Modelo de Detección de la Pérdida Ósea Radiográfica Basada en Deep Learning

Abstract

among diseases classified as most common in diagnosis we find dental cavities, toothache, sore gums, dental abscess. Which in some cases are underlining a more complex treatment for a periodontal disease for which 2 out of 5 adults in the USA are affected by any classification of a periodontal disease (P.I, B.A, L, G.O, & R.J, 2012). We present a labeling process for a database composed of periapical x ray using Periodontal Bone Loss attributes as information for training, validation, and test of a convolutional deep learning model. The x rays were extracted previous permission of the DAGSS of Universidad Santo Tomas, such x rays present variations accordingly to x ray protocol, radiolucency, radiopacity, contrast, specialist indications, mouth quadrants, among other things. Taking into consideration the periodontal workshop we identified the points of interest used for the identification of the Periodontal Bone Loss under supervision of periodontal experts using the bounding box technique we labeled 4175 tooth and 5229 points of interest out of 2379 x rays.