University of Twente Student Theses
Predicting facial hard tissue shape from soft tissue shape with a dense autoencoder and principal component analysis : a proof of concept
Westra, I. (2023) Predicting facial hard tissue shape from soft tissue shape with a dense autoencoder and principal component analysis : a proof of concept.
PDF
5MB |
Abstract: | Introduction: Facial feminization surgery (FFS) is an underdeveloped surgical field in need of objective methods to plan, execute, predict and evaluate FFS and its outcomes. If the desired postoperative facial shape is known, a method to translate that soft tissue shape into matching hard tissue shape will make objective surgical planning possible. A proof of concept for translating facial soft tissue shape towards matching hard tissue shape is provided in this research. Methods: Principal component analysis was performed on 252 facial soft and hard tissue meshes with 1:1 vertex correspondence. The soft and hard tissue principal component scores were used as input and output, respectively, for a dense autoencoder. The training, validation and test sets consisted of 200, 31, and 21 subjects respectively. The predicted hard tissue principal component scores were reconstructed to meshes and compared to the true meshes. Results: The reconstructed meshes had a median error in the x, y and z coordinates below 1.5 mm with median absolute deviation below 1.2 mm. The regions with errors above the third quartile corresponded with inaccurately predicted principal components. Conclusion: This study provides a proof of concept for the predictive value of facial soft tissue shape for facial hard tissue shape through a combination of PCA and a dense autoencoder applied to 3D meshes. Along with more data of subjects with comparable demographics and a stricter exclusion on image quality, the results should improve and the true possibilities for objective FFS methods should become clear. |
Item Type: | Essay (Master) |
Clients: | Radboudumc, Nijmegen, Netherlands Amsterdam UMC - location VUmc, Amsterdam, Netherlands |
Faculty: | TNW: Science and Technology |
Subject: | 44 medicine |
Programme: | Technical Medicine MSc (60033) |
Link to this item: | https://purl.utwente.nl/essays/94748 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page