Deep Learning–Based Prediction of the 3D Postorthodontic Facial Changes

Abstract

With the increase of the adult orthodontic population, there is a need for an accurate and evidence-based prediction of the posttreatment face in 3 dimensions (3D). The objectives of this study are 1) to develop a 3D postorthodontic face prediction method based on a deep learning network using the patient-specific factors and orthodontic treatment conditions and 2) to validate the accuracy and clinical usability of the proposed method. Paired sets ( n = 268) of pretreatment (T1) and posttreatment (T2) cone-beam computed tomography (CBCT) of adult patients were trained with a conditional generative adversarial network to generate 3D posttreatment facial data based on the patient’s gender, age, and the changes of upper (ΔU1) and lower incisor position (ΔL1) as input. The accuracy was calculated with prediction error and mean absolute distances between real T2 (T2) and predicted T2 (PT2) near 6 perioral landmark regions, as well as percentage of prediction error less than 2 mm using test sets ( n = 44). For qualitative evaluation, an online survey was conducted with experienced orthodontists as panels ( n = 56). Overall, PT2 indicated similar 3D changes to the T2 face, with the most apparent changes simulated in the perioral regions. The mean prediction error was 1.2 ± 1.01 mm with 80.8% accuracy. More than 50% of the experienced orthodontists were unable to distinguish between real and predicted images. In this study, we proposed a valid 3D postorthodontic face prediction method by applying a deep learning algorithm trained with CBCT data sets.

Deep learning synthesis of cone-beam computed tomography from zero echo time magnetic resonance imaging

“…Previous studies utilized the adversarial learning strategy 7,25 , which trains a synthesis model with a discriminator that tries to distinguish target images as either real or synthesized. However, adversarial learning is known to be challenging to optimize due to "mode collapse," in which a synthesis model keeps generating identical samples 26 .…”

Three-Dimensional Multi-Modality Registration for Orthopaedics and Cardiovascular Settings: State-of-the-Art and Clinical Applications

“…Nowadays, similarly to the cardiovascular fields, the state-of-the-art methods for orthopaedics/orthodontics registration are neural networks [ 26 , 27 ]. In the case of orthodontics image registration, for example, Park and colleagues [ 28 ] have tried to develop a 3D post orthodontic face prediction method using a deep learning network that incorporates patient-specific factors and orthodontic treatment conditions. To this end, soft tissue masks derived from T1 and T2 CTs were reoriented and registered through stable anatomic structures of the cranial base for extraction and training purposes.…”

Artificial Intelligence in Orthodontics: Critical Review

“…Similar to generating text, generative AI can help with patient communication through image generation. First efforts have been undertaken to predict changes in an adult’s facial morphology following orthodontic treatment using AI (Park et al 2022). Intermediate fusion, a multimodal fusion technique that we describe below, was used to combine data from pretreatment cone beam computed tomography (CBCT) together with patient characteristics (i.e., patients’ age, sex, and the amount of incisor movement provided by orthodontic treatment).…”

“…Notably, efforts toward intermediate or late fusion have only been undertaken for tasks such as classification of lateral cephalograms or treatment outcome prediction (Yu et al 2020; Park et al 2022). However, the application of intermediate or late fusion techniques in orthodontic treatment planning has not been assessed, but it may be worthwhile, as features extracted from distinct modalities could be more meaningful than those captured by human experts (Acosta et al 2022).…”