Using a Disentangled Neural Network to Objectively Assess the Outcomes of Midfacial Surgery in Syndromic Craniosynostosis.

Abstract

Advancements in artificial intelligence and the development of shape models that quantify normal head shape and facial morphology provide frameworks by which the outcomes of craniofacial surgery can be compared. In this work, the authors demonstrate the use of the swap disentangled variational autoencoder to assess changes after midfacial surgery objectively. The model is trained on a data set of 1405 3-dimensional meshes of healthy individuals and syndromic patients, which was augmented using a technique based on spectral interpolation. Patients with a diagnosis of Apert or Crouzon syndrome who had undergone sub- or transcranial midfacial procedures using rigid external distraction had their results interpreted using this model as the point of comparison. A total of 56 patients met the inclusion criteria: 20 with Apert syndrome and 36 with Crouzon syndrome. By using linear discriminant analysis to project the high-dimensional vectors derived by swap disentangled variational autoencoder onto a 2-dimensional space, the shape properties of Apert syndrome and Crouzon syndrome can be visualized in relation to the healthy population. In this way, the authors were able to show how surgery elicits global shape changes in each patient. To assess the regional movements achieved during surgery, the authors used a novel metric derived from the Mahalanobis distance to quantify movements through the latent space. Objective outcome evaluation, which encourages in-depth analysis and enhances decision-making, is essential for the progression of surgical practice. The authors demonstrate how artificial intelligence has the ability to improve our understanding of surgery and its effect on craniofacial morphology.