Latent disentanglement in mesh variational autoencoders improves the diagnosis of craniofacial syndromes and aids surgical planning.

Craniofacial syndromes Geometric deep learning Latent disentanglement

Journal

Computer methods and programs in biomedicine
ISSN: 1872-7565
Titre abrégé: Comput Methods Programs Biomed
Pays: Ireland
ID NLM: 8506513

Informations de publication

Date de publication:
26 Aug 2024
Historique:
received: 06 08 2023
revised: 29 05 2024
accepted: 23 08 2024
medline: 31 8 2024
pubmed: 31 8 2024
entrez: 30 8 2024
Statut: aheadofprint

Résumé

The use of deep learning to undertake shape analysis of the complexities of the human head holds great promise. However, there have traditionally been a number of barriers to accurate modelling, especially when operating on both a global and local level. In this work, we will discuss the application of the Swap Disentangled Variational Autoencoder (SD-VAE) with relevance to Crouzon, Apert and Muenke syndromes. The model is trained on a dataset of 3D meshes of healthy and syndromic patients which was increased in size with a novel data augmentation technique based on spectral interpolation. Thanks to its semantically meaningful and disentangled latent representation, SD-VAE is used to analyse and generate head shapes while considering the influence of different anatomical sub-units. Although syndrome classification is performed on the entire mesh, it is also possible, for the first time, to analyse the influence of each region of the head on the syndromic phenotype. By manipulating specific parameters of the generative model, and producing procedure-specific new shapes, it is also possible to approximate the outcome of a range of craniofacial surgical procedures. This work opens new avenues to advance diagnosis, aids surgical planning and allows for the objective evaluation of surgical outcomes. Our code is available at github.com/simofoti/CraniofacialSD-VAE.

Sections du résumé

BACKGROUND AND OBJECTIVE OBJECTIVE
The use of deep learning to undertake shape analysis of the complexities of the human head holds great promise. However, there have traditionally been a number of barriers to accurate modelling, especially when operating on both a global and local level.
METHODS METHODS
In this work, we will discuss the application of the Swap Disentangled Variational Autoencoder (SD-VAE) with relevance to Crouzon, Apert and Muenke syndromes. The model is trained on a dataset of 3D meshes of healthy and syndromic patients which was increased in size with a novel data augmentation technique based on spectral interpolation. Thanks to its semantically meaningful and disentangled latent representation, SD-VAE is used to analyse and generate head shapes while considering the influence of different anatomical sub-units.
RESULTS RESULTS
Although syndrome classification is performed on the entire mesh, it is also possible, for the first time, to analyse the influence of each region of the head on the syndromic phenotype. By manipulating specific parameters of the generative model, and producing procedure-specific new shapes, it is also possible to approximate the outcome of a range of craniofacial surgical procedures.
CONCLUSION CONCLUSIONS
This work opens new avenues to advance diagnosis, aids surgical planning and allows for the objective evaluation of surgical outcomes. Our code is available at github.com/simofoti/CraniofacialSD-VAE.

Identifiants

DOI: 10.1016/j.cmpb.2024.108395 PMID: 39213899
pubmed: 39213899
pii: S0169-2607(24)00388-2
doi: 10.1016/j.cmpb.2024.108395
pii:
doi:

Types de publication

Journal Article

Langues

eng

Sous-ensembles de citation

IM

Pagination

108395

Informations de copyright

Copyright © 2024 The Author(s). Published by Elsevier B.V. All rights reserved.

Déclaration de conflit d'intérêts

Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Athanasios Papaioannou and Eimear O’Sullivan are currently employed by Huawei Technologies Co., Ltd. They were with Imperial College London and University College London during the experiments. The other authors declare no competing interests.

Auteurs

Simone Foti (S)

Wellcome/EPSRC Centre for Interventional and Surgical Sciences, University College London, London, UK; Centre For Medical Image Computing, University College London, London, UK; Imperial College London, Department of Computing, London, UK. Electronic address: s.foti@cs.ucl.ac.uk.

Alexander J Rickart (AJ)

UCL Great Ormond Street Institute of Child Health and Craniofacial Unit, Great Ormond Street Hospital for Children, London, UK.

Bongjin Koo (B)

Wellcome/EPSRC Centre for Interventional and Surgical Sciences, University College London, London, UK; Centre For Medical Image Computing, University College London, London, UK; University of California, Santa Barbara, Department of Electrical & Computer Engineering, Santa Barbara, USA.

Eimear O' Sullivan (E)

UCL Great Ormond Street Institute of Child Health and Craniofacial Unit, Great Ormond Street Hospital for Children, London, UK; Imperial College London, Department of Computing, London, UK.

Lara S van de Lande (LS)

Department of Oral and Maxillofacial Surgery, Erasmus Medical Center, Rotterdam, The Netherlands.

Athanasios Papaioannou (A)

UCL Great Ormond Street Institute of Child Health and Craniofacial Unit, Great Ormond Street Hospital for Children, London, UK; Imperial College London, Department of Computing, London, UK.

Roman Khonsari (R)

Department of Maxillofacial Surgery and Plastic Surgery, Necker - Enfants Malades University Hospital, Paris, France.

Danail Stoyanov (D)

Wellcome/EPSRC Centre for Interventional and Surgical Sciences, University College London, London, UK; Centre For Medical Image Computing, University College London, London, UK.

N U Owase Jeelani (NUO)

UCL Great Ormond Street Institute of Child Health and Craniofacial Unit, Great Ormond Street Hospital for Children, London, UK.

Silvia Schievano (S)

UCL Great Ormond Street Institute of Child Health and Craniofacial Unit, Great Ormond Street Hospital for Children, London, UK.

David J Dunaway (DJ)

UCL Great Ormond Street Institute of Child Health and Craniofacial Unit, Great Ormond Street Hospital for Children, London, UK.

Matthew J Clarkson (MJ)

Wellcome/EPSRC Centre for Interventional and Surgical Sciences, University College London, London, UK; Centre For Medical Image Computing, University College London, London, UK.

Classifications MeSH