Dento maxillo facial radiology最新文献

Objectives: To develop a content-aware chatbot based on GPT-3.5-Turbo and GPT-4 with specialized knowledge on the German S2 Cone-Beam CT (CBCT) dental imaging guideline and to compare the performance against humans.

Methods: The LlamaIndex software library was used to integrate the guideline context into the chatbots. Based on the CBCT S2 guideline, 40 questions were posed to content-aware chatbots and early career and senior practitioners with different levels of experience served as reference. The chatbots' performance was compared in terms of recommendation accuracy and explanation quality. Chi-square test and one-tailed Wilcoxon signed rank test evaluated accuracy and explanation quality, respectively.

Results: The GPT-4 based chatbot provided 100% correct recommendations and superior explanation quality compared to the one based on GPT3.5-Turbo (87.5% vs. 57.5% for GPT-3.5-Turbo; P = .003). Moreover, it outperformed early career practitioners in correct answers (P = .002 and P = .032) and earned higher trust than the chatbot using GPT-3.5-Turbo (P = 0.006).

Conclusions: A content-aware chatbot using GPT-4 reliably provided recommendations according to current consensus guidelines. The responses were deemed trustworthy and transparent, and therefore facilitate the integration of artificial intelligence into clinical decision-making.

Objectives: Instance-level tooth segmentation extracts abundant localization and shape information from panoramic radiographs (PRs). The aim of this study was to evaluate the performance of a mask refinement network that extracts precise tooth edges.

Methods: A public dataset which consists of 543 PRs and 16211 labelled teeth was utilized. The structure of a typical Mask Region-based Convolutional Neural Network (Mask RCNN) was used as the baseline. A novel loss function was designed focus on producing accurate mask edges. In addition to our proposed method, 3 existing tooth segmentation methods were also implemented on the dataset for comparative analysis. The average precisions (APs), mean intersection over union (mIoU), and mean Hausdorff distance (mHAU) were exploited to evaluate the performance of the network.

Results: A novel mask refinement region-based convolutional neural network was designed based on Mask RCNN architecture to extract refined masks for individual tooth on PRs. A total of 3311 teeth were correctly detected from 3382 tested teeth in 111 PRs. The AP, precision, and recall were 0.686, 0.979, and 0.952, respectively. Moreover, the mIoU and mHAU achieved 0.941 and 9.7, respectively, which are significantly better than the other existing segmentation methods.

Conclusions: This study proposed an efficient deep learning algorithm for accurately extracting the mask of any individual tooth from PRs. Precise tooth masks can provide valuable reference for clinical diagnosis and treatment. This algorithm is a fundamental basis for further automated processing applications.

Objectives: To evaluate the effect of the presence and the number of high-density objects in the exomass on the volume of a high-density object in cone-beam CT (CBCT).

Methods: Cylinders of cobalt-chromium (Co-Cr), titanium (Ti), and zirconium (Zi) were inserted into a polymethylmethacrylate phantom in five different combinations of number and position: 1-no cylinder; 2-one cylinder in a posterior region; 3-one cylinder in an anterior region; 4-two cylinders in posterior regions; and 5-three cylinders in anterior and posterior regions. The phantom underwent CBCT scanning using OP300 and X800 systems, with the afore mentioned cylinders of the same composition placed in the exomass and an additional high-density cylinder placed in the centre of the field of view (FOV), corresponding to the left-anterior region. The tomographic volume of the cylinder inside the FOV was measured using semi-automatic segmentation. The volumetric alteration (VA) between the segmented and physical volumes, in percentage, was compared among the experimental groups using repeated measures ANOVA and Tukey post-hoc (α = 5%).

Results: The factors material, combination, and their interaction affected the volume or both CBCT systems. In OP300, more cylinders in the exomass reduced the VA, mainly for Co-Cr. In X800, more cylinders in the exomass tended to increase the VA inside the FOV, except for Zi.

Conclusions: In general, the presence of high-density objects in the exomass influences the VA of the object inside the FOV, although this oscillates according to object composition, number and position in the exomass, and CBCT system.

Objectives: The purpose of this study is to prospectively investigate the reference values of masseter and temporal muscle thicknesses by ultrasonography and muscle hardness values by shear wave elastography in healthy adults.

Methods: The sample of the study consisted of a total of 160 healthy individuals aged between 18 and 59, including 80 women and 80 men. By examining the right and left sides of each participant, thickness and hardness values were obtained for 320 masseter muscles and 320 temporal muscles in total.

Results: The mean masseter muscle thickness was found to be 1.09 cm at rest and 1.40 cm in contraction. The mean temporal muscle thickness was found to be 0.88 cm at rest and 0.98 cm in contraction. The thickness values of the masseter and temporal muscles were significantly greater in the male participants than in the female participants (P < .001). While there were significant differences between the right and left masseter muscle thickness values at rest and in contraction, the values of the temporal muscles did not show a significant difference between the sides. While the resting hardness (rSWE) of the masseter muscle was transversally 6.91 kPa and longitudinally 8.49 kPa, these values in contraction (cSWE) were found, respectively, 31.40 and 35.65 kPa. The median temporal muscle hardness values were 8.84 kPa at rest and 20.43 kPa in contraction. Masseter and temporal muscle hardness values at rest and in contraction were significantly higher among the male participants compared to the female participants (P < .001).

Conclusion: In this study, reference values for the thickness and hardness of the masseter and temporal muscles are reported. Knowing these values will make it easier to assess pain in the masseter and temporal muscles and determine the diagnosis and prognosis of masticatory muscle pathologies by allowing the morphological and functional assessments of these muscles, and it will identify ranges for reference parameters.

Objectives: The objectives of this study are to explore and evaluate the automation of anatomical landmark localization in cephalometric images using machine learning techniques, with a focus on feature extraction and combinations, contextual analysis, and model interpretability through Shapley Additive exPlanations (SHAP) values.

Methods: We conducted extensive experimentation on a private dataset of 300 lateral cephalograms to thoroughly study the annotation results obtained using pixel feature descriptors including raw pixel, gradient magnitude, gradient direction, and histogram-oriented gradient (HOG) values. The study includes evaluation and comparison of these feature descriptions calculated at different contexts namely local, pyramid, and global. The feature descriptor obtained using individual combinations is used to discern between landmark and nonlandmark pixels using classification method. Additionally, this study addresses the opacity of LGBM ensemble tree models across landmarks, introducing SHAP values to enhance interpretability.

Results: The performance of feature combinations was assessed using metrics like mean radial error, standard deviation, success detection rate (SDR) (2 mm), and test time. Remarkably, among all the combinations explored, both the HOG and gradient direction operations demonstrated significant performance across all context combinations. At the contextual level, the global texture outperformed the others, although it came with the trade-off of increased test time. The HOG in the local context emerged as the top performer with an SDR of 75.84% compared to others.

Conclusions: The presented analysis enhances the understanding of the significance of different features and their combinations in the realm of landmark annotation but also paves the way for further exploration of landmark-specific feature combination methods, facilitated by explainability.

Purpose: This study investigated the differences in subjective and objective image parameters as well as dose exposure of photon-counting CT (PCCT) compared to cone-beam CT (CBCT) in paranasal sinus imaging for the assessment of rhinosinusitis and sinonasal anatomy.

Methods: This single-centre retrospective study included 100 patients, who underwent either clinically indicated PCCT or CBCT of the paranasal sinus. Two blinded experienced ENT radiologists graded image quality and delineation of specific anatomical structures on a 5-point Likert scale. In addition, contrast-to-noise ratio (CNR) and applied radiation doses were compared among both techniques.

Results: Image quality and delineation of bone structures in paranasal sinus PCCT was subjectively rated superior by both readers compared to CBCT (P < .001). CNR was significantly higher for photon-counting CT (P < .001). Mean effective dose for PCCT examinations was significantly lower than for CBCT (0.038 mSv ± 0.009 vs. 0.14 mSv ± 0.011; P < .001).

Conclusion: In a performance comparison of PCCT and a modern CBCT scanner in paranasal sinus imaging, we demonstrated that first-use PCCT in clinical routine provides higher subjective image quality accompanied by higher CNR at close to a quarter of the dose exposure compared to CBCT.

Objectives: To evaluate the accuracy of major salivary gland ultrasonography (SGUS) in relation to minor salivary gland biopsy (mSGB) in the diagnosis of Sjögren's syndrome (SS).

Methods: A systematic review and meta-analysis were performed. Ten databases were searched to identify studies that compared the accuracy of SGUS and mSGB. The risk of bias was assessed, data were extracted, and univariate and bivariate random-effects meta-analyses were done.

Results: A total of 5000 records were identified; 13 studies were included in the qualitative synthesis and 10 in the quantitative synthesis. The first meta-analysis found a sensitivity of 0.86 (95% CI: 0.74-0.92) and specificity of 0.87 (95% CI: 0.81-0.92) for the predictive value of SGUS scoring in relation to the result of mSGB. In the second meta-analysis, mSGB showed higher sensitivity and specificity than SGUS. Sensitivity was 0.80 (95% CI: 0.74-0.85) for mSGB and 0.71 (95% CI: 0.58-0.81) for SGUS, and specificity was 0.94 (95% CI: 0.87-0.97) for mSGB and 0.89 (95% CI: 0.82-0.94) for SGUS.

Conclusions: The diagnostic accuracy of SGUS was similar to that of mSGB. SGUS is an effective diagnostic test that shows good sensitivity and high specificity, in addition to being a good tool for prognosis and for avoiding unnecessary biopsies. More studies using similar methodologies are needed to assess the accuracy of SGUS in predicting the result of mSGB. Our results will contribute to decision-making for the implementation of SGUS as a diagnostic tool for SS, considering the advantages of this method.

Objectives: The objective of this study is to assess the accuracy of computer-assisted periodontal classification bone loss staging using deep learning (DL) methods on panoramic radiographs and to compare the performance of various models and layers.

Methods: Panoramic radiographs were diagnosed and classified into 3 groups, namely "healthy," "Stage1/2," and "Stage3/4," and stored in separate folders. The feature extraction stage involved transferring and retraining the feature extraction layers and weights from 3 models, namely ResNet50, DenseNet121, and InceptionV3, which were proposed for classifying the ImageNet dataset, to 3 DL models designed for classifying periodontal bone loss. The features obtained from global average pooling (GAP), global max pooling (GMP), or flatten layers (FL) of convolutional neural network (CNN) models were used as input to the 8 different machine learning (ML) models. In addition, the features obtained from the GAP, GMP, or FL of the DL models were reduced using the minimum redundancy maximum relevance (mRMR) method and then classified again with 8 ML models.

Results: A total of 2533 panoramic radiographs, including 721 in the healthy group, 842 in the Stage1/2 group, and 970 in the Stage3/4 group, were included in the dataset. The average performance values of DenseNet121 + GAP-based and DenseNet121 + GAP + mRMR-based ML techniques on 10 subdatasets and ML models developed using 2 feature selection techniques outperformed CNN models.

Conclusions: The new DenseNet121 + GAP + mRMR-based support vector machine model developed in this study achieved higher performance in periodontal bone loss classification compared to other models in the literature by detecting effective features from raw images without the need for manual selection.

Objectives: To compare artificial intelligence (AI)-driven web-based platform and manual measurements for analysing facial asymmetry in craniofacial CT examinations.

Methods: The study included 95 craniofacial CT scans from patients aged 18-30 years. The degree of asymmetry was measured based on AI platform-predefined anatomical landmarks: sella (S), condylion (Co), anterior nasal spine (ANS), and menton (Me). The concordance between the results of automatic asymmetry reports and manual linear 3D measurements was calculated. The asymmetry rate (AR) indicator was determined for both automatic and manual measurements, and the concordance between them was calculated. The repeatability of manual measurements in 20 randomly selected subjects was assessed. The concordance of measurements of quantitative variables was assessed with interclass correlation coefficient (ICC) according to the Shrout and Fleiss classification.

Results: Erroneous AI tracings were found in 16.8% of cases, reducing the analysed cases to 79. The agreement between automatic and manual asymmetry measurements was very low (ICC < 0.3). A lack of agreement between AI and manual AR analysis (ICC type 3 = 0) was found. The repeatability of manual measurements and AR calculations showed excellent correlation (ICC type 2 > 0.947).

Conclusions: The results indicate that the rate of tracing errors and lack of agreement with manual AR analysis make it impossible to use the tested AI platform to assess the degree of facial asymmetry.

Objectives: This study aims to evaluate the morphological features of gubernacular tract (GT) for erupting permanent mandibular canines at different ages from 5 to 9 years old with a three-dimensional (3D) measurement method.

Methods: The cone-beam CT images of 50 patients were divided into five age groups. The 3D models of the GT for mandibular canines were reconstructed and analysed. The characteristics of the GT, including length, diameter, ellipticity, tortuosity, superficial area, volume, and the angle between the canine and GT, were evaluated using a centreline fitting algorithm.

Results: Among the 100 GTs that were examined, the length of the GT for mandibular canines decreased between the ages of 5 and 9 years, while the diameter increased until the age of 7 years. Additionally, the ellipticity and tortuosity of the GT decreased as age advanced. The superficial area and volume exhibited a trend of initially increasing and then decreasing. The morphological variations of the GT displayed heterogeneous changes during different periods.

Conclusions: The 3D measurement method effectively portrayed the morphological attributes of the GT for mandibular canines. The morphological characteristics of the GT during the eruption process exhibited significant variations. The variations in morphological changes may indicate different stages of mandibular canine eruption.