Correlation Between Quality Evaluation Metrics and Teeth Detection Results in Panoramic X-Rays Using Deep Learning.

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Panoramic radiography is a common dental examination for visualizing the entire mouth.
• Interpreting these X-rays can be time-consuming and prone to diagnostic errors due to professional fatigue or inexperience.

Purpose of the Study:

• To investigate the impact of various image enhancement techniques on the accuracy of teeth detection in panoramic radiography using deep learning.
• To identify which image features, after enhancement, correlate with improved object detection performance.

Main Methods:

• Applied five different object-detection deep learning architectures to a dataset of 300 panoramic images.
• Evaluated the effectiveness of various image enhancement techniques as a pre-processing step.
• Computed cross-correlation between image quality metrics and detection performance for each architecture.

Main Results:

• Observed a significant dependence of teeth detection performance on specific image enhancement techniques.
• Techniques that minimized noise introduction and preserved global image contrast demonstrated the most beneficial impact on detection accuracy.
• Different deep learning architectures showed varying sensitivities to the applied pre-processing methods.

Conclusions:

• Image enhancement as a pre-processing step can significantly improve the accuracy of automated teeth detection in panoramic dental images.
• The choice of enhancement technique is crucial, with noise reduction and contrast preservation being key factors for optimal performance.
• Further research can leverage these findings to develop more robust AI-driven diagnostic tools for dentistry.