Dento maxillo facial radiology最新文献

{"title":"Development and evaluation of a deep learning model to reduce exomass-related metal artefacts in cone-beam CT: an ex vivo study using porcine mandibles.","authors":"Matheus L Oliveira, Susanne Schaub, Dorothea Dagassan-Berndt, Florentin Bieder, Philippe C Cattin, Michael M Bornstein","doi":"10.1093/dmfr/twae062","DOIUrl":"10.1093/dmfr/twae062","url":null,"abstract":"<p><strong>Objectives: </strong>To develop and evaluate a deep learning (DL) model to reduce metal artefacts originating from the exomass in cone-beam CT (CBCT) of the jaws.</p><p><strong>Methods: </strong>Five porcine mandibles, each featuring six tubes filled with a radiopaque solution, were scanned using four CBCT units before and after the incremental insertion of up to three titanium, titanium-zirconium, and zirconia dental implants in the exomass of a small field of view. A conditional denoising diffusion probabilistic model, using DL techniques, was employed to correct axial images from exomass-related metal artefacts across the CBCT units and implant scenarios. Three examiners independently scored the image quality of all datasets, including those without an implant (ground truth), with implants in the exomass (original), and DL-generated ones. Quantitative analysis compared contrast-to-noise ratio (CNR) to validate artefact reduction using repeated measures analysis of variance in a factorial design followed by Tukey test (α = .05).</p><p><strong>Results: </strong>The visualisation of the hard tissues and overall image quality was reduced in the original and increased in the DL-generated images. The score variation observed in the original images was not observed in the DL-generated images, which generally scored higher than the original images. DL-generated images revealed significantly greater CNR than both the ground truth and their corresponding original images, regardless of the material and quantity of dental implants and the CBCT unit (P < .05). Original images revealed significantly lower CNR than the ground truth (P < .05).</p><p><strong>Conclusions: </strong>The developed DL model using porcine mandibles demonstrated promising performance in correcting exomass-related metal artefacts in CBCT, serving as a proof-of-principle for future applications of this approach.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"109-117"},"PeriodicalIF":2.9,"publicationDate":"2025-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11784918/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142727041","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Automated tooth segmentation in magnetic resonance scans using deep learning - A pilot study.","authors":"Tabea Flügge, Shankeeth Vinayahalingam, Niels van Nistelrooij, Stefanie Kellner, Tong Xi, Bram van Ginneken, Stefaan Bergé, Max Heiland, Florian Kernen, Ute Ludwig, Kento Odaka","doi":"10.1093/dmfr/twae059","DOIUrl":"10.1093/dmfr/twae059","url":null,"abstract":"<p><strong>Objectives: </strong>The main objective was to develop and evaluate an artificial intelligence model for tooth segmentation in magnetic resonance (MR) scans.</p><p><strong>Methods: </strong>MR scans of 20 patients performed with a commercial 64-channel head coil with a T1-weighted 3D-SPACE (Sampling Perfection with Application Optimized Contrasts using different flip angle Evolution) sequence were included. Sixteen datasets were used for model training and 4 for accuracy evaluation. Two clinicians segmented and annotated the teeth in each dataset. A segmentation model was trained using the nnU-Net framework. The manual reference tooth segmentation and the inferred tooth segmentation were superimposed and compared by computing precision, sensitivity, and Dice-Sørensen coefficient. Surface meshes were extracted from the segmentations, and the distances between points on each mesh and their closest counterparts on the other mesh were computed, of which the mean (average symmetric surface distance) and 95th percentile (Hausdorff distance 95%, HD95) were reported.</p><p><strong>Results: </strong>The model achieved an overall precision of 0.867, a sensitivity of 0.926, a Dice-Sørensen coefficient of 0.895, and a 95% Hausdorff distance of 0.91 mm. The model predictions were less accurate for datasets containing dental restorations due to image artefacts.</p><p><strong>Conclusions: </strong>The current study developed an automated method for tooth segmentation in MR scans with moderate to high effectiveness for scans with respectively without artefacts.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"12-18"},"PeriodicalIF":2.9,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11664100/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142727040","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluation of temporomandibular joint disc displacement with MRI-based radiomics analysis.","authors":"Hazal Duyan Yüksel, Kaan Orhan, Burcu Evlice, Ömer Kaya","doi":"10.1093/dmfr/twae066","DOIUrl":"10.1093/dmfr/twae066","url":null,"abstract":"<p><strong>Objectives: </strong>The purpose of this study was to propose a machine learning model and assess its ability to classify temporomandibular joint (TMJ) disc displacements on MR T1-weighted and proton density-weighted images.</p><p><strong>Methods: </strong>This retrospective cohort study included 180 TMJs from 90 patients with TMJ signs and symptoms. A radiomics platform was used to extract imaging features of disc displacements. Thereafter, different machine learning algorithms and logistic regression were implemented on radiomics features for feature selection, classification, and prediction. The radiomics features included first-order statistics, size- and shape-based features, and texture features. Six classifiers, including logistic regression, random forest, decision tree, k-nearest neighbours (KNN), XGBoost, and support vector machine were used for a model building which could predict the TMJ disc displacements. The performance of models was evaluated by sensitivity, specificity, and ROC curve.</p><p><strong>Results: </strong>KNN classifier was found to be the most optimal machine learning model for prediction of TMJ disc displacements. The AUC, sensitivity, and specificity for the training set were 0.944, 0.771, 0.918 for normal, anterior disc displacement with reduction (ADDwR) and anterior disc displacement without reduction (ADDwoR) while testing set were 0.913, 0.716, and 1 for normal, ADDwR, and ADDwoR. For TMJ disc displacements, skewness, root mean squared, kurtosis, minimum, large area low grey level emphasis, grey level non-uniformity, and long-run high grey level emphasis, were selected as optimal features.</p><p><strong>Conclusions: </strong>This study has proposed a machine learning model by KNN analysis on TMJ MR images, which can be used for TMJ disc displacements.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"19-27"},"PeriodicalIF":2.9,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142738694","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The impact of cone beam CT on outcomes associated with endodontic access cavity preparation: a controlled human analogue study using 3D-printed first maxillary molars.","authors":"Margarete B McGuigan, Henry F Duncan, Gabriel Krastl, Julia Ludwig, Bahman Honari, Keith Horner","doi":"10.1093/dmfr/twae048","DOIUrl":"10.1093/dmfr/twae048","url":null,"abstract":"<p><strong>Objectives: </strong>To identify if supplemental preoperative cone beam CT (CBCT) imaging could improve outcomes related to endodontic access cavity preparation, using 3D-printed maxillary first molars (M1Ms) in a rigorously simulated, controlled human analogue study.</p><p><strong>Methods: </strong>Eighteen operators with 3 experience-levels took part in 2 simulated clinical sessions, 1 with and 1 without the availability of CBCT imaging, in a randomized order and with an intervening 8-week washout period. Operators attempted the location of all 4 root canals in each of 3 custom-made M1Ms (2 non-complex and 1 complex mesiobuccal [MB] canal anatomy). The primary outcome was tooth volume removed. Secondary outcomes were linear cavity dimensions, canals located, and procedural time. Operator confidence and \"helpfulness\" of available imaging were recorded. Statistical analysis of data included: paired t-tests, Fisher's exact test, linear mixed-effect modelling, and Mann-Whitney U test, with an alpha level of .05 for all.</p><p><strong>Results: </strong>When supplemental preoperative CBCT was available, there were significant reductions in volume of the access cavity and procedural times, with significantly increased MB2 canal location, but only for teeth with non-complex anatomies and for more experienced operators. Linear mixed-effect modelling identified image type and operator experience as significant predictors of tooth volume removed and procedural time. There was significantly lower confidence in canal location and perceived \"helpfulness\" (all Experience Groups) when conventional imaging only was used compared with when CBCT was available.</p><p><strong>Conclusions: </strong>Supplemental preoperative CBCT had several beneficial impacts on access cavity preparation, although this only applied to teeth with non-complex anatomy and for more experienced operators.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"43-55"},"PeriodicalIF":2.9,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11664103/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142343506","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Carotid calcifications in panoramic radiographs can predict vascular risk.","authors":"Maria Garoff, Jan Ahlqvist, Eva Levring Jäghagen, Per Wester, Elias Johansson","doi":"10.1093/dmfr/twae057","DOIUrl":"10.1093/dmfr/twae057","url":null,"abstract":"<p><strong>Objectives: </strong>Carotid artery calcification (CAC) is occasionally detected in panoramic radiographs (PRs). Bilateral vessel-outlining (BVO) CACs are independent risk markers for future vascular events and have been associated with large plaque area. If accounting for plaque area, BVO CACs may no longer be an independent risk marker for vascular events. The aim of this study was to explore the association between BVO CACs and vascular events and its relationship with carotid ultrasound plaque area.</p><p><strong>Methods: </strong>In this cohort study we prospectively included 212 consecutive participants with CACs detected in PR that were performed to plan and evaluate odontologic treatment. Of these 212, 43 (20%) had BVO CACs. Plaque area was assessed with ultrasound at baseline. Primary outcome was major adverse cardiovascular events (MACEs) during follow-up.</p><p><strong>Results: </strong>Vessel-outlining CAC was associated with larger plaque area on the same side (P = .03) and BVO CACs were associated with larger total plaque area (both sides summed) than other CAC features (P = .004). Mean follow-up was 7.0 years and 72 (34%) participants had more than 1 MACE. In bivariable analyses, both BVO CACs (HR 2.5, P < .001) and total plaque area (HR 1.8 per cm2, P = .008) were associated with MACE. When entering BVO CACs, plaque area and other relevant co-variates in a multivariable model, BVO CACs were virtually unchanged (HR 2.4, P = .001), but total plaque area was no longer significant (HR 1.0, P = .92).</p><p><strong>Conclusion: </strong>Present results support the contention that BVO CACs are a stronger predictor for future vascular events than carotid ultrasound plaque area.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"28-34"},"PeriodicalIF":2.9,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11664099/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142681094","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Development and validation of a CT-based deep learning radiomics signature to predict lymph node metastasis in oropharyngeal squamous cell carcinoma: a multicentre study.","authors":"Tianzi Jiang, Hexiang Wang, Jie Li, Tongyu Wang, Xiaohong Zhan, Jingqun Wang, Ning Wang, Pei Nie, Shiyu Cui, Xindi Zhao, Dapeng Hao","doi":"10.1093/dmfr/twae051","DOIUrl":"10.1093/dmfr/twae051","url":null,"abstract":"<p><strong>Objectives: </strong>Lymph node metastasis (LNM) is a pivotal determinant that influences the treatment strategies and prognosis for oropharyngeal squamous cell carcinoma (OPSCC) patients. This study aims to establish and verify a deep learning (DL) radiomics model for the prediction of LNM in OPSCCs using contrast-enhanced computed tomography (CECT).</p><p><strong>Methods: </strong>A retrospective analysis included 279 OPSCC patients from 3 institutions. CECT images were used for handcrafted (HCR) and DL feature extraction. Dimensionality reduction for HCR features used recursive feature elimination (RFE) and least absolute shrinkage and selection operator (LASSO) algorithms, whereas DL feature dimensionality reduction used variance-threshold and RFE algorithms. Radiomics signatures were constructed using six machine learning classifiers. A combined model was then constructed using the screened DL, HCR, and clinical features. The area under the receiver operating characteristic curve (AUC) served to quantify the model's performance, and calibration curves were utilized to assess its calibration.</p><p><strong>Results: </strong>The combined model exhibited robust performance, achieving AUC values of 0.909 (95% CI, 0.861-0.957) in the training cohort, 0.884 (95% CI, 0.800-0.968) in the internal validation cohort, and 0.865 (95% CI, 0.791-0.939) in the external validation cohort. It outperformed both the clinical model and best-performing radiomics model. Moreover, calibration was deemed satisfactory.</p><p><strong>Conclusions: </strong>The combined model based on CECT demonstrates the potential to predict LNM in OPSCCs preoperatively, offering a valuable tool for more precise and tailored treatment strategies.</p><p><strong>Advances in knowledge: </strong>This study presents a novel combined model integrating clinical factors with DL radiomics, significantly enhancing preoperative LNM prediction in OPSCC.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"77-87"},"PeriodicalIF":2.9,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142281917","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Diffusion-weighted imaging of the head and neck: comparison between integrated slice-specific dynamic shimming and simultaneous multi-slice readout-segmented echo-planar sequences.","authors":"Tong Su, Xiaoli Zhu, Yu Chen, Zhentan Xu, Xingming Chen, Tao Zhang, Jinxia Zhu, Wei Liu, Xiaoye Wang, Zhuhua Zhang, Feng Feng, Zhengyu Jin","doi":"10.1093/dmfr/twae047","DOIUrl":"10.1093/dmfr/twae047","url":null,"abstract":"<p><strong>Objectives: </strong>To compare integrated slice-specific dynamic shimming (iShim) and simultaneous multi-slice (SMS) readout-segmented echo-planar imaging (RESOLVE) for diffusion-weighted imaging (DWI) of malignant head and neck tumours.</p><p><strong>Methods: </strong>In this prospective study, 45 patients with malignant head and neck lesions underwent iShim- and SMS-RESOLVE imaging with two b-values (0, 800 s/mm2) at 3 T. Subjective image quality scores (lesion distortion, signal loss, fat saturation, and artefacts), quantitative lesion distortion, quantitative image quality [signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and SNR efficiency], ADC values, and total acquisition times of iShim- and SMS-RESOLVE imaging were evaluated and compared.</p><p><strong>Results: </strong>iShim-RESOLVE (3 mins 7 s) had longer acquisition times than SMS-RESOLVE (2 min 1 s). iShim-RESOLVE images had much better overall image quality, geometric distortion, fat saturation, and signal loss than SMS-RESOLVE (P < .05). The quantitative tumour distortion in iShim-RESOLVE images was significantly smaller than those of SMS-RESOLVE, measured either with b = 0 s/mm2 or b = 800 s/mm2 (P < .05). Using the iShim technique, the SNR, CNR of high b-value images, and SNR efficiency were significantly higher than those of SMS-RESOLVE (P < .001). The ADC measurements with excellent agreement of iShim-RESOLVE images were higher than those of SMS-RESOLVE.</p><p><strong>Conclusions: </strong>iShim-RESOLVE can decrease lesion distortion and improve overall image quality compared with SMS-RESOLVE in head and neck region.</p><p><strong>Advances in knowledge: </strong>iShim-RESOLVE DWI may serve as a promising technique for reducing distortion and assessing head and neck lesions in clinical practice.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"35-42"},"PeriodicalIF":2.9,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142827444","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"In vitro early proximal caries detection using trilateral short-wave infrared reflection at 1050 and 1550 nm.","authors":"Katrin Heck, Nils Werner, Lea Hoffmann, Falk Schwendicke, Friederike Litzenburger","doi":"10.1093/dmfr/twae049","DOIUrl":"10.1093/dmfr/twae049","url":null,"abstract":"<p><strong>Objectives: </strong>This in vitro study evaluated the diagnostic potential of short-wave infrared reflection (SWIRR) at 1050 and 1550 nm for proximal caries detection from the occlusal, buccal, and lingual surfaces of posterior teeth under clinically relevant conditions. Bitewing radiography (BWR) was the alternative index test and micro-computed tomography (μCT) was the reference standard.</p><p><strong>Methods: </strong>Two hundred and fifty proximal surfaces of extracted human teeth were examined using SWIRR at 1050 and 1550 nm and BWR. SWIRR, BWR, and μCT findings were evaluated twice by 2 trained examiners. SWIRR images were evaluated from occlusal and trilateral (occlusal, buccal, and lingual combined) views. Sensitivity, specificity, and area under the curves were calculated. Reliability assessment was performed using κ statistics.</p><p><strong>Results: </strong>Short-wave infrared reflection (1050 nm) showed sensitivity of 0.44 for occlusal and 0.55 for trilateral assessment, paired with specificity of 0.96 and 0.90, whereas SWIRR (1550 nm) showed sensitivity of 0.73 and 0.85 paired with specificity of 0.76 and 0.59. Compared to occlusal view, trilateral SWIRR view revealed ≈10% higher sensitivity and lower specificity. BWR revealed lowest sensitivity (0.30) and highest specificity (0.99). Over- and underestimation of caries demonstrated opposite trends: from 1050 to 1550 nm, overestimation of trilateral SWIRR increased (0.08-0.29), while underestimation decreased (0.15-0.06).</p><p><strong>Conclusion: </strong>Trilateral SWIRR has higher sensitivity and lower specificity for proximal caries, than occlusal SWIRR. For trilateral SWIRR, wavelengths around 1050 nm are more suitable, while 1550 nm is better for occlusal examinations. A combination of SWIRR at 1050 and 1550 nm may exhibit a balanced sensitivity and specificity for proximal caries.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"70-76"},"PeriodicalIF":2.9,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142343505","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluation of temporomandibular joint osteoarthritis using a new FRACTURE sequence of 3.0T MRI.","authors":"Michihito Nozawa, Motoki Fukuda, Shinya Kotaki, Daisuke Tomoda, Ayaka Morishita, Hironori Akiyama, Yoshiko Ariji","doi":"10.1093/dmfr/twae065","DOIUrl":"10.1093/dmfr/twae065","url":null,"abstract":"<p><strong>Objectives: </strong>The purpose of this study was to determine the usefulness of a new MRI sequence, CT-like fast field echo with limited echo-spacing (FRACTURE), in diagnosing temporomandibular joint (TMJ) osteoarthritis compared with routine MRI TMJ sequences.</p><p><strong>Methods: </strong>The study sample comprised 76 patients (152 joints) who underwent MRI and CT examinations to diagnose TMJ disorders. Two specialists in oral and maxillofacial radiology assessed the bony changes of the TMJ on FRACTURE, proton density-weighted (PDw), and fat-suppression T2-weighted (T2wFS) sequences. Receiver operating characteristic curves were plotted for each sequence, and the accuracy, sensitivity, specificity, and area under the curve (AUC) were calculated. Additionally, the interobserver agreement (Cohen's kappa value) and sensitivity in assessing each osteoarthritis finding were calculated for each sequence.</p><p><strong>Results: </strong>The FRACTURE sequence had the highest diagnostic performance, with an accuracy of 0.85, sensitivity of 0.85, specificity of 0.84, and AUC of 0.84. These values were 0.84, 0.72, 0.91, and 0.80, respectively, for the PDw sequence, and 0.83, 0.72, 0.91, and 0.79, respectively, for the T2wFS sequence. The AUC did not significantly differ between the FRACTURE and PDw sequences (Delong test, P > .05), but did significantly differ between the FRACTURE and T2wFS sequences (P < .05). For all osteoarthritis findings, the FRACTURE sequence had the highest kappa values and the highest sensitivity.</p><p><strong>Conclusions: </strong>FRACTURE sequencing may be a promising tool for the diagnosis of TMJ osteoarthritis compared with other conventional sequences.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"64-69"},"PeriodicalIF":2.9,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142738695","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Can temporomandibular joint osteoarthritis be diagnosed on MRI proton density-weighted images with diagnostic support from the latest deep learning classification models?","authors":"Michihito Nozawa, Motoki Fukuda, Shinya Kotaki, Marino Araragi, Hironori Akiyama, Yoshiko Ariji","doi":"10.1093/dmfr/twae040","DOIUrl":"10.1093/dmfr/twae040","url":null,"abstract":"<p><strong>Objectives: </strong>This study aimed to clarify the performance of MRI-based deep learning classification models in diagnosing temporomandibular joint osteoarthritis (TMJ-OA) and to compare the developed diagnostic assistance with human observers.</p><p><strong>Methods: </strong>The subjects were 118 patients who underwent MRI for examination of TMJ disorders. One hundred condyles with TMJ-OA and 100 condyles without TMJ-OA were enrolled. Deep learning was performed with 4 networks (ResNet18, EfficientNet b4, Inception v3, and GoogLeNet) using 5-fold cross validation. Receiver operating characteristics (ROC) curves were drawn for each model and diagnostic metrics were determined. The performances of the 4 network models were compared using Kruskal-Wallis tests and post hoc Scheffe tests, and ROCs between the best model and human were compared using chi-square tests, with P < .05 considered significant.</p><p><strong>Results: </strong>ResNet18 had areas under the curves (AUCs) of 0.91-0.93 and accuracy of 0.85-0.88, which were the highest among the 4 networks. There were significant differences in AUC and accuracy between ResNet and GoogLeNet (P = .0264 and.0418, respectively). The kappa values of the models were large, 0.95 for ResNet and 0.93 for EfficientNet. The experts achieved similar AUC and accuracy values to the ResNet metrics, 0.94 and 0.85, and 0.84 and 0.84, respectively, but with a lower kappa of 0.67. Those of the dental residents showed lower values. There were significant differences in AUCs between ResNet and residents (P < .0001) and between experts and residents (P < .0001).</p><p><strong>Conclusions: </strong>Using a deep learning model, high performance was confirmed for MRI diagnosis of TMJ-OA.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"56-63"},"PeriodicalIF":2.9,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141787490","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}