AggregateNet: A deep learning model for automated classification of cervical vertebrae maturation stages

Abstract

ObjectiveA study of supervised automated classification of the cervical vertebrae maturation (CVM) stages using deep learning (DL) network is presented. A parallel structured deep convolutional neural network (CNN) with a pre‐processing layer that takes X‐ray images and the age as the input is proposed.MethodsA total of 1018 cephalometric radiographs were labelled and classified according to the CVM stages. The images were separated according to gender for better model‐fitting. The images were cropped to extract the cervical vertebrae automatically using an object detector. The resulting images and the age inputs were used to train the proposed DL model: AggregateNet with a set of tunable directional edge enhancers. After the features of the images were extracted, the age input was concatenated to the output feature vector. To have the parallel network not overfit, data augmentation was used. The performance of our CNN model was compared with other DL models, ResNet20, Xception, MobileNetV2 and custom‐designed CNN model with the directional filters.ResultsThe proposed innovative model that uses a parallel structured network preceded with a pre‐processing layer of edge enhancement filters achieved a validation accuracy of 82.35% in CVM stage classification on female subjects, 75.0% in CVM stage classification on male subjects, exceeding the accuracy achieved with the other DL models investigated. The effectiveness of the directional filters is reflected in the improved performance attained in the results. If AggregateNet is used without directional filters, the test accuracy decreases to 80.0% on female subjects and to 74.03% on male subjects.ConclusionAggregateNet together with the tunable directional edge filters is observed to produce higher accuracy than the other models that we investigated in the fully automated determination of the CVM stages.

Mapping an intelligent algorithm for predicting female adolescents’ cervical vertebrae maturation stage with high recall and accuracy

“…Kök H et al trained the artificial neural networks (ANN) model with 300 individuals aged between 8 and 17 years and proposed the ANN algorithm was stable in determining the CVM classes with a 2.17 average rank [ 32 ]. With ongoing improvements in AI, Atici SF et al proposed an innovative deep-learning model with parallel architecture and achieved a validation accuracy of 82.35% in CVM stage classification on female subjects [ 33 ]. However, applying deep learning algorithms to medical image analysis presents several unique challenges and obstacles, such as the issues of imbalanced data, deficiency in the confidence interval, and lack of properly labeled data [ 34 , 35 ].…”

Artificial Intelligence in Orthodontics: Critical Review

“…AI-based classification of vertical and sagittal skeletal patterns has been found highly accurate, with a mean area under the receiver operating characteristic curve (ROC AUC) of >95% (Yu et al 2020). In a similar fashion, lateral cephalograms were used for end-to-end predictions of the upper airway obstruction (Jeong et al 2023), degenerative temporomandibular joint diseases (Fang et al 2023), or the classification of cervical vertebrae maturation stages (Atici et al 2023). Notably, such end-to-end classifications are not inherently explainable, that is, humans cannot easily identify why a certain class was chosen, something we critically discuss below.…”

Artificial Intelligence and Its Clinical Applications in Orthodontics: A Systematic Review

“…The assessment of skeletal maturity is very important in treatment planning for optimal treatment timing. A custom-design CNN model consisting of two parts, feature extraction and classification, was used to classify CVM into six maturation stages (CS1-CS6) [50,51]. AggregateNet was utilized in the model for feature extraction, and as the preprocessing layer, it used directional filters to enrich the information.…”

“…AggregateNet was utilized in the model for feature extraction, and as the preprocessing layer, it used directional filters to enrich the information. The AggregateNet output feature was coupled with age input to produce conclusions as well as to boost network performance [51]. Hatice Kök et al [52] used seven algorithms to determine CVS: k-nearest neighbors (k-NN), naive Bayes (NB), decision tree (DT), artificial neural network (ANN), support vector machine (SVM), random forest (RF), and logistic regression (LR).…”