Artificial neural network in orthodontic therapeutic extractions

An unsubstantiated decision could result in several problems in its course as the extraction of teeth is an irreversible process. Therefore, it is of interest to develop an artificial intelligence decision making model for the diagnosis of extractions using neural network machine learning. The sample included 455 patients wherein input data consisted of 12 cephalometric variables and two additional indexes obtained from patients' records which were manually traced and digitized using NemoCeph 2D version 10software. It was observed that the accuracy of the binary classifier model, i.e., the decision of whether to extract or not, was 92.38 % and that of the multi- classifier model, i.e., the decision of which tooth to extract was also 92.38 %. Thus, we show that the technique of predicting orthodontic extractions using an artificial neural network is a reliable and valuable method.