Feasibility of iterative closest point algorithm for accuracy between virtual surgical planning and orthognathic surgery outcomes.

Purpose: To evaluate the feasibility of iterative closest point (ICP) algorithm for assessing the accuracy between virtual surgical planning (VSP) and outcomes in orthognathic surgery.

Materials And Methods: VSP and results of surface mesh (SM0 and SM1) from CBCT scans of 25 patients who had been undergone bi-maxillary orthognathic surgery were converted into STL-format files and then imported to Geomagic software for semi-automatic alignment. ICP algorithm was used to calculate mean deviations (MD) and root mean square (3D error) at different calibrations of ±2 mm (T1), ±5 mm (T2) and ±10 mm (T3), with workflow being performed by two evaluators. Colour maps were generated to assess the 3D congruence qualitatively. Linear regression analysis was used to estimate whether SM0 or SM1 could condition the ICP and t-tests were used to assess whether MD and 3D error values were ≤-2 mm and ≥2 mm. Descriptive statistics was used to assess the method's feasibility by comparing T2 to T1 and T3.

Results: High intra- and inter-rater correlations supported the workflow reproducibility with the software. SM0 conditioned the ICP algorithm regarding both evaluators, and t-tests demonstrated that MD and 3D error were >-2 mm and <2 mm. MD and 3D error at T3 were 30% higher than those at T1.

Conclusions: ICP algorithm provided a reproducible method, but its feasibility was limited due to underestimation or overestimation of the results as they affect the validity of the actual deviations.