Assessing the diagnostic performance of MediXpar texture analysis for C-shaped canal morphology on panoramic radiographs: a pilot study.

Authors' response.

PurposeThe aim of this study was to assess and compare the diagnostic performance of panoramic and occlusal radiographs in detecting submandibular sialoliths.Materials and MethodsA total of 40 patients (20 cases and 20 controls) were included in this retrospective study. Cases were defined as subjects with a submandibular sialolith confirmed by computed tomography (CT), whereas controls did not have any submandibular calcifications. Three observers with different expertise levels assessed panoramic and occlusal radiographs of all subjects for the presence of sialoliths. Intraobserver and interobserver agreement were assessed using the kappa test. Sensitivity, specificity, accuracy, positive and negative predictive values, and the diagnostic odds ratio of panoramic and occlusal radiographs in screening for submandibular sialoliths were calculated for each observer.ResultsThe sensitivity and specificity values for occlusal and panoramic radiographs all ranged from 80% to 100%. The lowest values of sensitivity and specificity observed among the observers were 82.6% and 80%, respectively (P=0.001). Intraobserver and interobserver agreement were higher for occlusal radiographs than for panoramic radiographs, although panoramic radiographs demonstrated a higher overall accuracy.ConclusionBoth panoramic and occlusal radiographic techniques displayed satisfactory diagnostic performance and should be considered before using a CT scan to detect submandibular sialoliths.

Predicting the configuration of a C-shaped canal system from panoramic radiographs

This study sought to analyse the importance of panoramic and carpal radiographs for the evaluation of edentulous individuals, demonstrating the contribution of different dental radiographic techniques in the detection of patients with signs of osteoporosis. 30 digital panoramic and carpal radiographs of women between 50 and 90 years old. In the panoramic radiographs, the thickness of the mandibular cortex was measured (Mentual Index - MI and Goniac Index - GA), and the morphology of the lower cortex was evaluated (Klemetti classification). In the carpal radiographs, the thickness of the cortices of the 2nd, 3rd and 4th metacarpals was measured (Metacarpal Index - IMC - or Nordin Index). The average age of the individuals was 68.43 years. In the analysis of the cortical thickness of the panoramic radiographs, the mean values of the mental index (MI) and the gonial index (GI) referring to the age group of 50 to 59 years had values considered within the normal range (? 3.1mm and ? 1.2mm , respectively). For the analysis of cortex morphology, group C3 corresponded to most cases (43.33%). For carpal radiographs, the highest metacarpal index (IMC) values were found in the younger age group (50-59 years). Finally, there was a positive correlation between age and quantitative (MI, GA, IMC) and qualitative (Klemetti analysis) assessment indices on panoramic and carpal radiographs.

Evaluation of the accuracy of detecting C-shaped canals in mandibular second molars identified by cone-beam computed tomography on panoramic radiographs using artificial intelligence algorithms developed with deep learning methods.

ABSTRACTBackground and AimsOdontogenic keratocyst (OKC) is a radiolucent jaw lesion often mistaken for similar conditions like ameloblastomas on panoramic radiographs. Accurate diagnosis is vital for effective management, but manual image interpretation can be inconsistent. While deep learning algorithms in AI have shown promise in improving diagnostic accuracy for OKCs, their performance across studies is still unclear. This systematic review and meta‐analysis aimed to evaluate the diagnostic accuracy of AI models in detecting OKC from panoramic radiographs.MethodsA systematic search was performed across 5 databases. Studies were included if they examined the PICO question of whether AI models (I) could improve the diagnostic accuracy (O) of OKC in panoramic radiographs (P) compared to reference standards (C). Key performance metrics including sensitivity, specificity, accuracy, and area under the curve (AUC) were extracted and pooled using random‐effects models. Meta‐regression and subgroup analyses were conducted to identify sources of heterogeneity. Publication bias was evaluated through funnel plots and Egger's test.ResultsEight studies were included in the meta‐analysis. The pooled sensitivity across all studies was 83.66% (95% CI:73.75%–93.57%) and specificity was 82.89% (95% CI:70.31%–95.47%). YOLO‐based models demonstrated superior diagnostic performance with a sensitivity of 96.4% and specificity of 96.0%, compared to other architectures. Meta‐regression analysis indicated that model architecture was a significant predictor of diagnostic performance, accounting for a significant portion of the observed heterogeneity. However, the analysis also revealed publication bias and high variability across studies (Egger's test, p = 0.042).ConclusionAI models, particularly YOLO‐based architectures, can improve the diagnostic accuracy of OKCs in panoramic radiographs. While AI shows strong capabilities in simple cases, it should complement, not replace, human expertise, especially in complex situations.

The aim of this study was to evaluate the use of a convolutional neural network (CNN) system for predicting C-shaped canals in mandibular second molars on panoramic radiographs. Panoramic and cone beam CT (CBCT) images obtained from June 2018 to May 2020 were screened and 1020 patients were selected. Our dataset of 2040 sound mandibular second molars comprised 887 C-shaped canals and 1153 non-C-shaped canals. To confirm the presence of a C-shaped canal, CBCT images were analyzed by a radiologist and set as the gold standard. A CNN-based deep-learning model for predicting C-shaped canals was built using Xception. The training and test sets were set to 80 to 20%, respectively. Diagnostic performance was evaluated using accuracy, sensitivity, specificity, and precision. Receiver-operating characteristics (ROC) curves were drawn, and the area under the curve (AUC) values were calculated. Further, gradient-weighted class activation maps (Grad-CAM) were generated to localize the anatomy that contributed to the predictions. The accuracy, sensitivity, specificity, and precision of the CNN model were 95.1, 92.7, 97.0, and 95.9%, respectively. Grad-CAM analysis showed that the CNN model mainly identified root canal shapes converging into the apex to predict the C-shaped canals, while the root furcation was predominantly used for predicting the non-C-shaped canals. The deep-learning system had significant accuracy in predicting C-shaped canals of mandibular second molars on panoramic radiographs.

Previous studies have implied that the panoramic radiograph was inferior to the bitewing radiograph for caries diagnosis. However, these clinical studies lacked a method of validation. The aim of this study was to use an electronic caries meter (ECM II, LODE, Groningen, The Netherlands) to validate occlusal caries diagnosis made from bitewing and panoramic radiographs. Forty-nine Army recruits were examined with the ECM, and had bitewing and panoramic radiographs taken. In total 299 molar occlusal surfaces were available for examination. Seven examiners viewed the bitewing and panoramic radiographs on two separate occasions and assessed each occlusal surface for dentine caries as 1: almost definitely no caries, 2: probably no caries, 3: unsure, 4: caries probably present, and 5: caries almost definitely present. This was repeated on 20% of the radiographs at two further separate sittings. ECM conductance readings greater than 9 were taken to indicate dentine caries. Examiner decisions that caries was probably and definitely considered to be present were taken as positive diagnoses. Bitewing and panoramic radiographs provided sensitivity values of 0.25 and 0.19 and specificity values of 0.93 and 0.97 respectively. ROC analysis indicated no statistically significant difference in diagnostic quality between the bitewing and panoramic radiographs. Intra-examiner reproducibility was found to be poor to moderate (Kappa values for bitewing = 0.31-0.44, panoramic = 0.07-0.54). No difference in overall diagnostic performance was found between bitewing and panoramic radiographs for the diagnosis of occlusal dentine caries.

This study aimed to develop a deep learning system for the detection of three-rooted mandibular first molars (MFMs) on panoramic radiographs and to assess its diagnostic performance. Panoramic radiographs, together with cone beam computed tomographic (CBCT) images of the same subjects, were retrospectively collected from 730 patients, encompassing a total of 1444 MFMs (367 teeth were three-rooted and the remaining 1077 teeth were two-rooted). Five convolutional neural network (CNN) models (ResNet-101 and − 50, DenseNet-201, MobileNet-v3 and Inception-v3) were employed to classify three- and two-rooted MFMs on the panoramic radiographs. The diagnostic performance of each model was evaluated using standard metrics, including accuracy, sensitivity, specificity, precision, negative predictive value, and F1 score. Receiver operating characteristic (ROC) curve analyses were performed, with the CBCT examination taken as the gold standard.Among the five CNN models evaluated, ResNet-101 demonstrated superior diagnostic performance, and the AUC value attained was 0.907, significantly higher than that of all other models (all P < 0.01). The accuracy, sensitivity, and specificity were 87.5%, 83.6%, and 88.9%, respectively. DenseNet-201, however, showed the lowest diagnostic performance among the five models (all P < 0.01), with an AUC value of 0.701 and an accuracy of 73.2%. Overall, the performance of the CNNs diminished when using image patches containing only the distal half of MFMs, with AUC values ranging between 0.680 and 0.800. In contrast, the diagnostic performance of the two clinicians was poorer, with AUC values of only 0.680 and 0.632, respectively. In conclusion, the CNN-based deep learning system exhibited a high level of accuracy in the detection of three-rooted MFMs on panoramic radiographs.

To describe the morphology and course of the inferior alveolar canal (IAC) as it appears in digital panoramic radiographs. Three hundred and eighty-six digital rotational panoramic radiographs (OPG) were studied using the Clinview Software (6.1.3.7 version, Instrumentarium). Among the 386 radiographs, 86 radiographs with 5-mm steel balls were used to calculate the magnification. The average magnification of radiographs in this study was 7.24+/-7.55%. The course of IAC as seen in the panoramic radiograph may be classified into four types: (1) linear curve, 12.75%, (2) spoon-shaped curve, 29.25%, (3) elliptic-arc curve, 48.5%, and (4) turning curve, 9.5%. On panoramic radiographs, the IAC appeared closest to the inferior border of the mandible in the region of the first molar. In relation to the teeth, on panoramic radiographs, the IAC appeared closest to the distal root tip of the third molar and furthest from the mesial root tip of the first molar. In the OPG, there are four types of IAC: linear, spoon shape, elliptic-arc, and turning curve. The data found in the study may be useful for dental implant, mandibule surgery, and dental anesthesia. The limitations of the panoramic radiograph in depicting the true three-dimensional (3D) morphology of the IAC are recognized, computed tomography (CT) and cone beam (CB)3D imaging being more precise.

PurposeAccurate detection of proximal dental caries on panoramic radiographs is essential for effective treatment planning and preventive care. While senior dental students gradually develop interpretative competence during their training, artificial intelligence (AI) systems have emerged as promising adjuncts to enhance diagnostic performance. This study aimed to compare the diagnostic competence of fourth- and fifth-year dental students in detecting proximal carious lesions on panoramic radiographs, with and without AI assistance. It was hypothesized that AI-assisted evaluation would significantly improve diagnostic accuracy compared to unaided interpretation.MethodsA total of 132 dental students (66 fourth-year and 66 fifth-year) from Biruni University participated in this cross-sectional study. Sixty anonymized panoramic radiographs, each representing a single diagnostic question, and depicting various depths of carious lesions (enamel, dentin, or pulpal) were evaluated by a specialist in oral and maxillofacial radiology and a restorative dentistry expert. Disagreements were resolved by consensus, and their evaluation served as the reference standard. Each student assessed the presence and depth of caries in two separate sessions: first unaided and then assisted by AI-based diagnostic software (DeepDent AI, DeepInsight Technologies), trained on 10,000 annotated panoramic images and designed to highlight carious regions using bounding boxes and probability scores. Statistical analyses were conducted using IBM SPSS Statistics Version 27, employing the Chi-square test, Mann–Whitney U test, and Cohen’s Kappa coefficient to compare diagnostic performance and inter-rater agreement. Intra- and inter-observer reliability were high (κ = 0.82 and κ = 0.79, respectively). A p-value < 0.05 was considered statistically significant.ResultsFifth-year students demonstrated higher overall diagnostic performance than fourth-year students (p < 0.05). The highest rate of accurate diagnoses was observed for pulpal caries (39.4%), followed by dentin (6.1%) and enamel (5.3%) lesions. No statistically significant difference was found between the two student groups across caries depths (p > 0.05). The AI system achieved a sensitivity of 88.3%, specificity of 91.7%, and an overall accuracy of 90.1%, significantly outperforming both student groups (p < 0.001). Cohen’s Kappa coefficients (κ = 0.41–0.60) indicated moderate inter-rater agreement between students and the reference standard.ConclusionPanoramic radiographs alone provide limited accuracy for detecting early or shallow proximal caries, but diagnostic performance improves with educational level. The incorporation of AI-assisted diagnostic tools into undergraduate dental curricula may enhance students’ interpretative accuracy, strengthen clinical decision-making, and support early caries detection. Future studies should develop and evaluate standardized AI-integrated educational modules to optimize diagnostic proficiency in dental training.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12909-025-08269-2.

ObjectivesThe observer score of the trabecular pattern on panoramic radiographs is known to be a strong predictor of bone fractures. The aim of this study was to enhance the predictive power of panoramic radiographs by means of texture analysis methods.Material and methodsThe study followed 304 postmenopausal women during 26 years. At the beginning of the study, panoramic radiographs were obtained. One observer assessed the trabecular pattern in the premolar region as dense, sparse, or alternating dense and sparse. In addition, on each radiograph, a region of interest was selected in the molar/premolar region and analyzed with texture analysis procedures. During 26 years of follow-up, 115 women suffered a fracture of the hip, spine, leg, or arm. Logistic regression was applied to test the predictive power of various variables with respect to fractures.ResultsOf all variables, the observer score of the trabecular pattern correlated strongest with the occurrence of fractures. By itself, the score yielded an ROC curve with an area of 0.80 under the curve. Combining the observer score with the texture analysis features increased the area under the ROC curve to 0.85.ConclusionsThe trabecular pattern on panoramic radiographs provides a strong predictor of fractures, at least for postmenopausal women. The assessment by an observer combined with texture analysis procedures yields a predictive power that parallels best known predictions in literature.Clinical relevanceThis study illustrates that panoramic radiographs are state of the art predictors of postcranial fractures.

Development and Validation of a Visually Explainable Deep Learning Model for Classification of C-shaped Canals of the Mandibular Second Molars in Periapical and Panoramic Dental Radiographs

Mandibular cortical erosion detected on panoramic radiographs may be useful for identifying post-menopausal women with low skeletal bone mineral density (BMD). The purposes of this study were to calculate the diagnostic performance of general dental practitioners (GDPs) who attended a lecture on identifying post-menopausal women with low BMD from findings on panoramic radiographs and to evaluate the influence of GDPs' age on diagnostic performance. After a 1 h lecture, 111 GDPs were asked to classify the mandibular cortex (normal or eroded) on panoramic radiographs obtained from 100 post-menopausal women who have had skeletal BMD assessment. Low BMD was defined as a BMD T score of -1.0 or less. Diagnostic performance was analysed by comparing two groups classified by mandibular cortex (women with normal cortex and women with any eroded cortex) with those classified by BMD (women with normal BMD and women with low BMD). The mean sensitivity, specificity, positive predictive value, negative predictive value, accuracy and likelihood ratio for a positive risk result were 73.0% (95% confidence interval [CI]; 71.3 to 74.7%), 49.0% (95% CI; 46.4 to 51.5%), 66.9% (95% CI; 66.0 to 67.8%), 57.0% (95% CI; 55.8 to 58.2%), 62.9% (95% CI; 62.1 to 63.7%) and 1.51 (95% CI; 1.44 to 1.58), respectively. GDPs' age did not influence diagnostic performance. Our results suggest that 73.0% of women who had low skeletal BMD can be identified by GDPs after a lecture on the use of panoramic radiographs as an aid in diagnosing low BMD; however, the diagnostic performance may not be influenced by GDPs' age.

The evaluation of the maxillary sinus is very important in dental practice such as tooth extraction and implantation because of its proximity to the teeth, but it is not easy to evaluate because of the overlapping structures such as the maxilla and the zygoma on panoramic radiographs. When doom-shaped retention pseudocysts are observed in sinus on panoramic radiographs, they are often misdiagnosed as cysts or tumors, and additional computed tomography is performed, resulting in unnecessary radiation exposure and cost. The purpose of this study was to develop a deep learning model that automatically classifies retention pseudocysts in the maxillary sinuses on panoramic radiographs. A total of 426 maxillary sinuses from panoramic radiographs of 213 patients were included in this study. These maxillary sinuses included 86 sinuses with retention pseudocysts, 261 healthy sinuses, and 79 sinuses with cysts or tumors. An EfficientDet model first introduced by Tan for detecting and classifying the maxillary sinuses was developed. The developed model was trained for 200 times on the training and validation datasets (342 sinuses), and the model performance was evaluated in terms of accuracy, sensitivity, and specificity on the test dataset (21 retention pseudocysts, 43 healthy sinuses, and 20 cysts or tumors). The accuracy of the model for classifying retention pseudocysts was 81%, and the model also showed higher accuracy for classifying healthy sinuses and cysts or tumors (98% and 90%, respectively). One of the 21 retention pseudocysts in the test dataset was misdiagnosed as a cyst or tumor. The proposed model for automatically classifying retention pseudocysts in the maxillary sinuses on panoramic radiographs showed excellent diagnostic performance. This model could help clinicians automatically diagnose the maxillary sinuses on panoramic radiographs.