Comparison between a human judge and automatic landmark identification on digital models.

Area of Science:

• Biomedical image analysis
• Dental imaging technology
• Orthodontic diagnostics

Background:

• Accurate landmark identification is crucial for reliable biomedical image analysis.
• Orthodontic study models are essential for diagnosis, treatment planning, and record-keeping.
• Evaluating automated tools for landmark location in dental imaging is necessary.

Purpose of the Study:

• To assess the reliability and validity of a software program for automatic landmark identification.
• To compare software-based landmark location with human expert identification in orthodontic models.

Main Methods:

• Digital intraoral scans of 10 dental arches were analyzed.
• Four calibrated judges manually identified landmarks (cusp tips, interproximal contacts).
• Software automatically identified the same landmarks; agreement was assessed using statistical analysis (ICC, Bland-Altman plots).

Main Results:

• Excellent intraclass correlation coefficients (ICC) were found between software and human judges for x, y, and z coordinates (1.0, 1.0, 0.98).
• Regression analysis and Bland-Altman plots indicated no systematic errors in landmark identification.
• Software-based landmark identification demonstrated high agreement with human expert identification.

Conclusions:

• The automatic landmark location software is highly accurate and reliable.
• The software's performance is nearly identical to that of human judges.
• The software and human landmark identification methods are interchangeable for orthodontic applications.