Fig. 4From: Prediction of orthognathic surgery plan from 3D cephalometric analysis via deep learningArchitecture of VSP transformer. Norm: normalization. CT-reconstructed CMF model was quantified using 3D cephalometric analysis. 12 cephalometric variables and gender were weighted using ridge coefficients. After embedding, the data were encoded with positional information and then input into encoder block to extract features. At last, the extracted features were decoded by a MLP and an embedded residual structure to obtain output variablesBack to article page