Teeth segmentation and carious lesions segmentation in panoramic X-ray images using CariSeg, a networks' ensemble

Area of Science:

• Artificial Intelligence in Dentistry
• Medical Imaging Analysis
• Computational Pathology

Background:

• Dental cavities are prevalent oral diseases with significant consequences, including tooth loss.
• Early detection and intervention are crucial for managing cavities and preventing adverse outcomes.
• The development of advanced diagnostic tools is essential for improving dental care.

Purpose of the Study:

• To introduce CariSeg, an intelligent system designed for the accurate detection of dental cavities in X-ray images.
• To leverage deep learning and neural networks for automated cavity identification.
• To enhance the early diagnosis of dental caries, thereby improving patient treatment and outcomes.

Main Methods:

• CariSeg employs a multi-stage approach utilizing four neural networks.
• The initial network, based on U-Net architecture, performs tooth segmentation and image cropping.
• Subsequent carious lesion segmentation is achieved through an ensemble of U-Net, Feature Pyramid Network, and DeeplabV3 architectures, trained on diverse dental radiography datasets.

Main Results:

• The CariSeg system achieved a high accuracy rate of 99.42% in detecting dental cavities.
• The system demonstrated a mean Dice coefficient of 68.2% for carious lesion segmentation.
• Experimental results validate the efficacy of the proposed AI approach in identifying cavities from dental X-rays.

Conclusions:

• Artificial intelligence, exemplified by CariSeg, significantly aids in detecting carious lesions by analyzing dental X-rays.
• AI systems can identify cavities potentially missed by human observation, facilitating earlier diagnosis and treatment.
• The implementation of AI in dental diagnostics promises improved oral health outcomes through timely intervention.