Enrico Giraudo
Deep Learning for Face-Bone Transformation and Postoperative Facial Prediction: Evaluation of TCFNet and DGCFP.
Rel. Federica Marcolin, Elena Carlotta Olivetti. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Biomedica, 2026
Abstract
Orthognathic surgical planning can define the repositioning of the maxilla, mandible, and other skeletal segments with accuracy. The resulting soft-tissue response is harder to anticipate. Soft-tissue deformation varies from one patient to another and is influenced by local anatomy, the direction of the surgical movement and the interactions among adjacent facial regions. Traditional approaches range from two-dimensional cephalometric prediction to Mass-Spring Models (MSMs), Mass-Tensor Models (MTMs), and Finite Element Models (FEMs). While these methods can estimate the soft-tissue response to skeletal repositioning, their use may involve simplified modelling assumptions, time-consuming patient-specific preparation, and, particularly for FEMs, considerable computational effort. This thesis examines two geometric deep learning frameworks developed for craniofacial modelling.
TCFNet performs bidirectional learning between facial and skeletal point clouds by estimating pointwise displacement fields through a coarse-to-fine architecture that combines a Transformer-based global stage with a Local Information Aggregation Network for local refinement
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Informazioni aggiuntive
Corso di laurea
Classe di laurea
URI
![]() |
Modifica (riservato agli operatori) |
