Convolutional neural networks for dental image diagnostics: A scoping review.

OBJECTIVES: Convolutional neural networks (CNNs) are increasingly applied for medical image diagnostics. We performed a scoping review, exploring (1) use cases, (2) methodologies and (3) findings of studies applying CNN on dental image material. SOURCES: Medline via PubMed, IEEE Xplore, arXiv were searched. STUDY SELECTION: Full-text articles and conference-proceedings reporting CNN application on dental imagery were included. DATA: Thirty-six studies, published 2015-2019, were included, mainly from four countries (South Korea, United States, Japan, China). Studies focussed on general dentistry (n = 15 studies), cariology (n = 5), endodontics (n = 2), periodontology (n = 3), orthodontics (n = 3), dental radiology (2), forensic dentistry (n = 2) and general medicine (n = 4). Most often, the detection, segmentation or classification of anatomical structures, including teeth (n = 9), jaw bone (n = 2) and skeletal landmarks (n = 4) was performed. Detection of pathologies focused on caries (n = 3). The most commonly used image type were panoramic radiographs (n = 11), followed by periapical radiographs (n = 8), Cone-Beam CT or conventional CT (n = 6). Dataset sizes varied between 10-5,166 images (mean 1,053). Most studies used medical professionals to label the images and constitute the reference test. A large range of outcome metrics was employed, hampering comparisons across studies. A comparison of the CNN performance against an independent test group of dentists was provided by seven studies; most studies found the CNN to perform similar to dentists. Applicability or impact on treatment decision was not assessed at all. CONCLUSIONS: CNNs are increasingly employed for dental image diagnostics in research settings. Their usefulness, safety and generalizability should be demonstrated using more rigorous, replicable and comparable methodology. CLINICAL SIGNIFICANCE: CNNs may be used in diagnostic-assistance systems, thereby assisting dentists in a more comprehensive, systematic and faster evaluation and documentation of dental images. CNNs may become applicable in routine care; however, prior to that, the dental community should appraise them against the rules of evidence-based practice.

Deep Learning for the Radiographic Detection of Periodontal Bone Loss.

We applied deep convolutional neural networks (CNNs) to detect periodontal bone loss (PBL) on panoramic dental radiographs. We synthesized a set of 2001 image segments from panoramic radiographs. Our reference test was the measured % of PBL. A deep feed-forward CNN was trained and validated via 10-times repeated group shuffling. Model architectures and hyperparameters were tuned using grid search. The final model was a seven-layer deep neural network, parameterized by a total number of 4,299,651 weights. For comparison, six dentists assessed the image segments for PBL. Averaged over 10 validation folds the mean (SD) classification accuracy of the CNN was 0.81 (0.02). Mean (SD) sensitivity and specificity were 0.81 (0.04), 0.81 (0.05), respectively. The mean (SD) accuracy of the dentists was 0.76 (0.06), but the CNN was not statistically significant superior compared to the examiners (p = 0.067/t-test). Mean sensitivity and specificity of the dentists was 0.92 (0.02) and 0.63 (0.14), respectively. A CNN trained on a limited amount of radiographic image segments showed at least similar discrimination ability as dentists for assessing PBL on panoramic radiographs. Dentists' diagnostic efforts when using radiographs may be reduced by applying machine-learning based technologies.

Deep Learning for the Radiographic Detection of Apical Lesions.

INTRODUCTION: We applied deep convolutional neural networks (CNNs) to detect apical lesions (ALs) on panoramic dental radiographs. METHODS: Based on a synthesized data set of 2001 tooth segments from panoramic radiographs, a custom-made 7-layer deep neural network, parameterized by a total number of 4,299,651 weights, was trained and validated via 10 times repeated group shuffling. Hyperparameters were tuned using a grid search. Our reference test was the majority vote of 6 independent examiners who detected ALs on an ordinal scale (0, no AL; 1, widened periodontal ligament, uncertain AL; 2, clearly detectable lesion, certain AL). Metrics were the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and positive/negative predictive values. Subgroup analysis for tooth types was performed, and different margins of agreement of the reference test were applied (base case: 2; sensitivity analysis: 6). RESULTS: The mean (standard deviation) tooth level prevalence of both uncertain and certain ALs was 0.16 (0.03) in the base case. The AUC of the CNN was 0.85 (0.04). Sensitivity and specificity were 0.65 (0.12) and 0.87 (0.04,) respectively. The resulting positive predictive value was 0.49 (0.10), and the negative predictive value was 0.93 (0.03). In molars, sensitivity was significantly higher than in other tooth types, whereas specificity was lower. When only certain ALs were assessed, the AUC was 0.89 (0.04). Increasing the margin of agreement to 6 significantly increased the AUC to 0.95 (0.02), mainly because the sensitivity increased to 0.74 (0.19). CONCLUSIONS: A moderately deep CNN trained on a limited amount of image data showed satisfying discriminatory ability to detect ALs on panoramic radiographs.