Accuracy of one-step automated orthodontic diagnosis model using a convolutional neural network and lateral cephalogram images with different qualities obtained from nationwide multi-hospitals

Abstract

Objective The purpose of this study was to investigate the accuracy of one-step automated orthodontic diagnosis of skeletodental discrepancies using a convolutional neural network (CNN) and lateral cephalogram images with different qualities from nationwide multi-hospitals. Methods Among 2,174 lateral cephalograms, 1,993 cephalograms from two hospitals were used for training and internal test sets and 181 cephalograms from eight other hospitals were used for an external test set. They were divided into three classification groups according to anteroposterior skeletal discrepancies (Class I, II, and III), vertical skeletal discrepancies (normodivergent, hypodivergent, and hyperdivergent patterns), and vertical dental discrepancies (normal overbite, deep bite, and open bite) as a gold standard. Pre-trained DenseNet-169 was used as a CNN classifier model. Diagnostic performance was evaluated by receiver operating characteristic (ROC) analysis, t-stochastic neighbor embedding (t-SNE), and gradient-weighted class activation mapping (Grad-CAM). Results In the ROC analysis, the mean area under the curve and the mean accuracy of all classifications were high with both internal and external test sets (all, > 0.89 and > 0.80). In the t-SNE analysis, our model succeeded in creating good separation between three classification groups. Grad-CAM figures showed differences in the location and size of the focus areas between three classification groups in each diagnosis. Conclusions Since the accuracy of our model was validated with both internal and external test sets, it shows the possible usefulness of a one-step automated orthodontic diagnosis tool using a CNN model. However, it still needs technical improvement in terms of classifying vertical dental discrepancies.

A Scoping Review of Machine Learning in Dental Radiography: Its Current Applications and Relevance in Dentistry

“…Yu et al 31 investigated the ability of a DenseNet AI model in identifying vertical skeletal and sagittal classes from lateral cephalograms. Similar results were seen as compared to Yim et al 30 where the sensitivity scores were 6-7% lower than the speci city scores across all reported skeletal and sagittal classes. However, the study authors reported an imbalanced dataset and opted to under-sample the data where 882 images were excluded from the study to ensure equal group weighting.…”

“…In this section, four studies were included, with two detecting supernumerary teeth (ST) 22,26 and the other two relating to skeletal classi cation 30,31 . Mine et al 22 investigated the effectiveness of an AI model in detecting single and double ST in the early mixed dentition stage using panoramic radiographs.…”

“…27 The consequences of deregulated AI have been noted on social media where models recommend harmful content to users 28 , or in automated job recruitment where racial bias has been observed. 29 Yim et al 30 investigated the ability of a DenseNet-169 AI model in correctly positioning landmarks on lateral cephalograms and managed to achieve an AUC score of 0.95 in both the internal and external (validation) testing datasets. However, the sensitivity scores were lower than the speci city scores by about 5-8% across all skeletal and dental discrepancies.…”

Three-Dimensional Facial Soft Tissue Changes After Orthognathic Surgery in Cleft Patients Using Artificial Intelligence-Assisted Landmark Autodigitization

“…The artificial intelligence-assisted automated detection of the landmarks and evaluation of the skeletal and dental discrepancies is getting popular. 23,[26][27][28] However, it is still necessary to develop the artificial intelligence-assisted virtual surgical planning to reduce clinician's laborious and time-consuming works, to simulate the amount and direction of the required surgical movement, and to provide a preview of surgical outcome. However, for accurate prediction of the soft tissue change, it is mandatory to check surgical accuracy between virtually surgical simulation and actual surgical movements of the maxilla and mandible in the 3D coordinates.…”

Accuracy and clinical validity of automated cephalometric analysis using convolutional neural networks

“…CNNs have been applied to cephalometric analysis to automatically detect cephalometric landmarks and analyze patients' facial skeletal, dental, and soft tissue patterns. 2,3 Since the first application of computer vision technologies to automatically identify the sella and menton in lateral cephalograms in 1984, 4 algorithms for automated cephalometric analysis have evolved. Owing to the increased capacity of computing power to analyze and extract features from a large amount of data in a short amount of time, various AI-based cephalometric analysis algorithms have been developed, with mean error values ranging from 1.1 to 4.09 mm.…”