Cephalometric analysis performance discrepancy between orthodontists and an artificial intelligence model using...

Area of Science:

• Dentistry
• Orthodontics
• Medical Imaging

Background:

• Cephalometric analysis is crucial in orthodontics for diagnosis and treatment planning.
• Variability in manual cephalometric analysis can affect treatment outcomes.
• Artificial intelligence (AI) offers potential for automating and standardizing these measurements.

Purpose of the Study:

• To compare cephalometric analyses (Ricketts and Steiner) performed by experienced orthodontists and an AI software program.
• To measure orthodontist variability in cephalometric landmark identification and measurements.
• To evaluate the accuracy and potential of AI in cephalometric analysis.

Main Methods:

• 50 lateral cephalometric radiographs were analyzed.
• Two groups: orthodontists (manual landmark identification + software) and AI software (automatic landmark identification + measurements).
• Ricketts and Steiner analyses (24 measurements total) were performed; data analyzed using t-test.

Main Results:

• Significant discrepancies (p<0.05) were found between orthodontists and AI in 7 of 24 measurements.
• Specific discrepancies noted in Ricketts analysis (corpus length, mandibular arc, lower face height, overjet, overbite) and Steiner analysis (occlusal to SN, GoGn-SN).
• High intra-operator reliability was observed between orthodontists, except for corpus length.

Conclusions:

• AI software demonstrated potential for automated cephalometric analysis.
• Significant measurement differences exist between AI and orthodontist assessments.
• Further research is required to validate the accuracy of AI-based cephalometric systems.