Deep learning-based quantitative analysis of dental caries using optical coherence tomography: an ex vivo study

摘要

The conventional approach for diagnosing dental caries is clinical examination supplemented by radiographicevaluation. However, studies based on the clinical and radiographic examination methods often show low sensitivity andhigh specificity. Machine learning and deep learning techniques can be used to supplement optical coherencetomography (OCT) to more accurately identify diseased and damaged tissue. In this paper, we present a novel approachcombining OCT imaging modality and deep convolutional neural network (CNN) for the detection of occlusal cariouslesions. A total of 51 extracted human permanent teeth were collected and categorized into three groups: Non-cariousteeth, caries extending into enamel, and caries extending into dentin. In data acquisition and ex vivo OCT imaging, thesamples were imaged using spectral-domain OCT system operating at 1300nm center wavelength with a scan rate of 5.5-76kHz, and axial resolution of 5.5μm in air. To acquire images with minimum inhomogeneity, imaging was performedmultiple times at different points. For deep learning, OCT images of extracted human carious and non-carious teeth wereinput to a CNN classifier to determine variations in tissue densities reflecting the demineralization process. The CNNmodel employs two convolutional and pooling layers to extract features and then classify each patch based on theprobabilities from the SoftMax classification layer. The sensitivity and specificity of distinguishing between carious andnon-carious lesions were found to be 98% and 100%, respectively. This proposed deep learning-based OCT method canreliably detect demineralization of dental tissues, and could be extremely valuable in early dental caries detection.