A Mathematical Morphological Deep Neural Network for the Classification of Periapical Radiographs in the Diagnosis and Treatment of Dental Diseases
No Thumbnail Available
Date
2022-12
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Lead City University
Abstract
One of the best ways to diagnose a disease in medical practice objectively is through medical imaging. The importance of medical imaging cannot be overemphasized. In dentistry, dentists often use radiographs, especially in finding hidden dental structure, bone loss, malignant or benign masses, and cavities that cannot be examined during a visual examination. The use of dental radiographs also helps dentists to detect hidden dental diseases early. This study is a continuation of previous work which bothered on the development of an expert system for the diagnosis and prognosis of 20 Common dental diseases using Bayesian network. The work was developed using several symptoms associated with dental disease for diagnosing dental diseases (D1- D20). The study was limited to symptomatic diagnosis which however has some obvious gaps such as the uncertainty in the reasoning associated with Bayes rule, the use of Pain as a parameter among others that were filled through the use of deep learning tools on dental periapical radiographs through an improved model. The improved model was developed integrating mathematical morphology (MM) operations (dilation, erosion, opening and closing) in the convolution layer of CNN, for data preprocessing and quality feature extraction. With its high sense of intelligence (artificial) obtained during training, the system receives dental images and analyses them automatically for various clinical findings with which 6 dental disease problems were solved. With an achieved accuracy of 99.78%, it can be established that this system can be used in dental clinics with high confidence giving very little or no-error-diagnosis. To make this system more scalable and robust, more dental diseases should nbe added through other MM based theory like lattice, topology and random functions other than set theory-based MM used in this study.
Keywords: Mathematical Morphology (MM), Dilation, Erosion, Opening, Closing,
Convolutional Neural Network (CNN)
Word Count: 285
Description
Keywords
Citation
Kate Turabian