Toward development of automated grading system for carious lesions classification using deep learning and OCT imaging

Dental caries remains the most prevalent chronic disease in both children and adults. Optical coherence tomography (OCT) is a noninvasive optical imaging modality extensively utilized to image oral samples to diagnose carious lesions, but detecting early stage dental caries with high-level accuracy remains challenging. Deep learning models have been employed to classify OCT images for various healthcare applications. In this paper, human tooth specimens were imaged ex vivo using OCT imaging systems, and a three-class grading system based on deep learning model for detection and classification of carious lesions was developed. Human extracted premolar and molar teeth were collected and categorized into three classes, Grade 0: healthy (non-carious teeth), Grade 1: early-stage caries (caries extending into enamel), and Grade 2: late-stage caries (caries extending into dentin). For OCT imaging, a spectral-domain OCT system and a swept-source OCT system were utilized. Advanced image processing and augmentation techniques were performed to prepare the image data and generate additional examples of each class prior to the deep learning process. For deep learning, ten deep convolutional neural networks (CNN) architectures were investigated to determine the optimal numbers of convolutional and fully connected layers for the classification tasks. The diagnostic accuracy, sensitivity, specificity, positive predictive value, and negative predictive value were calculated for detection and diagnostic performances of the CNN models. This study is a step forward in the development of automated deep learning/OCT imaging system for early dental caries diagnosis.