Evaluating the accuracy of automated cephalometric analysis based on artificial intelligence

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Cephalometric analysis is crucial in orthodontics and maxillofacial surgery.
• Accurate landmark identification and measurement are essential for diagnosis and treatment planning.
• Traditional manual cephalometric analysis is time-consuming and prone to inter-observer variability.

Purpose of the Study:

• To evaluate the accuracy of artificial intelligence (AI)-based automatic cephalometric landmark localization and measurements.
• To compare AI automatic analysis with computer-assisted manual analysis for cephalometric evaluations.

Main Methods:

• Reconstructed lateral cephalograms (RLCs) from 85 CBCT scans were analyzed.
• 19 landmarks were digitized and 23 measurements were obtained using both computer-assisted manual (Dolphin Imaging) and AI automatic (Planmeca Romexis) software.
• Mean radial error (MRE) and successful detection rate (SDR) were calculated to assess accuracy.

Main Results:

• The AI automatic program achieved a Mean Radial Error (MRE) of 2.07 ± 1.35 mm for 19 landmarks.
• Successful detection rates (SDR) within 2 mm and 4 mm were 58.58% and 91.39%, respectively.
• 15 out of 23 measurements met the clinical accuracy criteria (±2 mm or 2°), with consistency rates above 90%.

Conclusions:

• AI-driven cephalometric analysis demonstrates sufficient effectiveness for clinical application.
• Complete replacement of manual tracing by automatic cephalometry is not yet feasible.
• Manual supervision and adjustments can enhance the accuracy and efficiency of AI-based cephalometric analysis.