Anzeige
Erschienen in:
20.01.2022 | Original Article
Assessment of an Artificial Intelligence Mandibular Osteotomy Design System: A Retrospective Study
Erschienen in: Aesthetic Plastic Surgery | Ausgabe 3/2022
Einloggen, um Zugang zu erhaltenAbstract
Background
In this study, an AI osteotomy software was developed to design the presurgical plan of mandibular angle osteotomy, which is followed by the comparison between the software-designed presurgical plan and the traditional manual presurgical plan, thus assessing the practicability of applying the AI osteotomy software in clinical practices.
Methods
(1) Develop an AI osteotomy software: design an algorithm based on convolutional neural networks capable of learning feature point and processing clustering segmentation; then, select 2296 cases of successful 3D mandibular angle osteotomy presurgical plans, followed by using those 2296 cases to train the deep learning algorithm; (2) compare the osteotomy presurgical plan of AI osteotomy software and that of manual: first step: randomly selecting 80 cases of typical female head 3D CTs, and designing those 80 cases by means of AI osteotomy software designing (group A) and manually designing (group B), respectively; second step: comparing several indexes of group A and those of group B, including the efficiency index (time from input original CT data to osteotomy presurgical plan output), the safety index (the minimum distance from the osteotomy plane to the mandibular canal), the symmetry indexes (bilateral difference of mandibular angle, mandibular ramus height and mandibular valgus angle) and aesthetic indexes (width ratio between middle and lower faces (M/L), mandibular angle and mandibular valgus angle).
Results
The efficiency index: the design time of group A is 1.768 ± 0.768 min and that of group B is 26.108 ± 1.137 min, with P = 0.000; the safety index: The minimum distances from the osteotomy plane to the mandibular canal are 3.908 ± 0.361mm and 3.651 ± 0.437mm, p = 0.117 in groups A and B, respectively; The symmetry indexes: Bilateral differences of mandibular angle are 1.824 ± 1.834° and 1.567 ± 1.059° in groups A and B, respectively, with P = 0.278; bilateral differences of mandibular ramus height are 2.083 ± 1.263 and 2.965 ± 1.433, respectively, with P = 0.119 in groups A and B; Aesthetic indexes: M/L in groups A and B is 1.364 ± 0.074 and 1.371 ± 0.067, respectively, with P = 0.793; mandibular angles in groups A and B are 127.724 ± 5.800° and 127.242 ± 5.545°, respectively, with P = 0.681; Valgus angles in groups A and B are 11.474 ± 5.380 and 9.743 ± 4.620, respectively, with P = 0.273.
Conclusions
With high efficiency, as well as safety, symmetry and aesthetics equivalent to those of a manual design, the AI osteotomy software designing can be used as an alternative method for manual osteotomy designing.
Level of Evidence IV
This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266