Logo Logo
Hilfe
Kontakt
Switch language to English
Automatisierte Erkennung und Kategorisierung von Seitenzahnrestaurationen mit Hilfe von künstlicher Intelligenz
Automatisierte Erkennung und Kategorisierung von Seitenzahnrestaurationen mit Hilfe von künstlicher Intelligenz
An automated detection and categorization of posterior restoration using CNN could increase the efficiency in dental practice and create a good foundation for further research. The main objective of the present dissertation project was to establish the prerequisites for the development of an AI-based algorithm for the automated detection of posterior restorations and, furthermore, to determine the diagnostic accuracy of the developed AI-algorithm in comparison to the created expert or reference standard. The image set of a total 1,761 clinical photo- graphs was divided into a training set (N=1,407) and an independent test set (N=354). The expert diagnoses served as a reference standard for cyclic train- ing and repeated evaluation of the CNN (ResNeXt-101-32x8d), which was trained by using image augmentation and transfer learning. Further statistical analysis included the calculation of contingency tables, areas under the Receiv- er Operating Characteristic Curve (AUCs) and saliency maps. After completion of the training, the CNN was able to correctly differentiate and recognize poste- rior restorations on clinical photographs with an average value of 97.1 percent. Therefore, it can be stated that the aim to develop an AI-algorithm for automatic categorization and recognition of posterior restorations was successfully com- pleted. Dental diagnostics might be supported by artificial intelligence-based algorithms in the future. To develop an applied AI-assisted evaluation system on clinical images, further research in practicability needs to be done.
Not available
Engels, Paula
2023
Deutsch
Universitätsbibliothek der Ludwig-Maximilians-Universität München
Engels, Paula (2023): Automatisierte Erkennung und Kategorisierung von Seitenzahnrestaurationen mit Hilfe von künstlicher Intelligenz. Dissertation, LMU München: Medizinische Fakultät
[thumbnail of Engels_Paula.pdf] PDF
Engels_Paula.pdf

422kB

Abstract

An automated detection and categorization of posterior restoration using CNN could increase the efficiency in dental practice and create a good foundation for further research. The main objective of the present dissertation project was to establish the prerequisites for the development of an AI-based algorithm for the automated detection of posterior restorations and, furthermore, to determine the diagnostic accuracy of the developed AI-algorithm in comparison to the created expert or reference standard. The image set of a total 1,761 clinical photo- graphs was divided into a training set (N=1,407) and an independent test set (N=354). The expert diagnoses served as a reference standard for cyclic train- ing and repeated evaluation of the CNN (ResNeXt-101-32x8d), which was trained by using image augmentation and transfer learning. Further statistical analysis included the calculation of contingency tables, areas under the Receiv- er Operating Characteristic Curve (AUCs) and saliency maps. After completion of the training, the CNN was able to correctly differentiate and recognize poste- rior restorations on clinical photographs with an average value of 97.1 percent. Therefore, it can be stated that the aim to develop an AI-algorithm for automatic categorization and recognition of posterior restorations was successfully com- pleted. Dental diagnostics might be supported by artificial intelligence-based algorithms in the future. To develop an applied AI-assisted evaluation system on clinical images, further research in practicability needs to be done.