An overview of image processing for dental diagnosis

Abstract

Dental disease evaluation and clinical assessment are frequently accomplished through radiographic penetration. The difficulty of obtaining an accurate clinical diagnosis from radiographs rises due to the minimal mineral density change in demineralized tissue of tooth and gum disorders. Dental abnormalities may not be visible on radiographs until the demineralization is higher than 40%, according to the literature. As a result, a dental practitioner’s judgment can have a big impact on how accurately the radiography penetration depth is determined through visual inspection. To counteract this effect, image processing-based clinical diagnosis methods have become widely adopted, transforming dentistry from traditional to advance in recent years. The efforts made in the area of image processing-based digital dental diagnosis of the most challenging dental issues are outlined in the presented comprehensive literature evaluation, which also identifies any research gaps in the scope of work already done. The included studies’ quality was evaluated using Quality Assessment and Diagnostic Accuracy Tool-2 (QUADAS-2). A total of 52 out of 178 articles, published from 2012 to February 2023, were reviewed and data like image-processing approach, the size of datasets, approach results, advantages and disadvantages, name(s) of diagnosed diseases, imaging type, author, and publication year were extracted. Results show that, in 52 studies, more than 14 image-processing approaches were used on different types of radiographs for the diagnosis of a single or more than one disease by a single approach with an accuracy range from 64% to 93%. Most studies have used artificial intelligence (AI) for computer-aided diagnosis and used dental experts to label their dataset and validate the outcome of proposed methods. Efforts done by different research groups for image processing-based digital diagnosis are appreciable but still, they are lagging to meet clinically expected accuracy. There looks to be a great requirement for the development or standardization of existing methodology and it is also needed to construct standard public dental datasets to attract a greater number of research groups in the dental field.