Prediction of individual prosthesis size for valve-sparing aortic root reconstruction based on geometric features

Abstract

Valve-sparing aortic root reconstruction is an up- and-coming approach for patients suffering from aortic valve insufficiencies which promises to significantly reduce complications. However, the success of the treatment strongly depends on the challenging task of choosing the correct size of the prosthesis, for which, up to now, surgeons solely have to rely on their experience. Here, we present a novel machine learning based approach, which might make it possible to predict the size of the prosthesis from pre-operatively acquired ultrasound images. We utilize support vector regression to train a prediction model on three geometric features extracted from the ultrasound data. In order to evaluate the accuracy and robustness of our approach we created a large data base of porcine aortic root geometries in a healthy state and an artificially dilated state. Our results indicate that prediction of correct prosthesis sizes is feasible. Furthermore, they suggest that it is crucial that the training data set faithfully represents the diversity of aortic root geometries.

A Support Vector Regression-Based Data-Driven Leaflet Modeling Approach for Personalized Aortic Valve Prosthesis Development

“…One possibility to overcome this problem could be an additional estimation step predicting the healthy geometry based on the obtained one. A similar approach was already tested in a comparable cardiovascular problem [6].…”

“…2 (c)). Previous studies have shown that this simple geometric description allows for a reasonable representation of the individual geometry [6]. Based on these landmarks, different geometric key features were extracted, namely the commissure distances K 1 , K 2 and K 3 as well as the leaflets' free edge lengths S 1 , S 2 and S 3 (see Fig.…”

Generating Healthy Aortic Root Geometries from Ultrasound Images of the Individual Pathological Morphology Using Deep Convolutional Autoencoders

“…To evaluate our method, we tested it on an ex-vivo porcine data set. The data set was published in [3] and consists of 3D ultrasound images of 24 isolated aortic roots in the closed state under physiologically realistic pressure conditions. During the collection of the data set, at first, images of the natural aortic roots were taken.…”

“…In this first approach, we extracted one 2D slice image from each volume in which the commissure plane, i.e. the slice image were all three commissure points are visible, is shown [3].…”

“…As this problem is about infering from image information, machine learning is a promising approach to solve this problem. However, classical supervised learning techniques, where the prediction model learns the mapping from a dilated shape to the corresponding healthy shape based on annotated shape pairs (comparable to [3]), suffer from the unrealistic demand for training data as collecting these shape pairs in a clinical study is very unlikely.…”