Effectiveness of using artificial intelligence-based software for automated analysis of orthopantomograms to identify signs of periapical lesions

Abstract

Introduction. The development of machine learning and artificial intelligence technologies can potentially facilitate the process of identifying signs of periapical lesions on digital orthopantomograms, speeding up the stages of primary analysis of radiographic images and demarcation among such obvious and questionable problematic areas. Objective. To assess the diagnostic effectiveness of artificial intelligence-based software for identifying signs of periapical lesions during automated analysis of orthopantomograms. Methodology/Methods. The study design involved automatic analysis of orthopantomograms by software operating on the basis of artificial intelligence technology, in order to register signs of periapical lesions on digital images. For the pilot study, it was planned to use 60 orthopantomograms, randomly selected from the digital image database. To assess the comparative diagnostic effectiveness of the software used, the analysis of the selected 60 orthopantomograms with a targeted search among such areas with signs of periapical lesions was carried out independently by two dentists with experience in clinical practice for over 10 years and specialization in performing therapeutic dental manipulations. Results and Discussion. During the automated processing of 60 orthopantomograms by software based on artificial intelligence, 187 areas with signs of potential periapical changes were verified. Two dentists verified 158 and 132 areas with signs of potential periapical changes, respectively. Repeated analysis of the same sample of orthopantomograms by software based on artificial intelligence demonstrated 100% reproducibility of the obtained results in terms of the number of initially identified areas, as well as in terms of their distribution. The data obtained as a result of the analysis of orthopantomograms by dentists has shown show that agreement between them in terms of the number of identified areas with signs of periapical changes was 83,54%, while in terms of their distribution was 87,12%. Conclusions. Software based on artificial intelligence during the automated analysis of orthopantomograms provides 15,50–29,41% higher prevalence for identification of signs related with potential periapical lesions compared to the analysis of orthopantomograms performed by dentists. A significant advantage of the software for automatic analysis of orthopantomograms is the absolute reproducibility of the obtained results when repeating the diagnostic processing of digital images.