-specific aortic anatomy.
• Validation of the anatomical model on 198 CT images and prediction accuracy assessment using pre- and post-TAVI CT scans.

Main Results:

• The anatomical model achieved an accuracy of 1.30 +/- 0.23 mm on 198 CT images.
• Predicted implant deployment errors averaged 1.74 +/- 0.40 mm, with 1.32 mm in the aortic valve annulus region.
• The prediction accuracy in the annulus region is nearly three times better than the typical 3 mm gap between consecutive implant sizes.

Conclusions:

• The proposed integrated method offers a promising surrogate tool for TAVI planning.
• This computational approach can enhance the precision of implant selection and deployment prediction.
• Automated analysis of aortic anatomy and implant simulation can lead to improved TAVI outcomes.