In our last newsletter, we discussed the potential of artificial intelligence (AI) in the early detection of periodontal diseases and outlined the processes an AI model uses to identify symptoms of periodontal issues using intraoral camera (IOC) images. In this edition, we will explore AI models for automated cavity detection.
According to a World Health Organization (WHO) report from 2012, 60 to 90 percent of school-aged children and nearly all adults have experienced dental cavities. The incidence of smooth surface cavities has decreased the most due to their visibility on X-rays. Traditionally, dentists detect cavities by examining the tooth’s color and hardness and using X-rays. However, radiation-free imaging methods, such as intraoral cameras, offer a beneficial alternative for cavity detection.
Dental cavities are among the most prevalent chronic diseases, and early detection is crucial for effective management. Recent studies have focused on developing automated systems to detect cavities using intraoral images. These systems identify tooth boundaries and irregular regions and extract features from each image. These features may include statistical measures of color, grayscale, and advanced techniques such as Wavelet and Fourier Transforms.
A study by Leila Ghaedi et al. (2014) developed an automated system that detects and scores dental cavities by analyzing optical images of the tooth surface captured with an intraoral camera. The system outputs an ICDAS (International Caries Detection and Assessment System) score, which assesses cavity severity. ICDAS scores range from 0 (healthy tooth) to 6 (extensive cavity with visible dentin).
The proposed computational method begins by segmenting the tooth image into distinct components: the background, healthy enamel surface, and irregular regions. For dentists, these irregular regions are of particular interest due to variations in color, translucency, and porosity.
In this study, irregular regions are identified using spatial statistics and texture analysis. Incorporating texture information enables the system to detect not only visible changes in the enamel but also subtle textural changes typically identified through tactile examination. These features assist in determining the presence and severity of cavities in the irregular regions. The steps to automate dental cavity detection include:
- Segmentation: involves partitioning tooth surface images into components such as the background, normal tooth surface, and irregular areas, adhering to cavity detection guidelines.
- Feature Extraction: involves extracting relevant features from the tooth image. The method focuses on identifying the most significant features for automatic cavity scoring.
- Classification: a classification technique is employed to assign clinical scores based on the extracted features. An ensemble classifier, incorporating four distinct classification methods, is utilized to develop the model.
This study presents a system that eliminates the need for manual landmarking through advanced feature extraction methods, enhancing reliability in early cavity detection. The findings suggest that optical images hold promise for cavity diagnosis in clinical settings. However, further research is necessary to test and validate the system with larger datasets and in vivo samples.
