AI-initiated second opinions: a framework for advanced caries treatment planning

Area of Science:

• Dentistry
• Artificial Intelligence
• Medical Diagnostics

Background:

• Clinician distrust and risks of AI errors hinder AI integration in healthcare.
• Existing diagnostic methods for advanced caries may lack optimal accuracy.

Purpose of the Study:

• To develop and evaluate an AI-driven second opinion framework for predicting pulp status in advanced caries.
• To mitigate risks associated with AI distrust and erroneous predictions in clinical settings.

Main Methods:

• An AI model was developed to predict pulp status post-caries treatment.
• Clinicians (dentists and students) made predictions independently.
• A novel framework used AI to trigger second opinions when clinician-AI predictions disagreed, keeping AI results hidden.
• A simulation tested the framework over 1000 trials.

Main Results:

• The AI-assisted second opinion framework significantly improved clinician prediction accuracy.
• The average F1-score for clinicians increased from 0.586 to 0.645.
• The system effectively leveraged AI to enhance human diagnostic performance.

Conclusions:

• An AI-triggered second opinion system can enhance clinical decision-making in dentistry.
• This approach successfully addresses clinician distrust by prioritizing human judgment.
• The framework shows promise for improving diagnostic accuracy in dental applications of AI.