Assessment of Elapsed Time Between Dental Radiographs Using Siamese Network

Area of Science:

• Dentistry
• Medical Imaging
• Machine Learning

Background:

• Machine learning for dental disease progression often requires expensive annotated data and shows poor generalization.
• Predicting time intervals between dental treatments is crucial for understanding disease progression patterns.

Purpose of the Study:

• To evaluate Siamese networks for detecting subtle changes in longitudinal dental X-rays.
• To predict time span categories between dental treatments using periapical radiographs and demographic data.

Main Methods:

• Application of Siamese networks to analyze longitudinal periapical radiographs.
• Comparison with baseline Convolutional Neural Networks (CNNs) and Multilayer Perceptron (MLP) models.
• Inclusion of patient demographic features (age, gender) to assess their impact on model performance.

Main Results:

• Siamese network models significantly outperformed baseline CNN and MLP models.
• The top-performing Siamese network achieved 86.32% ± 1.60% accuracy in predicting time intervals.
• Adding demographic features reduced model performance variance, indicating enhanced stability.

Conclusions:

• Siamese networks are effective in capturing subtle temporal changes in dental radiographs for longitudinal analysis.
• These models show potential for integration into clinical workflows for dental disease progression monitoring.
• Future research should focus on self-supervised learning for dental disease progression, especially with limited labeled data.