Influence of head positioning errors on the accuracy of fully automated artificial intelligence-based cephalometric...

Area of Science:

• Orthodontics
• Medical Imaging Analysis
• Artificial Intelligence in Healthcare

Background:

• Cephalometric analysis is vital in orthodontics for diagnosis and treatment planning.
• Traditional manual cephalometric analysis is time-consuming and subject to inter-observer variability.
• Automated software aims to improve efficiency and consistency in cephalometric analysis.

Purpose of the Study:

• To assess the accuracy of three automated artificial intelligence (AI) cephalometric software systems.
• To compare the performance of AI software against manual cephalometric analysis.
• To evaluate the impact of head position errors on automated cephalometric analysis accuracy.

Main Methods:

• Retrospective analysis of 40 lateral cephalograms with correct and incorrect head postures.
• Manual cephalometric landmarking using Dolphin software as the reference standard.
• Evaluation of three AI software systems: WebCeph, Ceph Assistant, and AudaxCeph.

Main Results:

• Automated software showed good consistency for angular measurements but were more error-prone for linear measurements.
• AudaxCeph demonstrated the highest accuracy for skeletal parameters but struggled with soft tissue analysis.
• Positional errors in cephalograms significantly reduced measurement accuracy across all tested AI software.

Conclusions:

• AI-based cephalometric software accuracy varies based on measurement type and is sensitive to cephalogram positional errors.
• Expert clinical oversight is essential to mitigate errors and ensure reliable cephalometric analysis.
• Further development is needed to enhance the robustness of AI cephalometric tools.