Detection of dental treatments performed on paediatric patients using artificial intelligence algorithms developed with deep learning methods on panoramic images

Aim: This study aims to compare the detection and classification capabilities of artificial intelligence models developed using deep learning methods with different classification techniques for various types of dental treatments found in panoramic radiographs. Materials and Methods: A dataset consisting of 3001 panoramic radiographs obtained from paediatric patients was labelled using the CranioCatch software (CranioCatch, Eskişehir, Turkey) for six different types of dental treatment: restorative filling treatment, amputation treatment, root canal treatment, stainless steel crown application, fixed space maintainer application, and braces. After the labelling was completed, the success metrics of artificial intelligence models developed using single- class and multi-class classification methods were calculated using the obtained dataset. Results: The AP values of artificial intelligence models developed using single- class and multi-class classification techniques for the relevant dental treatments are as follows: 0.842-0.832 for restorative filling treatment, 0.782-0.759 for amputation treatment, and 0.748-0. 788 for root canal treatment, 0.967-0.934 for stainless steel crown application, 0.860-0.910 for fixed space maintainer application, and 0.992-0.993 for bracket application. The artificial intelligence developed using the single-class classification technique demonstrated higher success performance, particularly in classes that were similar in terms of anatomical location and morphology. Conclusion: This study has demonstrated that artificial intelligence models developed using different classification techniques show promising performance in identifying types of dental treatment. In particular, the model approach developed using a multi-class technique has the potential to offer a more comprehensive and inclusive diagnostic framework when integrated into clinical decision support systems. This is due to its capacity to simultaneously evaluate multiple types of dental treatment within a single model, offering dentists advantages in terms of time and efficiency. Keywords: Artificial intelligence, panoramic radiography, pediatric dentistry