Editing AI-Results and Understanding AI Confidence Levels

Introduction

Dx allows users to edit AI-generated results, providing full control over the classification. This guide will walk you through the steps to edit AI results and understand the AI confidence levels provided by the software. AI confidence levels reflect how certain the AI system is about its predictions. These levels help dental professionals interpret AI-generated findings and support clinical decision-making. Understanding how these confidence levels are derived and validated helps build trust in the system.

Editing AI-Results

Dx puts you in full control. If you wish to adjust an AI-finding, simply click on it to bring up the edit pop-up, where you can edit the classification or remove the result entirely.

Steps: 

1. Click a detected oral health condition on the scan (e.g. surface caries highlighted in yellow) 
2. In the pop-up window, click Edit to display the available classification categories.

3. In the drop down menu, do one of the following:  

a. Change classification: Choose a new classification category  

b. Delete result: Delete the assigned classification by clicking the delete icon

4. Click X to close the pop-up window

Dx will update the classification category according to the selected option and refreshes the scan to reflect the change.

AI Confidence Level

The AI Confidence level is a tool that provides a measure of how sure the AI is about what it sees in a scan. The higher the confidence, the more likely the AI is correct — but it’s still essential to confirm findings clinically.

Where to find it

Confidence levels are shown alongside AI-detected conditions surface caries, tooth wear, gingival recession, and plaque. By clicking on the individual indication, the AI confidence level is indicated via a bar.

Why it matters

These levels help you prioritize what to examine more closely. They are not a replacement for your clinical judgment but a tool to support it.

How to Interpret Accuracy Metrics

Each AI system in Dx has been validated against expert benchmarks. The following metrics are used to describe how well the AI performs:

• Sensitivity: The percentage of actual cases the AI correctly identifies (true positives). For example, if sensitivity is 85%, the AI detects 85 out of 100 true cases.
• Specificity: The percentage of non-cases the AI correctly ignores (true negatives). A specificity of 90% means the AI correctly rules out 90 out of 100 healthy cases.
• Limits of Agreement (LoA): The interval within 95% of the differences between AI and reference measurements are expected to fall.

These metrics help you understand the strengths and limitations of each AI system and how much weight to give its findings in your clinical workflow.

Accuracy Metrics by AI System

Next Steps

• Learn how to share a Dx scan with a patient via the DentalHealth app here
• Learn how to generate a Dx report here