Automated AI Classification of Midpalatal Suture Maturation Stages Using Deep Learning Framework

Knowing the developmental status of the midpalatal suture helps orthodontists treatment plan for maxillary transverse discrepancy. Current methods used to assess the maturity of the midpalatal suture include evaluation of cone beam computed tomography (CBCT) imaging, which can provide valuable insights into the anatomy of the suture. However, diagnosis of midpalatal suture maturation remains a tedious task that has poor reproducibility among evaluators. Artificial intelligence can reduce human error when evaluating the development of the suture, helping clinicians treatment plan appropriately. This study proposes a deep learning framework to classify midpalatal suture maturation stages using CBCT images. 1200 images were collected and oriented in three planes on Dolphin Imaging Software. After removing images with poorly defined sutures, the final sample consisted of 618. The midpalatal suture for each CBCT was annotated into five developmental stages (A, B, C, D, E) by expert evaluators. For each axial section, the suture regions were extracted and pre-processed with dynamic filtering and data augmentation before being fed into multiple pre-trained convolutional neural network (CNN) models. These baseline architectures were modified to improve classification accuracy, resulting in a novel multi-filter convolutional residual attention network (MFCRAN) architecture enhanced with discrete cosine transform (DCT) layers. Our method was validated with a k-fold cross validation protocol. The novel architecture demonstrated a classification accuracy of 79.02%, outperforming other methods and marking a substantial advancement in orthodontic diagnosis and treatment planning.

Author dc:creator

• Lauren Mills (19805210)

• Orthodontics
• Computer engineering
• Artificial Intelligence
• Growth and development

Statement dc:rights

• In Copyright

DOI dc:identifier

https://doi.org/10.25417/uic.32991818.v1

OAI identifier oai:identifier

oai:figshare.com:article/32991818