Dento maxillo facial radiology最新文献

{"title":"Artificial intelligence for reducing metal artifacts in dental CBCT images: a systematic review.","authors":"Parisa Soltani, Pedram Iranmanesh, Fateme Ebrahimzadeh, Francesca Angelone, Alfonso Maria Ponsiglione, Francesco Amato, Amirhossein Moaddabi, Niccolo Giuseppe Armogida, Gianrico Spagnuolo, Carlo Rengo, Hessamoddin Faghihian","doi":"10.1093/dmfr/twag034","DOIUrl":"10.1093/dmfr/twag034","url":null,"abstract":"<p><strong>Objectives: </strong>The presence of metallic restorations introduces severe artifacts that compromise the diagnostic accuracy of cone beam computed tomography (CBCT) images. This systematic review aims to evaluate the effectiveness of artificial intelligence (AI) techniques in reducing metal artifacts in dental CBCT images.</p><p><strong>Methods: </strong>A comprehensive literature search was conducted across 6 databases up until April 2025, using keywords related to AI and CBCT artifact reduction. Studies were selected based on the eligibility criteria, focusing on AI models trained on human CBCT scans targeting dental regions. Data extraction and risk of bias assessment (using the QUADAS-2 tool) were performed independently by several reviewers accordingly.</p><p><strong>Results: </strong>Ten studies published between 2019 and 2025 met the eligibility criteria. The majority of included studies had a high risk of bias. The most employed deep learning architectures were U-Net, GANs, and transformer-based models. Quantitative metrics like SSIM, PSNR, and RMSE demonstrated consistent improvements in image quality, while qualitative assessments confirmed superior artifact suppression and anatomical detail preservation compared to conventional methods. Transformer-based and physics-informed dual-domain networks performed superior to traditional image-only U-Nets in both quantitative fidelity and visual artifact suppression.</p><p><strong>Conclusions: </strong>AI-based approaches show promising potential in enhancing CBCT image fidelity by effectively mitigating metal-induced distortions. Despite encouraging results, the broader clinical adoption of such techniques requires standardized evaluation protocols, training models using open-access datasets, and further trials to validate diagnostic impact and generalizability.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"695-704"},"PeriodicalIF":4.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148041540","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":"Does the radiographic image format influence the detection of fractured endodontic instruments in gutta-percha-filled canals?","authors":"Maria Luiza Santos de Oliveira, Thaís Ocampo, Caio Alencar-Palha, Murilo Miranda-Viana, Deborah Queiroz Freitas, Francisco Haiter-Neto","doi":"10.1093/dmfr/twag042","DOIUrl":"10.1093/dmfr/twag042","url":null,"abstract":"<p><strong>Objective: </strong>To evaluate the influence of different image file formats on the accuracy of detecting fractured endodontic instruments in gutta-percha-filled root canals.</p><p><strong>Methods: </strong>Thirty-two mandibular molars were instrumented and divided into 2 groups: case (n = 15), which received a 2 mm fragment of a #35 K-file inside the mesial canal, and control (n = 17), without the fragment. Subsequently, all canals received gutta-percha, and each tooth was placed into a dry human mandible for acquiring radiographs. The radiographs were exported in TIFF, JPEG, BMP, PNG, and DICOM formats, resulting in 160 images (32 radiographs × 5 formats). Six radiologists evaluated the radiographs using a 5-point scale to determine the presence or absence of the fractured instrument. Area under the ROC curve, sensitivity, and specificity were calculated for each evaluator, along with a one-way ANOVA to compare the formats, adopting a 5% significance level.</p><p><strong>Results: </strong>The image file format did not significantly influence the accuracy (P = .767), sensitivity (P = .307), or specificity (P = .999) in detecting fractured endodontic instruments in filled canals.</p><p><strong>Conclusions: </strong>The study showed similar performance across the image file formats tested for detecting fractured instruments. These results suggest that commonly used and compatible formats can be used without compromising diagnostic accuracy.</p><p><strong>Advances in knowledge: </strong>Instrument fracture is one of the most challenging complications in endodontic treatments, and digital periapical radiographs are widely used in diagnosing this occurrence. The detection and localization of the fragment inside the canal are essential for a favorable prognosis.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"806-811"},"PeriodicalIF":4.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148162216","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":"SinusNet+: deep condition-label-free segmentation of maxillary sinus conditions in cone-beam CT images.","authors":"Da-El Kim, Su Yang, Ji Yong Han, Jo-Eun Kim, Kyung-Hoe Huh, Sam-Sun Lee, Min-Suk Heo, Won-Jin Yi","doi":"10.1093/dmfr/twag040","DOIUrl":"10.1093/dmfr/twag040","url":null,"abstract":"<p><strong>Objectives: </strong>Segmentation of maxillary sinus conditions (MSC) in cone-beam CT (CBCT) images may support preoperative assessment in the posterior maxilla, including implant planning and sinus floor augmentation. Supervised deep learning methods for MSC segmentation typically rely on labor-intensive manual annotation of MSC for network training. This study aimed to develop and evaluate a condition-label-free deep learning framework (SinusNet+) for MSC segmentation in CBCT images, in which network training does not require manual MSC annotations and instead relies on synthetic conditions generated within the normal maxillary sinus (MS).</p><p><strong>Methods: </strong>To generate synthetic MSC in normal MS, a synthetic condition generator was introduced to simulate MSC within the normal MS by varying texture, shape, and noise, thereby approximating a range of radiographic appearances of MSC in CBCT images.</p><p><strong>Results: </strong>SinusNet+ achieved an average Dice similarity coefficient of 0.820 ± 0.110, precision of 0.878±0.061, and recall of 0.777±0.145, respectively. The proposed method outperformed unsupervised baselines and achieved segmentation performance comparable to that of the supervised approach.</p><p><strong>Conclusions: </strong>The proposed framework demonstrates the feasibility of condition-label-free segmentation of MSC in CBCT images, while still requiring anatomical annotation of the normal MS during dataset preparation.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":"783-795"},"PeriodicalIF":4.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148224043","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":"CRANIOSEG: Deep Learning-Based Automated Segmentation of Craniomaxillofacial Anatomical Structures on Cone-Beam Computed Tomography.","authors":"İbrahim Şevki Bayrakdar, Alican Kuran, Mehmet Uğurlu, Hande Kazan, Kübra Altuntaş, Esin Kılıç, Ezgi Deniz Taştan, Hatice Şahiner, Orhan Kazan, Özer Çelik, Kaan Orhan, Antigoni Delantoni","doi":"10.1093/dmfr/twag065","DOIUrl":"https://doi.org/10.1093/dmfr/twag065","url":null,"abstract":"<p><strong>Objective: </strong>To develop and evaluate an nnU-Net v2-based deep-learning model for fully automated multiclass segmentation of 27 craniomaxillofacial anatomical structures on cone-beam computed tomography (CBCT).</p><p><strong>Methods: </strong>This retrospective study included CBCT scans from 106 adult patients. Twenty-seven anatomical structures were manually segmented in 3D Slicer to establish a consensus reference standard using a hierarchical annotation workflow. The dataset was divided at the patient level into training (n = 96) and held-out internal test (n = 10) sets. A three-dimensional full-resolution nnU-Net v2 model was trained and evaluated using the Dice similarity coefficient (DSC), Jaccard index (IoU), precision, recall, the 95th percentile Hausdorff distance (HD95), and the average symmetric surface distance (ASSD).</p><p><strong>Results: </strong>Across all 27 anatomical structures, the model achieved a mean DSC of 0.850 ± 0.132, a mean IoU of 0.773 ± 0.159, a mean precision of 0.863, a mean recall of 0.868, and a mean HD95 of 2.34 mm. The highest category-level performance was observed for the paranasal sinuses (DSC 0.961) and the craniofacial osseous framework (DSC 0.961), whereas the skull-base fissures and fossae showed the lowest performance (DSC 0.694), primarily because of the difficulty in segmenting the pterygopalatine fossa (DSC 0.377). Large, well-defined structures, including the mandible (DSC 0.976) and maxillary sinus (DSC 0.977), were segmented with excellent accuracy, whereas smaller, low-contrast structures such as the mandibular canal (DSC 0.654) remained more challenging.</p><p><strong>Conclusions: </strong>This study demonstrates the technical feasibility of simultaneously segmenting 27 craniomaxillofacial anatomical structures from CBCT images using a single nnU-Net v2 model on a selected internal dataset. Overall, the model achieved strong segmentation performance, with particularly high overlap values for large, well-defined structures such as the mandible, maxillary sinus, sphenoid sinus, and upper skull. Performance varied among structures, with lower values observed mainly for several small, low-contrast, and morphologically complex structures.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148976045","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":"Technical Report: Creation of a Cartesian coordinate system within the mandibular ramus for precise analysis of its morphology.","authors":"J Matschke, H Leonhardt, A Franke, Michaela Buckova, Fernando Guastaldi, Kristin Paetzold-Byhain, Philipp Sembdner","doi":"10.1093/dmfr/twag064","DOIUrl":"https://doi.org/10.1093/dmfr/twag064","url":null,"abstract":"<p><p>The purpose was to create a Cartesian coordinate system (CCS) within the mandibular ramus relating to specific anatomical landmarks, which can then be used to analyse the mandibular ramus further. Therefore, STL files of mandibular rami were imported into the specifically created program CTinA. Anatomical landmarks, such as the Loukota line and the posterior edge of the mandibular ramus, were marked by hand. Afterwards, a CCS was created, which is related to the marked structures. This means, a reproducible CCS for the mandibular ramus and condyle was established referencing anatomical landmarks. In conclusion, the introduction of the CCS enables standardised clinically interpretable analysis of morphology and may support implant design, such as osteosynthesis plates made from magnesium alloys. Compared to the best-fit alignment, it relies on anatomical landmarks and may be more robust in cases of morphological variation and remodelling.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148911114","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":"Periodontitis and carotid calcifications on panoramic radiographs in relation to carotid atherosclerosis on ultrasound.","authors":"Magnus Bladh, Nils Gustafsson, Gunnar Engström, Cecilia Kennbäck, Daniel Jönsson, Eva Levring Jäghagen","doi":"10.1093/dmfr/twag063","DOIUrl":"https://doi.org/10.1093/dmfr/twag063","url":null,"abstract":"<p><strong>Objectives: </strong>To investigate periodontitis combined with carotid artery calcifications on panoramic radiographs in relation to carotid ultrasound findings of atherosclerotic disease among participants in the Malmö Offspring Dental Study.</p><p><strong>Methods: </strong>In this population-based, cross-sectional cohort study, 410 participants (mean age 53.0±7.8 years) were examined by carotid ultrasound, panoramic and intraoral radiography. The panoramic radiographs were assessed for presence and shape of carotid artery calcifications and marginal bone loss, indicating current or past periodontitis. Groups with and without periodontitis stage III-IV and/or carotid artery calcifications were analyzed for associations to findings of three carotid ultrasound characteristics: total plaque area, number of plaques, and lumen reduction. Adjustments for covariates were performed.</p><p><strong>Results: </strong>Participants with combined periodontitis and carotid artery calcifications (all shapes) had significantly greater total plaque area, number of plaques, and lumen reduction (P < 0.001). The findings for vessel-outlining shape combined with periodontitis was more severe but non-significant. Participants with combined findings had 13.7-times higher probability of having plaques on ultrasound, whereas those with carotid artery calcifications alone had 7.65-times higher probability than participants with no carotid artery calcifications or periodontitis (P < 0.001).</p><p><strong>Conclusions: </strong>Participants with periodontitis and carotid artery calcifications had more ultrasound findings, indicating more advanced atherosclerotic disease. Dentists should consider recommending patients with these findings to seek medical attention for cardiovascular risk factor evaluation and prevention.</p><p><strong>Advances in knowledge: </strong>This study explore that participants with periodontitis combined with carotid artery calcification on panoramic radiographs have more advanced signs of atherosclerotic disease on ultrasound than participants with carotid artery calcifications but without periodontitis.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148863987","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 Differential Diagnosis of Odontogenic Lesions in Cone Beam Computed Tomography Images Using Radiomics-Based Machine Learning.","authors":"Zarif Ece Hammudıoglu, Merve Kucuk Kurtgoz, Elif Çeçen Erol, Ceren Aktuna Belgın, Kaan Orhan","doi":"10.1093/dmfr/twag062","DOIUrl":"https://doi.org/10.1093/dmfr/twag062","url":null,"abstract":"<p><strong>Objective: </strong>This study aimed to evaluate the structural characteristics of mandibular alveolar bone in patients with Type 1 diabetes mellitus (T1DM), Type 2 diabetes mellitus (T2DM), and systemically healthy controls using panoramic radiography-based radiomic analysis combined with machine learning algorithms.</p><p><strong>Materials and methods: </strong>A total of 225 panoramic radiographs (75 T1DM, 75 T2DM, 75 healthy controls) were retrospectively analyzed. ROIs were segmented from eight anatomical mandibular segments per subject, and 107 radiomic features were extracted using PyRadiomics. Interobserver reliability was confirmed by two-way random-effects ICC (≥0.85). A leakage-free pipeline was applied. Four machine learning algorithms were evaluated: Random Forest, ExtraTrees, SVM-RBF, and Logistic Regression.</p><p><strong>Results: </strong>Significant differences were identified in age (Kruskal-Wallis p = 0.0003) and sex (χ²=17.857, p = 0.0001). The best single-segment performance was achieved in the left mandibular corpus with Logistic Regression (Accuracy=0.833, F1=0.832, AUC=0.958). All segments showed significant radiomic differences (FDR q < 0.001). Sensitivity of 1.000 was achieved for T1DM and AUC=1.000 for T2DM. The feature glszm_SizeZoneNonUniformityNormalized showed the strongest discriminative power (H = 86.928, ε²=0.578).</p><p><strong>Conclusion: </strong>Panoramic radiography-based radiomic analysis demonstrates high diagnostic performance in non-invasively distinguishing mandibular bone alterations among T1DM, T2DM, and healthy individuals, with potential as a clinical bone monitoring tool.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148788890","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":"Comparative Evaluation of MRI and Ultrasonography for Temporomandibular Joint Pathologies with Emphasis on Bone Marrow Edema.","authors":"İrem Öztürk, Kaan Orhan","doi":"10.1093/dmfr/twag059","DOIUrl":"https://doi.org/10.1093/dmfr/twag059","url":null,"abstract":"<p><strong>Objectives: </strong>This retrospective diagnostic accuracy study compared magnetic resonance imaging and ultrasonography in 98 temporomandibular joints from 49 patients for evaluating disc position, disc configuration, joint effusion, condylar degenerative changes, and bone marrow edema.</p><p><strong>Methods: </strong>A total of 98 temporomandibular joints from 49 patients who underwent both modalities were retrospectively evaluated by two independent, blinded observers using established classification systems. All imaging was performed prior to the initiation of any specific TMJ treatment. Intra- and inter-observer agreement, the diagnostic accuracy of ultrasonography relative to the magnetic resonance imaging reference standard, diagnostic likelihood ratios, and associations between parameters were analysed statistically.</p><p><strong>Results: </strong>Intra- and inter-observer agreement was good to excellent for both modalities (Kappa 0.71-0.91, p < 0.05). Ultrasonography showed high sensitivity and specificity for disc displacement (94.4-97.4% and 61.5-70.0%) and effusion (87.5-92.3% and 84.0-87.0%), but more limited performance for bone marrow edema, particularly in specificity (60.0-61.1%). On the left side, magnetic resonance imaging identified bone marrow edema in 63.3% of cases compared with 36.7% on ultrasonography (p = 0.015). Folded and convex disc configurations and marked or severe effusion were both associated with displacement without reduction (p ≤ 0.046).</p><p><strong>Conclusions: </strong>Ultrasonography appears suitable as a first-line screening modality for selected TMJ abnormalities, whereas magnetic resonance imaging remains indispensable whenever bone marrow pathology or complex internal derangement is suspected.</p><p><strong>Advances in knowledge: </strong>This study provides direct, same-patient diagnostic accuracy data comparing ultrasonography with magnetic resonance imaging across multiple temporomandibular joint pathologies, including an indirect ultrasonographic assessment of bone marrow edema via cortical irregularity, clarifying which findings ultrasonography can reliably screen for and which still require magnetic resonance imaging confirmation.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148788775","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":"Information matters, but cues are dangerous: a comparative evaluation of three multimodal AI chatbots in oral and maxillofacial radiology.","authors":"Chang-Ki Min","doi":"10.1093/dmfr/twag061","DOIUrl":"https://doi.org/10.1093/dmfr/twag061","url":null,"abstract":"<p><strong>Objectives: </strong>To compare the diagnostic performance of three multimodal AI chatbots on oral and maxillofacial radiographic images and to examine how additional information, delivery mode, and user-suggested diagnoses affect accuracy.</p><p><strong>Methods: </strong>Three AI chatbots (GPT-5.1, Gemini 3 Flash, Claude Opus 4.7) were tested on 90 cases comprising normal controls, osteomyelitis, and benign jaw lesions. Inputs were combinations of a panoramic image, a cropped panoramic image, an axial CBCT image or a text-based cue. Inputs were delivered all at once or sequentially. Accuracy was scored at category and specific-diagnosis levels using non-parametric tests with false-discovery-rate correction.</p><p><strong>Results: </strong>With the panoramic image alone, accuracy for diseased cases was low (0-60%) but rose to as high as 38-92% in each model's best condition with added information. A cropped image was the most consistently beneficial additional visual input, whereas an axial CBCT image provided less improvement. GPT-5.1 recognised normal cases well but missed lesions, Gemini 3 was sensitive but less specific, and Claude 4.7 defaulted to benign diagnoses. Correct verbal cues increased accuracy, whereas a misleading cue caused decline. Gemini 3 accepted a false benign suggestion in 96% of cases it had initially classified as normal. For benign lesions, specific-diagnosis accuracy was almost half of category-level accuracy.</p><p><strong>Conclusions: </strong>Accuracy was strongly affected by model choice, information type and user-suggested diagnoses, but no model reached a level that would be acceptable for clinical use.</p><p><strong>Advances in knowledge: </strong>This multi-model comparison isolates the effects of information type, delivery mode, and sycophancy in oral and maxillofacial radiology.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148788903","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":"Multimodal Large Language Models and Dental Students in Radiographic Landmark Identification: A Novel Grid-Based Assessment.","authors":"Mehmet Egemen Aydemir, Burcu Sayin, Burcu Yeliz Kollayan, Can Günel, Kadir Cem, Mehmet Ali Gül","doi":"10.1093/dmfr/twag060","DOIUrl":"https://doi.org/10.1093/dmfr/twag060","url":null,"abstract":"<p><strong>Objectives: </strong>To introduce a novel grid-based coordinate framework for evaluating spatial landmark localisation competence in multimodal large language models (MLLMs) applied to dental radiography, and to benchmark two MLLMs against dental student consensus.</p><p><strong>Methods: </strong>GPT-5.4 and Gemini 3.1 Pro were evaluated on 900 queries spanning 200 dental radiographs (100 panoramic, 50 periapical, 50 cephalometric) and 12 landmarks (9 point, 3 area) under zero-shot and guided prompting across three repetitions (10,800 API calls). Performance was scored against a two-rater adjudicated ground truth and a team-adjudicated fourth-year dental student consensus (n = 40). Metrics included Euclidean distance, the successful detection rate, and the Jaccard index.</p><p><strong>Results: </strong>GPT-5.4 matched student-level performance on cephalometric landmarks but was substantially outperformed on periapical and panoramic tasks. Gemini 3.1 Pro outperformed GPT-5.4 on panoramic and periapical landmarks, while GPT-5.4 retained an advantage on cephalometric points. Guided prompting produced heterogeneous effects, improving localisation of some landmarks while causing severe regression on others; most notably, GPT-5.4 directed the majority of lower-left canine apex predictions to the incorrect lower-right molar region, a prompt-resistant mislocalisation pattern absent in Gemini 3.1 Pro. Both models remained inferior to the student consensus on all periapical and panoramic comparisons, with Gemini 3.1 Pro approaching parity only on cephalometric points under guided prompting.</p><p><strong>Conclusion: </strong>Current MLLMs demonstrate modality-specific spatial competence insufficient for autonomous clinical deployment. The grid-based framework provides a reproducible, modality-agnostic benchmark for tracking MLLM spatial reasoning across future model generations.</p>","PeriodicalId":11261,"journal":{"name":"Dento maxillo facial radiology","volume":" ","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148788913","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}