ISSN: 2332-0702

Journal d'hygiène bucco-dentaire et de santé

Accès libre

Notre groupe organise plus de 3 000 séries de conférences Événements chaque année aux États-Unis, en Europe et en Europe. Asie avec le soutien de 1 000 autres Sociétés scientifiques et publie plus de 700 Open Access Revues qui contiennent plus de 50 000 personnalités éminentes, des scientifiques réputés en tant que membres du comité de rédaction.

Les revues en libre accès gagnent plus de lecteurs et de citations
700 revues et 15 000 000 de lecteurs Chaque revue attire plus de 25 000 lecteurs

Abstrait

A Novel Advance for Orthodontic Landmarks Recognition Using an Artificial Neural Network

Ali Mohammad Saghiri

Background: Cephalometric analysis is the clinical application of dental cephalometry. It is investigation of the dental and skeletal connections of a human skull. Cephalometric analysis is one of most difficult part for orthodontic and orthogenetic surgical treatments. Most of time landmark identifications is time consuming and has high dependency to operator. the aim of current investigation is to find a new approach for orthodontic landmarks identification using an artificial neural network to enhance identification of cephalometric landmarks.

Materials and Methods: 110 lateral cephalograms were randomly selected from orthodontic private office and spited in two parts, First for training artificial neural network (ANN) and the remain cephalograms used for the evaluation of the software. In blind manner we asked three orthodontists to locate 5 landmarks on software and used these information for training. After that, our algorithm identified 5 landmarks on rest of cephalograms automatically. Eventually the result of both Algorithm evaluation and orthodontists landmarks for second part were compared with each other by "paired T test".

Results: Current Investigation showed, in four points out of five, the mean average distance between the point determined by ANN and the orthodontist’s points, was less than 1mm accuracy for all four landmarks.

Conclusion: With the limitation of this study, the results confirmed that that the landmark locating errors by ANN algorithms has near enough accuracy to realization, therefore it could be a proper substitute for manual method.