Aortic root sizing for transcatheter aortic valve implantation using a shape model parameterisation

Bart Bosmans, Toon Huysmans, Patricia Lopes, Eva Verhoelst, Tim Dezutter, Peter de Jaegere, Jan Sijbers, Jos Vander Sloten, Johan Bosmans

Research output: Contribution to journalArticleScientificpeer-review

1 Citation (Scopus)
82 Downloads (Pure)


During a transcatheter aortic valve implantation, an axisymmetric implant is placed in an irregularly shaped aortic root. Implanting an incorrect size can cause complications such as leakage of blood alongside or through the implant. The aim of this study was to construct a method that determines the optimal size of the implant based on the three-dimensional shape of the aortic root. Based on the pre-interventional computed tomography scan of 89 patients, a statistical shape model of their aortic root was constructed. The weights associated with the principal components and the volume of calcification in the aortic valve were used as parameters in a classification algorithm. The classification algorithm was trained using the patients with no or mild leakage after their intervention. Subsequently, the algorithms were applied to the patients with moderate to severe leakage. Cross validation showed that a random forest classifier assigned the same size in 65 ± 7% of the training cases, while 57 ± 8% of the patients with moderate to severe leakage were assigned a different size. This initial study showed that this semi-automatic method has the potential to correctly assign an implant size. Further research is required to assess whether the different size implants would improve the outcome of those patients.
Original languageEnglish
Pages (from-to)2081-2092
Number of pages12
JournalMedical & Biological Engineering & Computing
Issue number10
Publication statusPublished - 2019

Bibliographical note

Accepted Author Manuscript


  • Aortic root sizing
  • Statistical shape modelling
  • Transcatheter aortic valve implantation


Dive into the research topics of 'Aortic root sizing for transcatheter aortic valve implantation using a shape model parameterisation'. Together they form a unique fingerprint.

Cite this