Quantitative Analysis of Endodontic Treatment for Periapical Lesions in Intraoral Radiographs

Abstract

Intraoral radiographs have been taken to diagnose periapical lesions. Subsequent endodontic treatment needs to be evaluated quantitatively, that is often difficult due to various imaging factors as well as subjective visual interpretation. Therefore, we sought to establish an image analysis based quantitative model to evaluate endodontic treatments (40 effective and 43 noneffective cases). To normalize an image, the dentin area and the background were used as references. In each pair of images representing before and after treatment, the lesion area was manually selected by experts and segmented by tophat operation. Numerous features representing the effective bone healing were calculated. Using relative differences of selected features, an evaluation model was derived by logistic regression analysis. Gray level intensity and textural differences obtained from lesions significantly increased in the effectively treated cases. The model provided the accuracy of 80.7%. Our quantitative model may be helpful to evaluate endodontic treatment in clinical settings and in animal studies.

Detection of Dental Apical Lesions Using CNNs on Periapical Radiograph

“…Using Equation (4), as derived from the results, the accuracy of detection of apical lesions based on the technique presented on this paper has been success-fully increased to 92.75%. This is in contrast to the study in [23] which reported an accuracy of 80.7%. In [23], TOPHAT technology was used to crop the periapical, and the model was determined by logistic regression analysis.…”

“…This is in contrast to the study in [23] which reported an accuracy of 80.7%. In [23], TOPHAT technology was used to crop the periapical, and the model was determined by logistic regression analysis. Although this study showed a significant improvement.…”

“…The reported accuracy reached as high as 91.67%. Based on the above information, this article proposes apical lesion detection by using a periapical radiograph based on transfer learning with convolutional neural networks (CNN), using the literature in [23,27] as important references for comparison. This provides objective radiography interpretation and potentially saves more clinical time for dentists, allowing them to focus on treatment modality and clinical operation.…”

“…Semantic segmentation was first used to distinguish between targets of interest with an accuracy of up to 89%. Other technologies do not use artificial intelligence [23] in determining the feature of the tooth decay of the bitewing image, but use edge identification and Otsu thresholding for the image preprocessing. The affected areas were highlighted by the connected component and mask to find out where dental caries are [24] and analyze panoramic images on the mandibular bone for osteoporosis detection.…”

Noninvasive differential diagnosis of dental periapical lesions in cone‐beam CT scans

“…34,35 On periapical lesions, Mol and van der Stelt reported one of the early studies on periapical bone defects, detecting the lesions semiautomatically and characterizing their size via texture analysis. [36][37][38] More recently, Li et al 39 and Lee et al 40 introduced semiautomatic CAD methods for segmentation of periapical/bifurcation lesions by level-set method and for detection of bone remodeling at treated sites by logistic regression, respectively. However, these studies neither address the differential diagnosis between granulomas and cysts nor deal with 3D CT data.…”