Tufts Dental Database: A Multimodal Panoramic X-Ray Dataset for Benchmarking Diagnostic Systems

Area of Science:

• Artificial Intelligence in Healthcare
• Medical Imaging Analysis
• Dental Diagnostics

Background:

• Dental radiographs are crucial for diagnosis, with AI, particularly Convolutional Neural Networks, showing promise in analysis.
• Existing datasets lack comprehensive expert annotations and multimodal data, hindering AI development in dentistry.

Purpose of the Study:

• Introduce the Tufts Dental Database, a novel multimodal dataset of 1000 panoramic dental radiographs.
• Facilitate the integration of human expertise into AI for improved dental abnormality detection.
• Establish benchmark performance for AI systems in dental radiograph analysis.

Main Methods:

• Curated a dataset of 1000 panoramic dental radiographs with expert labels for abnormalities and teeth.
• Classified images based on anatomical location, peripheral characteristics, radiodensity, surrounding effects, and abnormality category.
• Captured radiologist expertise using eye-tracking and think-aloud protocols.

Main Results:

• Presented the Tufts Dental Database, a publicly available multimodal resource.
• Provided benchmark performance analysis for deep learning-based dental radiograph enhancement and segmentation.
• Conducted an in-depth review of existing dental image datasets and AI systems.

Conclusions:

• The Tufts Dental Database will accelerate the development of AI for automated abnormality detection and classification in dental radiographs.
• This resource aims to enhance tooth segmentation algorithms and distill radiologist expertise into AI models.
• The dataset supports research in creating more robust and accurate AI tools for dental healthcare.