Comparison of automated algorithm-based versus manual dental landmark placement for model analysis and evaluation of orthodontic quality indices.

BACKGROUND: Digital approaches have fundamentally transformed orthodontics. This study compares an automated, rule-based algorithm with manual landmark placement, evaluating the results of manual versus semi-automated orthodontic quality indices. METHODOLOGY: This retrospective method-comparison study evaluated 100 digital models from 50 orthodontic patients using the Onyxceph3™ diagnostic software. Primary automated landmark placement was assessed for four predefined dental landmark types across 14 tooth groups and was then compared with expert-based manual landmark placement, which was used as the reference standard. Deviations were quantified using Euclidean distances and signed coordinate differences along the X-, Y- and Z-axes. Secondly, we evaluated semi-automated versus manual orthodontic quality assessment using the ABO eCRE and PAR index. RESULTS: Cusps and incisal landmarks show low levels of error when using automatic placement compared with manual placement, with a threshold of ≤1 mm. Vestibular marginal gingival landmarks show the highest deviation. Across tooth groups, mean deviations ranged from 0.54 mm for second molars (95% CI 0.49-0.59) to 0.83 mm for canines (95% CI 0.80-0.85). There was no relevant global directional bias in the X-, Y- and Z-axes. Secondary analysis revealed mean differences of 1.90 points and 3.98 points for ABO eCRE, and -0.08 points and 0.74 for PAR in initial and final models, respectively. CONCLUSION: The automatic placement of landmarks generally showed acceptable deviations, with the exception of vestibular marginal gingival landmarks. However, the automatic detection of landmarks and orthodontic indices should always be reviewed by a clinician.