Project Home
PDF access
Authors
Abduallah ELmaraghy, Ganna Ayman, Mohamed Khaled, Sara Tarek, Maha Sayed, Mennat Allah Hassan, Yomna M.I. Hassan
Publishing Date
December 15, 2021
Abstract
Dental pathology is a wide field of study as it passes by several stages of diagnoses and treatments on the patients. As a part of the diagnosis process for some dental diseases that cause facial asymmetry, dentists need a lot of tests and diagnoses to detect the dental problem that causes facial asymmetry. The classification is either by imaging or scanning. 3D models are more accurate than 2D images due to less error rate in information loss. Less information loss helps to provide a successful classification process. This document proposes a system that takes one 2D facial image as an input and constructs it into a 3D facial model, then classifies facial asymmetry using photogrammetry and deep learning techniques. The reconstruction of facial models uses several input images with different poses and angles. In this system, the challenges are to reach higher accuracy in the reconstruction process, to reduce the loss rate of any facial features during that process, and finally. To be able to classify any facial asymmetry that could indicate dental problems. The research was carried out on the proposed method to validate it and compare its performance. Research’s results prove that the proposed technique is time-efficient and accurate.
1.1 Background
Orthodontic pathology is a wide field of study. Orthodontics pass by several stages for diagnosing to pick the right plan for treatment. The diagnoses are so important and sensitive. Facial asymmetry in some cases could indicate some dental diseases. Dental diagnoses must be very accurate to follow the correct treatment. 3D Facial models are more informative and easier for classification and analysis than 2D ones. 3D models will be more helpful for dentists than 2D images as 2D images don’t reach the accuracy that 3D models reach with facial asymmetry classification. A proposed solution will be an assistant system, which takes the 2D facial images of patients and converts them into a 3D facial model. In addition, the system will classify the existence of any facial abnormality or asymmetry. Dentists could then diagnose and identify the patient’s disease within a brief time, which finally would lead to correct and accurate treatment.
1.2 Motivation
Dentistry is a very wide field of study, especially dental pathology. Diagnosing and detecting dental diseases is not an easy process. The process of diagnosing the defect or the disease is very important and sensitive in dentistry to find the appropriate treatment. With the spread of viruses nowadays especially COVID-19, most of the clinics are trying to decrease the gathering in the clinics. They do so to decrease the infection rate. Some clinics already offer the diagnosis process online, so they may need an automated system for facial asymmetry classification. Providing a system for facial asymmetry classification will also be helpful as an assistant for dentists. Dentists need to do different types of scanning and imaging of the patients’ faces and teeth for accurate diagnoses and treatment. Providing dentists with accurate and high-quality data is highly important for appropriate diagnosis. Studies have proven that 3D models of faces and teeth are more accurate providing more information and correct diagnoses than the 2D images or scans . Dentists use Cone-beam computed tomography (CBCT) which is an imaging method in dentistry as it shows information that can’t be seen in the normal 2D imaging. They tried this way as a solution for the problem faced when working with 2D images. But unfortunately, CBCT image is a time-consuming process that requires a physician to work with complicated software. A solution for this time-consuming process could be an application that assists dentists in diagnosing and detecting diseases in teeth by calculating and classifying the facial asymmetry. The system will take 2D facial images as input, then convert them to a 3D model (for easy and accurate diagnoses). Finally, after processing these models classifies the facial asymmetry for further diagnoses and treatment.
1.3 Problem Statement
Facial asymmetry is a common issue that causes both functional and aesthetic problems. Orthodontists use a multi-stage diagnostic process to identify the dental disease-causing facial asymmetry. Classification of facial asymmetry from 2D images is currently the most commonly used method of classifying facial asymmetry to assist dentists. Since 2D images do not provide enough information about the face, it is more efficient to use 3D models to address this issue. Reconstructing a 3D facial model requires many input images. It is not an efficient method as time and effort are the main factors for dentists in the classification process. It’s important to reconstruct the 3D facial model from the minimum number of input images without losing much information about the face.