Investigation of the best effective fold of data augmentation for training deep learning models for recognition of contiguity between mandibular third molar and inferior alveolar canal on panoramic radiographs.

OBJECTIVES: This study aimed to train deep learning models for recognition of contiguity between the mandibular third molar (M3M) and inferior alveolar canal using panoramic radiographs and to investigate the best effective fold of data augmentation. MATERIALS AND METHODS: The total of 1800 M3M cropped images were classified evenly into contact and no-contact. The contact group was confirmed with CBCT images. The models were trained from three pretrained models: AlexNet, VGG-16, and GoogLeNet. Each pretrained model was trained with the original cropped panoramic radiographs. Then the training images were increased fivefold, tenfold, 15-fold, and 20-fold using data augmentation to train additional models. The area under the receiver operating characteristic curve (AUC) of the 15 models were evaluated. RESULTS: All models recognized contiguity with AUC from 0.951 to 0.996. Ten-fold augmentation showed the highest AUC in all pretrained models; however, no significant difference with other folds were found. VGG-16 showed the best performance among pretrained models trained at the same fold of augmentation. Data augmentation provided statistically significant improvement in performance of AlexNet and GoogLeNet models, while VGG-16 remained unchanged. CONCLUSIONS: Based on our images, all models performed efficiently with high AUC, particularly VGG-16. Ten-fold augmentation showed the highest AUC by all pretrained models. VGG-16 showed promising potential when training with only original images. CLINICAL RELEVANCE: Ten-fold augmentation may help improve deep learning models' performances. The variety of original data and the accuracy of labels are essential to train a high-performance model.

Evaluation of cone-beam computed tomography diagnostic image quality using cluster signal-to-noise analysis.

OBJECTIVES: (1) We sought to assess correlation among four representative parameters from a cluster signal-to-noise curve (true-positive rate [TPR] corresponding to background noise, accuracy corresponding to background noise, maximum TPR, and maximum accuracy) and the diagnostic accuracy of the identification of the mandibular canal using data from observers in a previous study, under the same exposure conditions. (2) We sought to clarify the relationship between the hole depths of a phantom and diagnostic accuracy. METHODS: CBCT images of a Teflon plate phantom with holes of decreasing depths from 0.7 to 0.1 mm were analyzed using the FindFoci plugin of ImageJ. Subsequently, we constructed cluster signal-to-noise curves by plotting TPRs against false-positive rates. The four parameters were assessed by comparing with the diagnostic accuracy calculated from the observers. To analyze image contrast ranges related to detection of mandibular canals, we determined five ranges of hole depths, to represent different contrast ranges-0.1-0.7, 0.1-0.5, 0.2-0.6, 0.2-0.7 and 0.3-0.7 mm-and compared them with observers' diagnostic accuracy. RESULTS: Among the four representative parameters, accuracy corresponding to background noise had the highest correlation with the observers' diagnostic accuracy. Hole depths of 0.3-0.7 and 0.1-0.7 mm had the highest correlation with observers' diagnostic accuracy in mandibles with distinct and indistinct mandibular canals, respectively. CONCLUSIONS: The accuracy corresponding to background noise obtained from the cluster signal-to-noise curve can be used to evaluate the effects of exposure conditions on diagnostic accuracy.

Determination of optimum exposure parameters for dentoalveolar structures of the jaws using the CB MercuRay system with cluster signal-to-noise analysis.

OBJECTIVE: To determine the optimum cone beam computed tomography exposure parameters for specific diagnostic tasks. METHODS: A Teflon phantom attached to a half-mandible in a large container was scanned in dental (D), implant (I), and panoramic (P) modes. An identical phantom in a small container was scanned in D mode. Both were scanned at 60, 80, 100, and 120 kV. We evaluated the image quality of five anatomical structures [dentinoenamel junction (1), lamina dura and periodontal ligament space (2), trabecular pattern (3), cortex-spongy bone junction (4), and pulp chamber and root canal (5)] and analyzed the diagnostic image quality with cluster signal-to-noise analysis. We then evaluated correlations between the two image qualities and calculated the threshold of acceptable diagnostic image quality. Optimum exposure parameters were determined from images with acceptable diagnostic image quality. RESULTS: For the small container, the optimum exposure parameters were D mode, 80 kV for (1), (3), and (4) and D mode, 100 kV for (5). For the large container, they were D mode, 120 kV for (1), (3), and (5) and D mode, 100 kV for (4). I mode, 120 kV reached the acceptable level for (4). No images reached the acceptable level for (2). CONCLUSIONS: No optimum exposure parameters were identified for the evaluation of the lamina dura and periodontal ligament space. D mode was sufficient for the other structures; however, the tube voltage required for each structure differed. Smaller patients required lower tube voltage. I mode, 120 kV may be used for larger lesions.

Sonographic diagnosis in the head and neck region: from an educational lecture presented at the 56th General Assembly and Annual Scientific Congress of the Japanese Society for Oral and Maxillofacial Radiology.

Sonography is a simple, inexpensive, and non-invasive diagnostic modality. Although tissues behind bony structures and deep tissues are not delineated, sonography can depict superficial soft tissues very clearly. In the head and neck region, however, it has not yet been used widely, as the anatomical structures are complicated, and considerable experience is needed both to perform an examination and to make a diagnosis. To perform examinations efficiently, operators must be familiar with the sonographic system in use, and take images at standard planes. To make a correct diagnosis, operators require knowledge of the sonographic anatomy on standard planes, representative sonographic signs and artifacts, and common diseases and their typical sonographic findings. In this paper, we have explained the sonographic anatomy on standard planes, and the sonographic findings of common diseases in the oral and maxillofacial region.

Prediction of detectability of the mandibular canal by quantitative image quality evaluation using cone beam CT.

OBJECTIVES: To compare the results of a new quantitative image quality evaluation method that requires no observers with the results of receiver operating characteristic (ROC) analysis in detecting the mandibular canal (MC) in cone beam CT (CBCT) images. METHODS: A Teflon (polytetrafluoroethylene) plate phantom with holes of different depths was scanned with two CBCT systems. One CBCT system was equipped with an image intensifier (Experiment 1), and the other was equipped with a flat panel detector (Experiment 2). Holes that were above the threshold gray value (ΔG), calculated using just-noticeable difference (JND), were extracted. The number of extracted holes was used as the index of the image quality, and was compared with the Az values calculated by ROC analysis to detect the MC. RESULTS: The number of extracted holes reflected the influence of different scanning conditions, and showed a strong correlation with the Az values calculated by ROC analysis. Indices of the number of extracted holes corresponding to high Az values for detecting the MC were obtained in both experiments. CONCLUSIONS: Our image quality evaluation method applying JND to images of a standardized phantom is a quantitative method that could be useful for evaluating the detectability of the MC in CBCT images.

Cluster signal-to-noise analysis for evaluation of the information content in an image.

OBJECTIVES: (1) To develop an observer-free method of analysing image quality related to the observer performance in the detection task and (2) to analyse observer behaviour patterns in the detection of small mass changes in cone-beam CT images. METHODS: 13 observers detected holes in a Teflon phantom in cone-beam CT images. Using the same images, we developed a new method, cluster signal-to-noise analysis, to detect the holes by applying various cut-off values using ImageJ and reconstructing cluster signal-to-noise curves. We then evaluated the correlation between cluster signal-to-noise analysis and the observer performance test. We measured the background noise in each image to evaluate the relationship with false positive rates (FPRs) of the observers. Correlations between mean FPRs and intra- and interobserver variations were also evaluated. Moreover, we calculated true positive rates (TPRs) and accuracies from background noise and evaluated their correlations with TPRs from observers. RESULTS: Cluster signal-to-noise curves were derived in cluster signal-to-noise analysis. They yield the detection of signals (true holes) related to noise (false holes). This method correlated highly with the observer performance test (R2 = 0.9296). In noisy images, increasing background noise resulted in higher FPRs and larger intra- and interobserver variations. TPRs and accuracies calculated from background noise had high correlation with actual TPRs from observers; R2 was 0.9244 and 0.9338, respectively. CONCLUSIONS: Cluster signal-to-noise analysis can simulate the detection performance of observers and thus replace the observer performance test in the evaluation of image quality. Erroneous decision-making increased with increasing background noise.

Effects of exposure parameters and slice thickness on detecting clear and unclear mandibular canals using cone beam CT.

OBJECTIVES: The purpose of this study was to clarify the effects of exposure parameters and image-processing methods when using CBCT to detect clear and unclear mandibular canals (MCs). METHODS: 24 dry half mandibles were divided into 2 groups with clear and unclear MCs based on a previous CBCT study. Mandibles were scanned using a CBCT system with varying exposure parameters (tube voltages 60 kV, 70 kV and 90 kV; and tube currents 2 mA, 5 mA, 10 mA and 15 mA) to obtain a total of 144 scans. The images were processed with different slice thicknesses using ImageJ software (National Institutes of Health, Bethesda, MD). Five radiologists evaluated the cross-sectional images of the first molar region to detect the MCs. The diagnostic accuracy of varying exposure parameters and image-processing conditions was compared with the area under the curve (Az) in receiver-operating characteristic analysis. RESULTS: The Az values for clear MCs were higher than those for unclear MCs (p < 0.0001). With increasing exposure voltages and currents, Az values increased, but no significant differences were found with high voltages and currents in clear MCs (p = 1.0000 and p = 0.9340). The Az values of serial images were higher than those of overlaid images (p < 0.0001), and those for thicker slices were higher than those for thinner slices (p < 0.0001). CONCLUSIONS: Our findings indicate that detection of unclear MCs requires either higher exposure parameters or processing of the images with thicker slices. To detect clear MCs, lower exposure parameters can be used.

A new method to evaluate image quality of CBCT images quantitatively without observers.

OBJECTIVES: To develop an observer-free method for quantitatively evaluating the image quality of CBCT images by applying just-noticeable difference (JND). METHODS: We used two test objects: (1) a Teflon (polytetrafluoroethylene) plate phantom attached to a dry human mandible; and (2) a block phantom consisting of a Teflon step phantom and an aluminium step phantom. These phantoms had holes with different depths. They were immersed in water and scanned with a CB MercuRay (Hitachi Medical Corporation, Tokyo, Japan) at tube voltages of 120 kV, 100 kV, 80 kV and 60 kV. Superimposed images of the phantoms with holes were used for evaluation. The number of detectable holes was used as an index of image quality. In detecting holes quantitatively, the threshold grey value (ΔG), which differentiated holes from the background, was calculated using a specific threshold (the JND), and we extracted the holes with grey values above ΔG. The indices obtained by this quantitative method (the extracted hole values) were compared with the observer evaluations (the observed hole values). In addition, the contrast-to-noise ratio (CNR) of the shallowest detectable holes and the deepest undetectable holes were measured to evaluate the contribution of CNR to detectability. RESULTS: The results of this evaluation method corresponded almost exactly with the evaluations made by observers. The extracted hole values reflected the influence of different tube voltages. All extracted holes had an area with a CNR of ≥1.5. CONCLUSIONS: This quantitative method of evaluating CBCT image quality may be more useful and less time-consuming than evaluation by observation.

Effectiveness of imaging modalities for screening IgG4-related dacryoadenitis and sialadenitis (Mikulicz's disease) and for differentiating it from Sjögren's syndrome (SS), with an emphasis on sonography.

INTRODUCTION: The aim of this study was to clarify the effectiveness of various imaging modalities and characteristic imaging features in the screening of IgG4-related dacryoadenitis and sialadenitis (IgG4-DS), and to show the differences in the imaging features between IgG4-DS and Sjögren's syndrome (SS). METHODS: Thirty-nine patients with IgG4-DS, 51 with SS and 36 with normal salivary glands were enrolled. Images of the parotid and submandibular glands obtained using sonography, 2-[(18)F]-fluoro-2-deoxy-D-glucose positron emission tomography/computed tomography (FDG-PET/CT), computed tomography (CT) and magnetic resonance imaging (MRI) were retrospectively analyzed. Six oral and maxillofacial radiologists randomly reviewed the arranged image sets under blinded conditions. Each observer scored the confidence rating regarding the presence of the characteristic imaging findings using a 5-grade rating system. After scoring various findings, diagnosis was made as normal, IgG4-DS or SS, considering all findings for each case. RESULTS: On sonography, multiple hypoechoic areas and hyperechoic lines and/or spots in the parotid glands and obscuration of submandibular gland configuration were detected mainly in patients with SS (median scores 4, 4 and 3, respectively). Reticular and nodal patterns were observed primarily in patients with IgG4-DS (median score 5). FDG-PET/CT revealed a tendency for abnormal (18)F-FDG accumulation and swelling of both the parotid and submandibular glands in patients with IgG4-DS, particularly in the submandibular glands. On MRI, SS had a high score regarding the findings of a salt-and-pepper appearance and/or multiple cystic areas in the parotid glands (median score 4.5). Sonography showed the highest values among the four imaging modalities for sensitivity, specificity and accuracy. There were significant differences between sonography and CT (p = 0.0001) and between sonography and FDG-PET/CT (p = 0.0058) concerning accuracy. CONCLUSIONS: Changes in the submandibular glands affected by IgG4-DS could be easily detected using sonography (characteristic bilateral nodal/reticular change) and FDG-PET/CT (abnormal (18)F-FDG accumulation). Even inexperienced observers could detect these findings. In addition, sonography could also differentiate SS. Consequently, we recommend sonography as a modality for the screening of IgG4-DS, because it is easy to use, involves no radiation exposure and is an effective imaging modality.