Reliability and accuracy of Artificial intelligence-based software for cephalometric diagnosis. A diagnostic study

Area of Science:

• Orthodontics
• Medical Imaging
• Artificial Intelligence

Background:

• Artificial intelligence (AI) is transforming orthodontic diagnostics.
• AI streamlines patient assessments using cephalometric analysis.
• This study evaluates AI software against conventional methods for 2D lateral cephalograms.

Purpose of the Study:

• To assess the reliability and accuracy of AI-based cephalometric analysis software.
• To compare AI software performance (automatic and semi-automatic) with conventional digital techniques.
• To evaluate the time efficiency of AI-driven cephalometric tracing.

Main Methods:

• Analyzed 408 lateral cephalograms using manual, automatic AI, and semi-automatic AI landmark localization.
• Measured 15 skeletal, dental, and soft tissue variables.
• Compared AI software accuracy and time consumption against conventional digital methods using Student's t-test.

Main Results:

• Statistically significant differences in landmark positioning accuracy were found (p < 0.01), but most were not clinically significant.
• Semi-automatic AI showed slightly less accuracy than conventional methods but was faster.
• Automatic AI was the fastest method, followed by semi-automatic AI and conventional tracing (p < 0.000).

Conclusions:

• AI software demonstrates statistically significant but generally not clinically significant accuracy differences compared to conventional cephalometric analysis.
• Semi-automatic AI offers a balance of accuracy and speed, outperforming automatic AI in accuracy and conventional methods in speed.
• Further research is recommended to validate AI's role in precise cephalometric tracing.