Abstract:

Introduction: Digital technologies are rapidly entering dentistry and artificial intelligence is incorporated into the diagnostic processes. In orthodontic practice, is has been integrated into the cephalometric analyses in order to improve the diagnostic accuracy, the treatment plan development, and outcome prediction. The automation of the cephalometry still has systematic and random errors. The aim is to establish the accuracy of a fully automatic cephalometric analysis versus automatic one in which the orthodontist controls the accuracy of the reference points.
Materials and methods: The present study utilized lateral teleradiographs of 30 patients before orthodontic treatment. The study includes measuring the difference of six parameters – ∢SNA, ∢SNB, ∢I/SN, ∢i/MP, ∢MP/SN, ∢MP/FH. The cephalometric analysis was first performed by the software. The second analysis was done on the same radiograph with orthodontist controlling the placement of the reference points after the automatic positioning. The analysis was conducted using the Planmeca Romexis Cephalometric Analysis Software and the data was processed IBM SPSS Statistics 25.0 and MedCalc Version 19.6.3.
Results: The results demonstrate variations in the measurements obtained from the two methods, ranging from 0.13º to 2.14º. The indices ∢I/SN – 0.15º and ∢I/MP – 0.13º show the smallest disparities. The most substantial disparities were identified in ∢MP/SN and ∢MP/FH, with the highest difference in the ∢MP/FH – 2.14º.
Conclusion: The obtained values are within the tolerance range for the angles most commonly used in a cephalometric analysis, making automated cephalometry fast and precise enough to enhance the efficiency of the daily orthodontic practice.
Keywords: Artificial intelligence, cephalometry, cephalometric analysis, orthodontic diagnosis