Digital image analysis of dental radiography for cavity detection

Abstract

Healthcare research is becoming an increasingly important newlinefield for medical imaging. Technological advances in physical imaging newlineconstitute a step in the direction of effective healthcare and patient newlinefriendly devices. With such advances, the possibilities of early disease newlinedetection and their accuracy are on the rise. Initial detection of oral newlinecavity reduces the risk of losing the teeth. The main objective of this newlineresearch is to design a deep learning technique to detect the dental newlinecavity which helps the dentist in treatment plans. Research work is newlinedivided into four modules. newlineThe first module provides different filtering techniques used newlineunder different noise conditions to produce better output noise-free newlineimages in the medical sector. The input images are seen with different newlinenoise such as salt and pepper noise, speckle noise and Gaussian noises. newlineThe different filters namely, Average media filter, linear filter, and newlinewiener filter are applied to the image and performance is analysed by newlinemeasuring MSE values. The method of designing a selective median newlinefilter is obtained by the above analysis. Selective median filter acts as a newlinemedian filter for salt and pepper noise and acts as a wiener filter for other newlineix newlinenoises. MSE parametric values for the selective median filter are low in newlinecomparison with conventional filters. newlineIn the second module, a canny edge detection model is used newlinewhich extracts the edge features and then it is given to Otsu s newlinethresholding method. In segmentation, teeth images are segmented in newlinelayer wise such as enamel, dentin, root, pulp and bone. Layer wise newlinesegmented output can be used for finding cavity presence in the newlineparticular region efficiently. This helps the physician in easy newlineidentification of the cavity region in the teeth. newlineThird module deals with the process of dental caries newlineclassification system using the convolutional neural network. There are newlinetotally 418 images taken for the training of the learning algorithm. Each newlineimage in the dataset is pre-processed using a selective m