Fig. 4: Dataset composition, surgical movement distribution, and patient characteristics.

a Paired lateral cephalograms before and after surgery (N = 707). Pre- and post-cephs were aligned using Sella and Nasion landmarks. b All pairs (N = 650) of pre- and post-cephs from three university hospitals and one dental hospital (A, B, C, and D) were used as internal datasets. Of these, 550, 50, and 50 pairs were used for training, validation, and internal tests, respectively. c Distribution of surgical movement for representative landmarks of hard and soft tissues, including (from upper left to lower right) anterior nasal spine (ANS), posterior nasal spine (PNS), B-point, Md 1crown (Mandible 1 crown), upper lip, lower lip, soft tissue pogonion, and soft-tissue menton. d Distribution of surgery types across the entire dataset and proportion of each type per hospital, where the horizontal axis of the bar graph represents the hospitals, and the vertical axis indicates the ratio of patients. Consequently, 57 pairs of pre- and post-cephs from six university hospitals (E, F, G, H, I, and J) were used for external tests. e Gender distribution of patients. f Distribution of pre-surgical orthodontic treatment time, where the horizontal axis represents the number of months, and the vertical axis denotes the number of patients.