A statistical shape modelling method for predicting the human head from the face
Vi Do Tran, Phong Phu Vo, Ngoc Lan Nhi Tran, Tien Tuan Dao, Tan Nhu Nguyen · International Journal of Biomedical Engineering and Technology · 2024
Predicting the back head based only on the face is necessary for generating the full head and skull. We prepared a dataset of 329 surface meshes of the head. These meshes were reconstructed and post-processed from computed tomography (CT) image sets of adult subjects having normal head structures. A novel back head and face sampling technique was also developed for acquiring back head and face features. The relation between the face features and the back head features was trained using four strategies: non-rigid scaling, SSM optimization, partial least squares regression (PLSR), and principal component analysis (PCA). A ten-fold cross-validation procedure was conducted for selecting the optimal training strategies and tuning their parameters. The face features and the predicted back head features formed the head. The mean mesh-to-mesh distances between the predicted and the ground truth back head were (mean ± SD) 1.15 ± 0.21 mm. The method will enhance the accuracy of face-to-skull prediction.