Date: February 2022.
Source: Scientific Reports, vol. 12, Article 2230. DOI: 10.1038/s41598-021-02411-y.
Research Summary: Clinical diagnosis of craniofacial syndromes normally requires expert knowledge, and most AI-based analysis tools work from 2D photographs, analyzing texture and color rather than shape, which limits their use with medical imaging modalities like ultrasound, MRI, or CT. Researchers from Great Ormond Street Hospital, Imperial College London, Alder Hey Children’s Hospital, and Necker Hospital in Paris built a research framework using Convolutional Mesh Autoencoders (CMAs), a shape-based successor to 3D morphable models (3DMMs) that works directly from surface topography rather than texture, making it usable across both photography and medical imaging. A single model was trained on a combined pool of real-world 3dMD facial and head scans, drawn from the datasets underlying the LSFM and LYHM shape models, alongside CT-derived meshes of infants and children with three genetically distinct craniofacial syndromes, Apert, Crouzon, and Muenke.

Article: Convolutional Mesh Autoencoders for the 3-dimensional Identification of FGFR-related Craniosynostosis.
Authors: Eimear O’Sullivan, Lara S van de Lande, Athanasios Papaioannou, Richard WF. Breakey, N Owase Jeelani, Allan Ponniah, Christian Duncan, Silvia Schievano, Roman H. Khonsari, Stefanos Zafeiriou, David J Dunaway. UCL Great Ormond Street Institute of Child Health; Necker-Enfants Malades University Hospital, Paris; Department of Computing, Imperial College London; Royal Free Hospital, London; Craniofacial Unit, Alder Hey Children’s Hospital, Liverpool.

Related Articles:

  1. The Identification of Beckwith-Wiedemann Syndrome Through Swap Disentangled Variational Autoencoder. T Rijlaarsdam, L Smith, A Rickart, S Schievano, D Dunaway, E Wolvius, L van de Lande, S Foti, J Ong.
  2. Oral and craniofacial research in the Generation R study: an executive summary. AW van Meijeren-van Lunteren, X Liu, FCH Veenman, et al.
  3. What can a morphometric study of unoperated children teach us about the natural history of metopic synostosis? AM Elawadly, LPJ Smith, A Borghi, KI Abdelaziz, AH Dulanka Silva, DJ Dunaway, NU Owase Jeelani, J Ong, RD Hayward, G James.
  4. Using Facial Morphology Traits Related to Attention Problems in Children to Identify Prenatal Exposures History. A Shchetinina, X Liu, N Slopen, C Chambers, M Kayser, E Wolvius, G Roshchupkin, H Tiemeier.