Search

[eng] Objectives: The objective is to provide a methodology to obtain a categorized database without segmentation by sex or age that can be used in dental object detection applications and that may help in the diagnosis and usual clinical practice. Methods: A total of 10,677 panoramic images were analyzed by four examiners. In each tooth, the examiner indicated if the tooth exists or not and the position on FDI notation. After that, and for each tooth that exists, the examiner detailed whether or not there were the variables analyzed. Those variables were filled teeth, crown, implant, endodontic treatment, caries, and prosthetic. A descriptive study of inter-observer and intra-observer concordance-consistency was performed. Results: The results were statistically significant. Both teams obtained for all variables an almost perfect concordance k = 0.9 except in filled teeth where the kappa was k=0.8 and caries where a moderate agreement was obtained. The intra-examiner agreement was poor in caries variable and almost perfect in the rest of variables. Conclusions: A correctly categorized database is essential to obtain correct results in applications with artificial intelligence and neural networks. This study shows how to categorize a database of dental images for use in object detection applications in the field of dentistry., [spa] Objetivos: El objetivo de este estudio es proporcionar una metodología para categorizar imágenes dentales que serán usadas para detectar objetos dentales sin que exista segmentación por sexo o edad y la cual ayude en el diagnóstico y la práctica clínica. Métodos: 10.677 imágenes panorámicas fueron examinadas por 4 examinadores. En cada diente, los examinadores indicaron la existencia de dicho diente y la posición del mismo de acuerdo a la notación FDI. Posteriormente, para cada diente existente, se detalló la existencia o no de las variables analizadas. Dichas variables son: empastes, coronas, implantes, endodoncias, caries y prótesis. Se realizó un estudio de la concordancia inter e intra examinador. Resultados: Los resultados alcanzados son estadísticamente significativos. Ambas parejas obtuvieron una concordancia casi perfecta, k=0.9, en todas las variables excepto en los empastes, donde kappa toma un valor de k=0.8 y en las caries, donde la concordancia fue moderada. La concordancia intra-examinador fue pobre para las caries y casi perfecta para el resto de variables. Conclusiones: Una correcta categorización es esencial para obtener buenos resultados en las aplicaciones en las que se emplean las redes neuronales y la inteligencia artificial. Este estudio muestra cómo categorizar una base de datos de imágenes panorámicas que vayan a ser empleadas para detectar objetos en el campo de la odontología.

Dental radiography plays an important role in clinical diagnosis, treatment and making decisions. In recent years, efforts have been made on developing techniques to detect objects in images. The aim of this study was to detect the absence or presence of teeth using an effective convolutional neural network, which reduces calculation times and has success rates greater than 95%. A total of 8000 dental panoramic images were collected. Each image and each tooth was categorized, independently and manually, by two experts with more than three years of experience in general dentistry. The neural network used consists of two main layers: object detection and classification, which is the support of the previous one. A Matterport Mask RCNN was employed in the object detection. A ResNet (Atrous Convolution) was employed in the classification layer. The neural model achieved a total loss of 0.76% (accuracy of 99.24%). The architecture used in the present study returned an almost perfect accuracy in detecting teeth on images from different devices and different pathologies and ages., Asisa Dental S.A.U., Depto. de Medicina, Fac. de Medicina, TRUE, pub