Classification Algorithms for Early Tooth Demineralization Assessment by Impedance Spectroscopy.

Area of Science:

• Biomedical Engineering
• Dental Diagnostics
• Artificial Intelligence in Healthcare

Background:

• Oral caries, a prevalent global oral disease, begins with enamel demineralization and can lead to tooth loss.
• Current diagnostic methods for demineralization lack sensitivity, rely on operator expertise, and may involve ionizing radiation.
• There is a significant need for non-invasive, rapid, and accurate early detection tools for oral caries.

Purpose of the Study:

• To investigate the feasibility of impedance spectroscopy for non-invasive, radiation-free early detection of oral caries.
• To develop and validate AI-based classifiers for automated assessment of demineralization using impedance data.
• To create a prototype portable diagnostic tool for early caries detection.

Main Methods:

• In vitro demineralization of human teeth was performed to obtain impedance spectroscopy data.
• Two binary classifiers were developed: a single-neuron classifier using a 15 Hz impedance phase measurement and a Multi-Layer Perceptron (MLP) using equivalent circuit element values.
• A prototype diagnostic tool was built to measure impedance phase reliably.

Main Results:

• The single-neuron classifier achieved 88% accuracy, offering a low-power solution for rapid measurements.
• The MLP classifier achieved 86% accuracy.
• The prototype tool demonstrated reliable impedance phase measurements with uncertainty less than 2°.
• Classifier performance met or exceeded existing AI-based methods using radiographic data.

Conclusions:

• Impedance spectroscopy, combined with AI, presents a novel and effective approach for early oral caries detection.
• This non-invasive, radiation-free method offers a promising alternative to current diagnostic techniques.
• The developed technology lays the foundation for an accessible and accurate tool for early caries management.