# Application of Artificial Intelligence in Detecting Dental Anomalies: Current Models, Imaging Modalities, and Future Directions

## TL;DR

This paper reviews how AI models can detect dental anomalies in imaging, showing high accuracy and pointing to future improvements.

## Contribution

The study evaluates current AI models for detecting various dental anomalies and highlights the need for multi-class models.

## Key findings

- Deep learning models like EfficientDet-D3, nnU-Net, and ResNeXt achieved over 85% accuracy for specific dental anomalies.
- AI performance varied across anomalies and imaging types, indicating a need for optimization.
- Future research should focus on multi-class AI models integrating clinical and radiographic data.

## Abstract

As dental anomalies can significantly affect esthetic and function, early detection and diagnosis are crucial for treatment and minimizing potential negative effects. Artificial intelligence (AI) has emerged as a promising tool for the segmentation and detection of dental anomalies in number, morphology, size, position, and structure that may be missed by dentists. This study aimed to investigate the application of various AI models in dental anomaly detection and diagnosis, including supernumerary teeth, tarodontism, impaction, ectopic eruption, and molar‐incisor hypomineralization in both dental radiography and photography.

A comprehensive literature search was conducted in PubMed/Medline, Scopus, Web of Science, and Google Scholar for studies published from the initiate up to 2023 on AI applications in dental anomaly detection. Inclusion criteria encompassed recent AI models utilizing imaging modalities for identifying dental abnormalities, with full‐text availability in English. Studies lacking imaging‐based AI applications or methodological clarity were excluded.

A total of 20 studies assessed various AI models for detecting dental anomalies in radiographic and photographic imaging. Deep learning models, particularly EfficientDet‐D3, nnU‐Net, and ResNeXt, demonstrated the highest accuracy for supernumerary teeth, ectopic eruption, and molar‐incisor hypomineralization, respectively, with most models achieving accuracy rates above 85%. These findings underscore AI's significant potential for automated dental anomaly detection; however, performance varied across different anomalies and imaging modalities, highlighting the need for further optimization. Given the complexity of simultaneous dental anomalies, future research should focus on developing multi‐class AI models capable of detecting multiple conditions concurrently and integrating clinical and radiographic data for improved diagnostic accuracy and treatment planning.