科研之友
微信
新浪微博
Facebook
分享链接
摘要
Subjects The sample consisted of a data set of 12 cephalometric variables and additional 6 indices, collected from 156 orthodontic patients (learning set, n=96; test set, n=60). Key exposure/study factor Subject selection 1) Exclusion criteria included those who exhibited unerupted permanent teeth or missing teeth (except for third molars), malformed teeth, previous histories of orthodontic treatment and/or orthognathic surgery, and the presence of maxillofacial deformities. 2) Inclusion criteria included those who were judged as in need of undertaking one of the following 5 treatment plans, ie, nonextraction, maxillary and mandibular first premolar extractions (Ext_type_44-44), maxillary and mandibular second premolar extractions (Ext_type_55-55), maxillary first premolar and mandibular second premolar extractions (Ext_type_44-55), and maxillary first premolar extractions only (Ext_type_44-00). 3) For all subjects, the treatment plans were determined by 1 orthodontic specialist. Feature elements used in prediction modeling 1) Lateral cephalometric measures-12 variables that characterized the sagittal and vertical jaw relationships, incisor inclinations, and soft tissue profiles are ANB angle, overjet, Bjork sum, overbite, maxillary central incisor to SN angle, maxillary central incisor to occlusal plane angle, IMPA, mandibular central lip to E-line, lower lip to E-line, and nasolabial angle. 2) Additional 6 measures-the maxillary arch length discrepancy index, the mandibular arch length discrepancy index, the molar key index, the large overjet index, the protrusion index, and the chief complaint for protrusion index. Multimodal elements should be clarified. Calculation cut-off criterion Iterative learning was stopped at the minimum error point of the validation set to prevent overfitting. Output data 1) The extraction patterns for premolar teeth were sub-categorized into 1 of the following 3 bits: (a) Dx_ext-whether or not the extractions were needed. (b) Dx_diff-whether differential extractions between the maxillary and mandibular arches were needed or not. (c) Dx_more-whether more retraction of anterior teeth was needed. 2) For comparison with an actual diagnosis, the decision-making success rates of Dx_ext, Dx_diff, and Dx_more were calculated. Finally, the total success rate of the system's recommendations for extractions was calculated. Main outcome measure In the judgment of requiring extraction vs nonextraction of premolars for orthodontic treatment, the decision-making success rates were 92% in the training set, 94% in the validation set, 93% in the test set, and 93% in total. In the judgment of identical vs differential extraction, the success rates were 88% in the training set, 100% in the validation set, 85% in the test set, and 89% in total. In the judgment of more retraction in identical extraction, the success rates were 88% in the training set, 75% in the validation set, 85% in the test set, and 84% in total. In the decicion-making of more retraction in differential extraction, the success rates were 95% in the training set, 100% in the validation set, 95% in the test set, and 96% in total. Through the sequential application of decision-making models, the final success rates were 85% in the learning set, 82% in the test set, and 84% in total. Six cases were reversed between Ext_type_55-55 and nonextraction. In the 25 cases of failed diagnosis, unacceptable decisions were found in 4 cases. The decisions for the other cases were acceptable because they were borderline. Conclusions The authors concluded that the success rates of the models were 93% for the system's recommendations for extraction vs nonextraction and 84% for the detailed recommendations for the extraction patterns. It was suggested that artificial intelligence expert systems with neural network machine learning could be useful in orthodontics and that improved performance was achieved by components such as proper selection of the input data, appropriate organization of the modeling, and preferable generalization.
- 出版日期2016-9
