The impact of measurement metrics on data robustness and repeatability when measuring 3D change

Tooth wear alters the tooth surface topography, which can be analysed through sequential intraoral scans. This study investigated the Coefficient of Variation (CoV) and agreement of three outcome measures of topographical change: volume loss, mean profile loss, and maximum point loss on single-surface wear and multi-surface wear.

An upper full arch typodont was scanned and manipulated to generate representative mild, moderate, and severe wear changes involving single-surface (non-carious cervical lesions) or multi-surface (palatal/incisal erosion) features (n = 36 teeth). Differences between each scan were analysed three times per metric using two software. Data were normal. CoV of the metric output within software per wear type was assessed. Metric agreement between software was assessed using correlations, paired t-tests and Bland-Altman plots.

A range of features were generated on the datasets from 1.34 mm³ on representative mild wear cases to 72.13 mm³ on severe wear cases. The CoV for both software was low for mean profile and maximum point loss. CoV was higher for volumetric measurements, particularly for multi-surface compared to single-surface measurements (Geomagic 3.10 %, 0.53 %, WearCompare 8.53 %, 4.71 % respectively). Measurements between software were significantly correlated (ICC > 0.98, p < 0.001). Mean profile and maximum point loss had overall good agreement, with mean difference lines close to zero. However, statistical differences (p = 0.006) and low levels of agreement (95 % LoA 13.3, -5.80) were observed for multi-surface volumetric measurements.

Both maximum point loss and mean profile loss demonstrated consistent low variation and high agreement. As wear and surface complexity increased, volume measurements showed higher variation and lower agreement.

Volume change displayed the highest variability and methodological challenges, especially in complex multi-surface measurements. Mean profile and maximum point loss showed superior repeatability and agreement across platforms but present a limited view of the change that has occurred. Further work is required to ensure volumetric measurements are robust.