Akademik Veri Yönetim Sistemi
European annals of dental sciences (Online), cilt.49, sa.1, ss.1-4, 2022 (Hakemli Dergi)
Purpose: The aim of this study is to create a model that enables the detection of dentigerous cysts on panoramic radiographs in order to enable dentistry students to meet and apply artiőcial intelligence applications. Materials & Methods: E.O. and I.T. who are 5th-year students of the faculty of dentistry, detected 36 orthopantomographs whose histopathological examinations were determined as Dentigerous Cyst, and the affected teeth and cystic cavities were segmented using CranioCatch’s artiőcial intelligence supported clinical decision support system software. Since the sizes of the images in the data set are different from each other, all images were resized as 1024x514 and augmented as vertical ŕip, horizontal ŕip and both ŕips were applied on the train-validation. Within the obtained data set, 200 epochs were trained with PyTorch U-Net with a learning rate of 0.001, train: 112 images (112 labels), val: 16 images (16 labels). With the model created after the segmentations were completed, new dentigerous cyst orthopantomographs were tested and the success of the model was evaluated. Results: With the model created for the detection of dentigerous cysts, the F1 score (2TP / (2TP+FP+FN)) precision (TP/ (TP+N)) and sensitivity (TP/ (TP+FN)) were found to be 0.67, 0.5 and 1, respectively. Conclusion: With a CNN approach for the analysis of dentigerous cyst images, the precision has been found to be 0.5 even in a small database. These methods can be improved, and new graduate dentists can gain both experience and save time in the diagnosis of cystic lesions with radiographs.