Combining Deformation Modeling and Machine Learning for Personalized Prosthesis Size Prediction in Valve-Sparing Aortic Root Reconstruction

Abstract

Finding the individually optimal prosthesis size is an intricate task during valve-sparing aortic root reconstruction. Previous work has shown that machine learning based prosthesis size prediction is possible. However, the very high demands on the underlying training data set prevent the application in a clinical setting. In this work, the authors present an alternative approach combining simplified deformation modeling with machine learning to mimic the surgeon’s decision making process. Compared to the previously published approach, the new method provides a similar prediction accuracy whith a dramatic decrease of demand on the training data. This is an important step towards the clinical application of machine learning based planning of personalized valve-sparing aortic root reconstruction.