Dento maxillo facial radiology最新文献

Objectives: To evaluate artefact-reducing filters as a means to optimize and ensure the accuracy of carious lesion diagnosis testing in 2 experimental models (presence and absence of adjacent metallic objects).

Methods: Fifty molar teeth were used, randomly divided into 5 groups (n = 10): G1-Sound teeth; G2-Carious teeth; G3-teeth with Class I cavity preparation restored with resin (Cl I + R); G4-Cl I + R with the use of a hyperdense lining material; and G5-Cl I + R with the use of a hypodense lining material. The Carestream CS 9600 tomograph was used, testing 2 experimental models (presence and absence of adjacent metallic objects), with tube voltages of 100 kV and 120 kV, voxel sizes of 75 and 150 µm, and applying the metal artefact reduction (MAR) filter. Three examiners scored according to the Likert scale. The Fleiss' kappa test was performed to analyse intra- and inter-examiner agreement, in addition to Cochran's Q test with a significance level of 5%, to compare the parameters of tube voltage, voxel size, and MAR filter.

Results: The Fleiss' kappa test showed excellent inter- and intraobserver agreement for all groups. All modalities of tube voltage, voxel size, and MAR filter showed very high accuracy, sensitivity, and specificity, providing diagnoses consistent with reality, achieving 99% accuracy when the model did not present adjacent metallic objects to the tooth, and 95% accuracy when such objects were present.

Conclusions: It is concluded that, although cone-beam computed tomography is not the exam of choice for diagnosing carious lesions, optimizing acquisition parameters and using MAR filters allows a reliable concomitant diagnosis in exams already indicated for other purposes.

Objectives: This study aimed to compare crestal facial bone measurements obtained from ultrasonography (USG) and cone-beam computed tomography (CBCT).

Methods: A total of 200 teeth from 15 systemically healthy individuals were included. Teeth were categorized as maxillary anterior (n = 50), maxillary posterior (n = 50), mandibular anterior (n = 50), and mandibular posterior (n = 50). Marginal bone level (MBL) and facial bone thickness at 1 mm (MBT-1), 2 mm (MBT-2), and 3 mm (MBT-3) apical to the bone crest were measured using both USG and CBCT. USG imaging utilized an 18 MHz transducer in B-mode, with standardized settings. Measurements were repeated twice by 2 independent examiners to assess intra- and inter-observer reliability. Interclass correlation coefficients (ICCs) and Bland-Altman plots were used for statistical comparisons.

Results: The ICCs between examiners ranged from 0.812 to 0.980. MBL, MBT-1, and MBT-2 measurements between ultrasound and CBCT readings showed excellent agreement (ICCs > 0.75). The agreement for MBT-3 in mandibular anterior was fair (ICC: 0.528). Overall, mean difference between the 2 methods for MBL was 0.06 mm and for MBT-1 was 0.018 mm, without systematic bias.

Conclusions: Ultrasound can be a valuable and reproducible tool for MBL and MBT-1 measurements, and it can serve as an alternative to CBCT. Despite reasonable agreement in MBT-2 and MBT-3, potential variability should be considered.

Advances in knowledge: While widely used for soft tissue measurements, USG has limited application in marginal alveolar bone assessment in living humans. This study demonstrates the potential use of USG in the evaluation of facial marginal alveolar bone in different regions of the oral cavity.

Objective: Inferior alveolar nerve (IAN) injury is a critical complication of mandibular third-molar extraction. This study aimed to construct and evaluate a deep learning framework that integrates contrastive learning and Bayesian optimization to enhance predictive performance using cone-beam computed tomography (CBCT) and panoramic radiographs.

Methods: A retrospective dataset of 902 panoramic radiographs and 1,500 CBCT images was used. Five deep learning architectures (MobileNetV2, ResNet101D, Vision Transformer, Twins-SVT, and SSL-ResNet50) were trained with and without contrastive learning and Bayesian optimization. Model performance was evaluated using accuracy, F1-score, and comparison with oral and maxillofacial surgeons (OMFSs).

Results: Contrastive learning significantly improved the F1-scores across all models (e.g., MobileNetV2: 0.302 to 0.740; ResNet101D: 0.188 to 0.689; Vision Transformer: 0.275 to 0.704; Twins-SVT: 0.370 to 0.719; SSL-ResNet50: 0.109 to 0.576). Bayesian optimization further enhanced the F1-scores for MobileNetV2 (from 0.740 to 0.923), ResNet101D (from 0.689 to 0.857), Vision Transformer (from 0.704 to 0.871), Twins-SVT (from 0.719 to 0.857), and SSL-ResNet50 (from 0.576 to 0.875). The AI model outperformed OMFSs on CBCT cross-sectional images (F1-score: 0.923 vs. 0.667) but underperformed on panoramic radiographs (0.666 vs. 0.730).

Conclusions: The proposed single-step deep learning approach effectively predicts IAN injury, with contrastive learning addressing data imbalance and Bayesian optimization enhancing model performance. While artificial intelligence surpasses human performance on CBCT images, panoramic radiograph analysis still benefits from expert interpretation. Future work should focus on multi-center validation and explainable artificial intelligence for broader clinical adoption.

Objectives: This study aimed to evaluate the diagnostic performance of machine learning (ML) algorithms based on radiomic features extracted from cone-beam CT (CBCT) images in differentiating the nasopalatine duct (NPD) from the nasopalatine duct cyst (NPDC), and to compare their performance with that of a dentomaxillofacial radiologist.

Methods: CBCT scans from 101 histopathologically confirmed NPDC cases and 101 age- and sex-matched controls with normal NPD were retrospectively analysed. Manual segmentation was performed to extract 1037 radiomic features (original, Laplacian of Gaussian, and wavelet-transformed). After dimensionality reduction, 5 ML models (support vector machine [SVM], random forest [RF], decision tree [DT], k-nearest neighbours [KNN], and logistic regression [LR]) were trained using 5-fold cross-validation. Performance was evaluated using the area under the ROC curve (AUC), sensitivity, specificity, precision, recall, and F1-score.

Results: Among the 11 optimal features identified through feature selection, large area high grey level emphasis and zone variance from the grey level size zone matrix (GLSZM) class were the most prominent. SVM achieved the highest performance in the test set (AUC and all other metrics = 1.00). The radiologist showed comparable but slightly lower overall performance than SVM (AUC = 0.94, with other metrics between 0.93 and 0.95).

Conclusions: ML algorithms based on radiomic features extracted from CBCT images can effectively differentiate NPD from NPDC. Unlike standard visual interpretation, this approach analyses quantitative image features via mathematical models, yielding objective and reproducible results. It may serve as a non-invasive, complementary decision-support tool, particularly in diagnostically challenging cases.

Objectives: To evaluate the impact of ambient exposure of photostimulable phosphor (PSP) plates and digital enhancement on detecting internal root resorption (IRR).

Methods: Thirty-five single-rooted teeth were selected, including 15 with artificially induced IRR (via 3-hour immersion in 37% hydrochloric acid) and 20 controls. Three repeated periapical radiographs were acquired of each tooth using the parallelling technique and PSP plates from the Express, VistaScan Mini, and CS 7600 digital radiographic imaging systems. For each set of 3 X-ray exposures, prior to scanning, one PSP plate was kept shielded from ambient light, another was exposed to ambient light for 5 seconds, while the third was exposed for 10 seconds. The presence of IRR in the total sample of 315 radiographs was assessed by 4 independent examiners using a 5-point scale. Initially, digital enhancement was not allowed, and these images were considered originals. A second round was conducted with adjustments permitted (enhanced radiographs). Sensitivity, specificity, and area under the receiver operating characteristic curve were calculated and compared using 2-way analysis of variance (α = 0.05).

Results: No significant differences were found among different light exposure times across all systems (P > .05). In the CS 7600, enhanced radiographs showed significantly higher sensitivity and lower specificity compared to originals (P < .05).

Conclusions: Ambient light exposure of PSP for up to 10 seconds does not compromise IRR diagnosis. Digital enhancement in CS 7600 may increase detection but reduce specificity, requiring cautious interpretation to avoid overdiagnosis.

Objectives: This study aimed to evaluate the diagnostic ability of relative values of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and diffusion kurtosis imaging (DKI) quantitative parameters for salivary gland tumours (SGTs).

Methods: A total of 107 patients with histopathologically confirmed SGTs (18 malignant [MTs], 38 pleomorphic adenomas [PAs], 31 Warthin tumours [WTs], 20 basal cell adenomas [BCAs]) underwent MRI with DKI and DCE-MRI sequences. Quantitative parameters included DCE-MRI-derived volume transfer constant (Ktrans), rate constant (Kep), fractional volume of the extravascular-extracellular space (Ve), plasma fraction (Vp), and DKI-derived mean kurtosis (MK) and mean diffusion (MD).The receiver operating characteristic (ROC) curve were used for statistical analysis. Statistical significance was set at P < 0.05.

Results: PAs exhibited the lowest MK (0.49 ± 0.15) among all groups (P < 0.05). Compared to WTs, PAs showed lower Kep (366.89 [260.06, 568.32]×10-³ min-1), higher MD (2.02 ± 0.42 × 10-³ mm2/s), and higher Ve (551.83 [388.10, 883.19]×10-³). PAs also displayed higher Ve, lower Kep, and lower Vp (85.42 [20.53, 332.72])×10-³) than BCAs, and lower Vp with higher Ve than MTs (all P < 0.05). WTs had significantly lower Ve (218.86 [142.07, 341.76]×10-³) than MTs (P = 0.001). BCAs demonstrated lower MK (0.61 ± 0.23) and higher MD (1.97 ± 0.44 × 10-³ mm2/s) compared to WTs and MTs (P < 0.05), alongside lower Ktrans (355.25 [211.88, 506.92]×10-³ min-1) and Ve (380.89 [271.28, 589.53]×10-³) than WTs (P < 0.05). Logistic regression analysis revealed enhanced discrimination: MK+Ve (AUC = 0.895) and MK + MD + Ve + Kep (AUC = 0.936) differentiated PAs from WTs; Kep + Vp + Ve distinguished PAs from BCAs (AUC = 0.843); MD + MK + Vp separated PAs from MTs (AUC = 0.854); Ktrans + MK differentiated WTs from BCAs (AUC = 0.856).

Conclusions: DKI and DCE-MRI parameters complement each other, enabling accurate SGT subtype differentiation. Combined use of these parameters achieves high diagnostic accuracy, and a stepwise diagnostic flowchart was designed to facilitate systematic discrimination among the four tumour types.

Objectives: To compare the measurement accuracy of cone-beam computed tomography (CBCT), high-resolution ultrasound (US), and intraoral scanning (IOS) with the gold-standard direct method in the measurement of peri-implant bone defects.

Methods: Forty standard-threaded and thirty-eight aggressive-threaded (Aggressor®) implants-identical in diameter and length (4.3/10 mm) but differing in macro-thread design-were placed into bovine rib bones in vitro. Dehiscence, 2/3-wall, and 4-wall (circumferential) defects were prepared around the implants. Each defect was measured for maximum width, depth, and height using CBCT, US, IOS, and direct manual measurement. Analyses were performed using the General Linear Model (ANOVA). A p-value < 0.05 was considered statistically significant.

Results: Intra-operator and inter-operator agreement showed high reliability (Gage R&R below 10%). For maximum width, defect type (F = 894.81, p < 0.001), method (F = 6.76, p < 0.001), and implant type (F = 5.39, p = 0.021) were significant. For maximum depth, defect type (F = 861.12, p < 0.001) and method (F = 3.39, p = 0.018) were significant. For maximum height, method (F = 12.62, p < 0.001) and defect type (F = 38.91, p < 0.001) were significant. The model demonstrated high explanatory power for width (R2=75.9%) and depth (R2=76.6%) measurements but lower for height (R2=20.7%). CBCT provided the most consistent results relative to direct measurements, followed by US, whereas IOS showed greater deviations.

Conclusions: CBCT showed the highest agreement with direct measurements, followed by US, while IOS exhibited greater variability. Defect type and measurement modality were the primary determinants of accuracy. These findings indicate that CBCT and US can be considered reliable tools for assessing peri-implant bone defects.

Objective: Radiation exposure remains a major concern in dentomaxillofacial imaging, especially for children and young adults who require repeated examinations. A precise, non-ionizing alternative for craniofacial measurement would have significant implications for clinical practice. Computed tomography (CT) and cephalometry are widely used but expose patients to ionizing radiation. Zero Echo Time Magnetic Resonance Imaging (ZTE-MRI) enable to visualize short-T2 tissues, including cortical bone, without radiation by minimizing TE toward zero; however, clinical head-and-neck use has been limited. The goal of the present study was to investigate the possibility of performing precise three-dimensional measurements of the craniofacial regions using ZTE-MRI.

Materials and methods: Twenty-nine participants (mean age 32.14±17.09 years; 23 female, 6 male) underwent MDCT and ZTE-MRI. Standard cephalometric landmarks were identified, yielding 27 linear and 21 angular variables. For each variable, linear regression analysis was performed for the results on MDCT (Multi-Detector CT) images and ZTE-MRI.

Results: Nineteen linear and eight angular variables had regression coefficients between 0.9 and 1.1 and 28 of the linear and 16 of the angular variables had R-square values of 0.81 or higher. 25 linear and 16 angular variables showed Cohen's effect sizes within the ±0.8 range.

Conclusions: Variables derived from skeletal reference points showed clinically acceptable results, indicating that ZTE-MRI is useful for these measurements. In contrast, variables involving dental reference points demonstrated lower agreement, making accurate measurement in these regions difficult.

Advances in knowledge: To our knowledge, this study is the first to show that ZTE-MRI allows feasible linear and angular craniofacial measurements used in dental practice and demonstrates good agreement with MDCT for major skeletal landmarks. These findings support its potential future clinical applicability for radiation-free bone assessment; however, further research is warranted to establish its performance across all measurements.

Objectives: This study aimed to analyze the correlation between disc position and condylar position and morphology through fused cone-beam computed tomography (CBCT) and magnetic resonance (MR) images.

Methods: Patients exhibiting temporomandibular disorder symptoms were included, and joints with poor osseous consistence were excluded. Angle of disc was measured in the fused image using the method proposed in this study. Joint spaces were measured, and condylar morphology was assessed in cone-beam computed tomography images. Statistical analysis was performed to examine the reliability of measurement method and the correlation between disc position and condylar position/morphology. A logistic regression model was used for identifying factors associated with anterior disc displacement.

Results: Our results showed that inter- and intra-observer agreement for measurements of disc angle and joint space were excellent (Intraclass correlation coefficient >0.9). Superior joint space, posterior joint space, and natural logarithm of the posterior-to-anterior joint space ratio showed significant correlations with the angle (P < 0.01) and significant differences between groups (P < 0.01). The posterior-to-anterior joint space ratio was significantly smaller in the mild displacement group. The logistic regression model demonstrated that beak-like shape in oblique sagittal view (OR = 5.235, P < 0.05) and reduced posterior-to-anterior ratio (OR = 0.301, P < 0.05) significantly increased the risk of anterior disc displacement.

Conclusions: Condylar position and morphology demonstrated statistically significant association with disc position. Multivariate logistic regression analysis revealed that condylar position and morphology in sagittal views in cone-beam computed tomography images can serve as indicators for disc displacement.

Objectives: Federated learning (FL) may mitigate privacy constraints, heterogeneous data quality, and inconsistent labeling in dental diagnostic artificial intelligence (AI). FL was compared with centralized (CL) and local learning (LL) for tooth segmentation in panoramic radiographs across multiple data corruption scenarios.

Methods: An Attention U-Net was trained on 2066 radiographs from six institutions across four settings: baseline (unaltered data); label manipulation (dilated/missing annotations); image-quality manipulation (additive Gaussian noise); and exclusion of one faulty client with corrupted data. FL was implemented via the Flower AI framework. Per-client training- and validation loss trajectories were monitored for anomaly detection and a set of metrics (Dice, IoU, HD, HD95 and ASSD) were evaluated on a hold-out test set. From these metrics significance results were reported through Wilcoxon signed-rank test. CL and LL served as comparators.

Results: Baseline: FL achieved a median Dice of 0.94889 (ASSD: 1.33229), slightly better than CL at 0.94706 (ASSD: 1.37074) and LL at 0.93557-0.94026 (ASSD: 1.51910-1.69777). Label manipulation: FL maintained the best median Dice score at 0.94884 (ASSD: 1.46487) versus CL's 0.94183 (ASSD: 1.75738) and LL's 0.93003-0.94026 (ASSD: 1.51910-2.11462). Similar performance was observed when two faulty clients were introduced. Image noise: FL led with Dice at 0.94853 (ASSD: 1.31088); CL scored 0.94787 (ASSD: 1.36131); LL ranged from 0.93179-0.94026 (ASSD: 1.51910-1.77350). Similar performance was observed when two faulty clients were introduced, with CL performing slightly better than FL. Faulty-client exclusion: FL reached Dice at 0.94790 (ASSD: 1.33113) better than CL's 0.94550 (ASSD: 1.39318). Loss-curve monitoring reliably flagged the corrupted site.

Conclusions: FL matches or exceeds CL and outperforms LL across corruption scenarios while preserving privacy. Per-client loss trajectories provide an effective anomaly-detection mechanism and support FL as a practical, privacy-preserving approach for scalable clinical AI deployment.