Diagnostic and interventional radiology最新文献

{"title":"Radiologic patterns of interstitial lung disease: a pictorial review based on the 2025 European Respiratory Society/American Thoracic Society Classification.","authors":"Furkan Ufuk","doi":"10.4274/dir.2026.263751","DOIUrl":"10.4274/dir.2026.263751","url":null,"abstract":"<p><p>Interstitial lung diseases (ILDs) are a heterogeneous group of diffuse parenchymal disorders with overlapping clinical and physiological features, making imaging central to diagnosis and management. Thin-slice computed tomography (CT) enables recognition of characteristic patterns and disease monitoring, serving as a cornerstone of multidisciplinary discussion. The 2025 European Respiratory Society/American Thoracic Society classification introduces refinements to the 2013 framework, grouping ILDs into three categories: interstitial patterns, alveolar filling disorders, and rare or unclassified disorders. This pictorial essay illustrates the updated classification with representative CT examples.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":"643-654"},"PeriodicalIF":2.7,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13530495/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147303501","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Single-center retrospective comparative study of 40-120 μm versus 100-300 μm trisacryl gelatin microspheres as the main embolic agents for prostatic artery embolization.","authors":"Hippocrates Moschouris, Çağın Şentürk, Konstantinos Stamatiou, Apostolos Mouroukis, Chrysovalantis Stylianou, Dimitrios Kontothanasis","doi":"10.4274/dir.2026.263602","DOIUrl":"10.4274/dir.2026.263602","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;To compare the safety and efficacy of trisacryl gelatin microspheres (TGM) with diameters of 40-120 μm (TGM 40-120) vs. 100-300 μm (TGM 100-300) as the primary embolic agents for prostatic artery (PA) embolization (PAE).&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;This single-center retrospective comparative study included patients with symptomatic benign prostatic hyperplasia treated with PAE using TGM 40-120 (group A) or TGM 100-300 (group B) over a two-year period. Evaluation included International Prostate Symptom Score (IPSS), quality of life score (QoL), prostatic volume (PV), post-void residual (PVR), and percentage of prostatic infarction (pPI) measured by contrast-enhanced ultrasound. Clinical success was defined as post-PAE reduction of IPSS ≤ 15 points with a decrease of at least 25% from the baseline, QoL score ≤ 3 points or a decrease of at least 1 point from baseline, and successful bladder catheter removal (for catheter-dependent patients). Adverse events were recorded and graded using the modified CIRSE classification system for complication reporting.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Most baseline characteristics did not differ significantly between group A (n = 21) and group B (n = 31). At 3 months post-PAE, there were no statistically significant differences between groups in improvement of IPSS (group A: 49.5 ± 26.1%, group B: 51.8 ± 23.3%, &lt;i&gt;P&lt;/i&gt; = 0.403) or QoL (group A: 54.7 ±3 1.6%, group B: 51.2 ± 14.1%, &lt;i&gt;P&lt;/i&gt; = 0.589), nor in reduction of PV (group A: 35.5 ± 16.9%, group B: 28.4 ± 11.0%, &lt;i&gt;P&lt;/i&gt; = 0.104) or PVR (group A: 59.8 ± 23.9%, group B: 60.4 ± 28.5%, &lt;i&gt;P&lt;/i&gt; = 0.934). However, pPI (calculated 1-4 days post-PAE) was significantly higher in group A (48.3 ± 16.2% vs. 30.3 ± 11.9%, &lt;i&gt;P&lt;/i&gt; = 0.007). Mid-term PV reduction was also significantly greater in group A (31.4 ± 21.4% vs. 12.2 ± 10.2%, &lt;i&gt;P&lt;/i&gt; = 0.030). There were more complications in group A (7/21 vs. 4/31, &lt;i&gt;P&lt;/i&gt; = 0.095). Three complications in group A were grade 3a events (bladder wall ischemia, n = 2; persistent severe anal and perineal pain requiring hospitalization and opioids, n = 1). No grade 3 events were observed in group B (&lt;i&gt;P&lt;/i&gt; = 0.060). Clinical success rates at 6, 12, 24, and 36 months post-PAE were 76.2%, 76.2%, 71.1%, and 71.1%, respectively, for group A, and 87.1%, 83.9%, 79.9%, and 79.9%, respectively, for group B (&lt;i&gt;P&lt;/i&gt; = 0.451).&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusion: &lt;/strong&gt;Compared with TGM 100-300, PAE with TGM 40-120 appears to provide similar clinical benefit but results in significantly higher pPI and greater mid-term prostate shrinkage. Ischemic complications may be more frequent and more severe with the smaller microspheres.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Clinical significance: &lt;/strong&gt;This is the first systematic comparison between the smallest and the second smallest commercially available TGM sizes in the clinical context of PAE. These findings may help guide ongoing research to identify the most appropriate embolic mater","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":"700-708"},"PeriodicalIF":2.7,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13530443/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147270089","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Ultrasound-guided microwave ablation for breast tumors: current status and future perspectives.","authors":"Jin Wang, Hongjuan Wei, Liming Zhou, Xin Chang, Jing Zhao, Hui Chen","doi":"10.4274/dir.2026.263639","DOIUrl":"10.4274/dir.2026.263639","url":null,"abstract":"<p><p>Ultrasound (US)-guided microwave ablation (MWA) has emerged as a promising minimally invasive therapy for both benign and malignant breast tumors. This review comprehensively examines the current clinical status, technical principles, and therapeutic outcomes of US-guided MWA in breast tumor management. We discuss the biophysical mechanisms of MWA, its advantages over other ablation techniques-such as rapid temperature elevation, the ability to create more extensive coagulation areas, and diminished impact from heat sink phenomena-and the critical role of real-time US guidance in enhancing procedural precision and safety. Clinical evidence supports the efficacy of US-guided MWA in achieving high rates of complete ablation and significant volume reduction for benign tumors, such as fibroadenomas, with minimal complications and excellent cosmetic results. For early-stage breast cancers, initial studies indicate that US-guided MWA provides local tumor control comparable with surgical resection in the short- to mid-term, while also offering the benefits of shorter operation times, reduced hospitalization, and stimulation of systemic antitumor immune responses. However, challenges remain, including technical limitations in treating tumors near critical structures, the lack of long-term oncological data, and operator dependence. Future directions involve technological refinements, integration with artificial intelligence and advanced imaging, combination with immunotherapy, and standardization of protocols. US-guided MWA represents an important advancement toward personalized, organ-preserving breast tumor therapy, with ongoing innovations poised to expand its clinical applicability.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":"675-691"},"PeriodicalIF":2.7,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13530549/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146112594","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Reply: comments on the diagnostic value of ADC texture analysis in PI-RADS 5 lesions.","authors":"Yan Bai, Chen Jiang Wu","doi":"10.4274/dir.2026.263881","DOIUrl":"10.4274/dir.2026.263881","url":null,"abstract":"","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":"533-534"},"PeriodicalIF":2.7,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13530622/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146112634","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Keros-Net: a convolutional block attention module-squeeze-and-excitation-integrated hybrid learning framework for olfactory fossa depth classification.","authors":"Zülküf Akdemir, Murat Canayaz, Abdulaziz Yalınkılıç, Suat Orak","doi":"10.4274/dir.2026.263749","DOIUrl":"10.4274/dir.2026.263749","url":null,"abstract":"<p><strong>Purpose: </strong>This study aimed to develop a hybrid decision support framework combining deep learning (DL) and machine learning (ML) to automatically classify olfactory fossa depth on paranasal computed tomography (CT) images according to the Keros classification. The goal was to enhance accuracy, reduce observer variability, and support safer endoscopic sinus surgery.</p><p><strong>Methods: </strong>A retrospective dataset of 481 individuals (1,549 cropped coronal CT slices) was analyzed and labeled as Keros types I-III by an experienced radiologist. Deep features were extracted using DenseNet121, DenseNet169, and DenseNet201 architectures enhanced with convolutional block attention modules (CBAMs) and squeeze-and-excitation (SE) blocks. Four feature selection techniques-SHapley Additive exPlanations, recursive feature elimination (RFE), principal component analysis, and SelectKBest-were applied to reduce dimensionality. Selected features were classified using support vector machines (SVMs), random forest, XGBoost, logistic regression, and Naive Bayes. Five-fold cross-validation was used to assess accuracy, precision, recall, and F1-score metrics.</p><p><strong>Results: </strong>Among the baseline models, DenseNet169 achieved the highest accuracy (88.37%). After feature selection, the RFE + SVM and RFE + logistic regression combinations yielded the best performance with an accuracy of 97.90%, demonstrating a substantial improvement over DL models alone. The most effective feature selection technique was RFE, and SVMs consistently produced well-balanced classification results.</p><p><strong>Conclusion: </strong>Integrating CBAM-SE-enhanced DenseNet architectures with optimized feature selection and classic ML classifiers enables highly accurate and reliable automatic classification of Keros types. The proposed hybrid approach outperforms conventional DL models and provides a robust framework for objective radiological assessment.</p><p><strong>Clinical significance: </strong>Accurate preoperative identification of olfactory fossa depth is essential for preventing complications such as cribriform plate injury and cerebrospinal fluid leakage during endoscopic sinus surgery. The proposed system offers an efficient, reproducible, and objective tool that may enhance surgical planning, reduce operator dependency, and increase patient safety.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":"600-607"},"PeriodicalIF":2.7,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13530439/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147472989","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Association of cardiac magnetic resonance imaging parameters with N-terminal pro-B-type natriuretic peptide levels.","authors":"İffet Doğan, Kemal Buğra Memis, Defne Şahin, Önder Durmaz, Serhat Hayme, Mecit Kantarcı","doi":"10.4274/dir.2026.264340","DOIUrl":"https://doi.org/10.4274/dir.2026.264340","url":null,"abstract":"<p><strong>Purpose: </strong>This study aimed to evaluate the associations between serum N-terminal pro-B-type natriuretic peptide (NT-proBNP) levels and conventional and tissue-characterization cardiac magnetic resonance (CMR) imaging parameters.</p><p><strong>Methods: </strong>This retrospective, single-center study included 204 patients who underwent CMR and had NT-proBNP measured within 7 days before or after the examination. Patients were categorized into three groups based on NT-proBNP levels: normal (< 125 pg/mL), high (125-500 pg/mL), and extremely high (> 500 pg/mL). A comparison of CMR-derived structural and functional parameters, including left ventricular ejection fraction, myocardial mass, left atrial (LA) diameter, late gadolinium enhancement (LGE), extracellular volume (ECV), and short tau inversion recovery (STIR)-based T2 signal-intensity (SI) ratio, was performed across groups. Associations were assessed using ordinal and multinomial logistic regression analyses.</p><p><strong>Results: </strong>Higher NT-proBNP levels were associated with increased myocardial mass, wall thickness, LA enlargement, and reduced systolic function. The LGE burden, LA cavity diameter, and ECV increased significantly across NT-proBNP groups. In adjusted multinomial logistic regression, the ECV and LA cavity diameter were independently associated with both the high and extremely high NT-proBNP groups. The STIR-based T2 SI ratio demonstrated an inverse association with NT-proBNP; however, an independent association was observed only in the extremely high NT-proBNP group (adjusted odds ratio: 0.80, <i>P</i> = 0.034), and this finding should be considered exploratory.</p><p><strong>Conclusion: </strong>Higher NT-proBNP categories were associated with greater structural abnormalities and tissue-characterization changes on CMR, particularly a higher ECV and larger LA diameter. The inverse association observed with the STIR-based T2 SI ratio was limited to the extremely high NT-proBNP group and should be considered exploratory.</p><p><strong>Clinical significance: </strong>Elevated NT-proBNP levels were associated with CMR findings consistent with myocardial remodeling, particularly increased ECV expansion.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":2.7,"publicationDate":"2026-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148817817","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Explainable artificial intelligence in medical imaging: how to interpret, evaluate, and use artificial intelligence explanations.","authors":"Gorkem Durak, Halil Ertugrul Aktas, Tugba Akinci D'Antonoli, Ulas Bagci","doi":"10.4274/dir.2026.264319","DOIUrl":"https://doi.org/10.4274/dir.2026.264319","url":null,"abstract":"<p><p>Most artificial intelligence (AI) models used in radiology are black boxes-they produce predictions without explaining the basis of their outputs, raising concerns about clinical safety, accountability, and trust. To address this, a growing body of methods has been developed to help clinicians understand and evaluate AI predictions. This field, known as explainable AI (XAI), aims to help clinicians interrogate, interpret, and critically evaluate AI predictions by identifying factors associated with model outputs. In this educational and practical review, we provide an accessible overview of XAI tailored for practicing radiologists and physicians. We cover the major categories of explanation methods, including saliency maps, perturbation-based and feature-attribution approaches, concept- based methods, and example-based reasoning, as well as uncertainty quantification as a complementary approach for assessing prediction reliability, along with common misconceptions and emerging regulatory obligations. We aim to make XAI easier for healthcare professionals to understand, as effective oversight of AI tools has become a core competency for the modern radiologist.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":2.7,"publicationDate":"2026-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148789490","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"3D Slicer-based preprocedural trajectory planning for transvenous mediastinal biopsy using automated segmentation: a technical note.","authors":"Yohei Takeshita, Kanako Furukawa, Yo Matsui, Junji Moriya","doi":"10.4274/dir.2026.263983","DOIUrl":"https://doi.org/10.4274/dir.2026.263983","url":null,"abstract":"<p><p>Histological diagnosis of mediastinal tumors can be challenging when safe puncture access is limited by adjacent airway and vascular structures. This technical note describes a 3D Slicer software-based preprocedural planning workflow for transvenous mediastinal biopsy using automated segmentation of normal anatomical structures. In a woman in her 60s with a large mediastinal tumor and superior vena cava (SVC) syndrome, a transpulmonary approach was considered undesirable because of the risk of pneumothorax and hemorrhage, and a transvenous approach was advantageous because biopsy and subsequent endovascular management of SVC syndrome could be incorporated into the same strategy. Contrast-enhanced computed tomography data were imported into 3D Slicer, and critical organs were segmented using artificial intelligence-based tools, and the tumor was manually segmented. With a fixed entry point in the SVC, six candidate puncture trajectories were generated and compared using three-dimensional distance-based metrics. A composite risk score was used to rank the trajectories as a heuristic within-case decision-support tool rather than a validated predictor. The workflow did not automatically determine the puncture direction but provided practical decision support by visualizing and avoiding high-risk trajectories. Transvenous biopsy was successfully performed under biplane fluoroscopic guidance using a transjugular liver biopsy kit without procedural complications.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":2.7,"publicationDate":"2026-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148789434","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Advancements in magnetic resonance imaging and magnetic resonance-driven techniques for evaluating pituitary macroadenoma consistency: current progress and future perspectives.","authors":"Chenxin Tian, Qinglei Shi, Xinzhi Zhao, Shijia Wang, Junlong Qin, Xuanxun Guo, Ming Feng, Ge Chen, Xiaohai Liu, Renzhi Wang","doi":"10.4274/dir.2026.263937","DOIUrl":"10.4274/dir.2026.263937","url":null,"abstract":"<p><p>Pituitary macroadenomas (PMAs) are among the most common intracranial tumors and pose significant surgical challenges, especially when the tumor consistency is increased due to high fibrous content. Accurate preoperative assessment of tumor consistency is crucial for optimizing surgical strategies and patient outcomes. Although conventional T2-weighted magnetic resonance imaging (MRI) is widely employed, its predictive value remains uncertain. Emerging techniques, such as diffusion-weighted imaging and magnetic resonance elastography, provide insights into tumor stiffness by assessing microstructural and mechanical properties, enhancing the prediction of PMA consistency. Furthermore, advancements in machine learning (ML) and deep learning, particularly convolutional neural networks (CNNs) and hybrid CNN-transformer models, have improved the extraction of complex imaging features, leading to greater predictive accuracy. This review summarizes recent developments in MRI and MR-driven imaging techniques for predicting PMA consistency, emphasizing their clinical application and the role of ML-based methods in refining predictive models. Despite numerous candidate imaging biomarkers, the field remains constrained by insufficient reproducibility, limited cross-study comparability, heterogeneity in MRI acquisition and post-processing protocols, and poor model generalizability. Continued integration of advanced MRI with ML-driven imaging analyses may further enhance preoperative prediction of PMA consistency and facilitate surgical planning.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":2.7,"publicationDate":"2026-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148720456","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Diagnostic performance and short-interval reproducibility of multimodal large language models in differentiating cholesteatoma from chronic otitis media using key-image temporal bone high-resolution computed tomography.","authors":"Buket Yağcı, Sergen Palaz, Erdem Atalay Çetinkaya, Utku Şenol","doi":"10.4274/dir.2026.264082","DOIUrl":"https://doi.org/10.4274/dir.2026.264082","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;This study aimed to assess whether multimodal large language models (LLMs) can distinguish cholesteatoma from non-cholesteatomatous chronic otitis media (COM) on representative key-image temporal bone high-resolution computed tomography (HRCT) and to evaluate the short-interval reproducibility of their outputs.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;This retrospective, single-center study (2019-2024) included 101 patients (48 with cholesteatoma, 53 with non-cholesteatomatous COM) who underwent surgical treatment. The reference standard was intraoperative diagnosis with histopathological confirmation for cholesteatoma and surgical documentation for COM. For each case, six anonymized representative HRCT images reflecting standard diagnostic criteria were selected by consensus between a 4&lt;sup&gt;th&lt;/sup&gt;-year radiology resident and a board-certified head and neck radiologist, both unaware of the diagnosis. Subsequently, the same images were analyzed by the GPT-5 and Gemini 2.5 Pro LLMs through their official web interfaces, utilizing structured prompts and a zero-shot approach. These evaluations were conducted in two distinct sessions (S1 and S2) with a 1-week interval. The primary endpoint was accurate binary classification. Accuracy, sensitivity, specificity, positive predictive value, and negative predictive value were calculated with 95% confidence intervals [(CIs); Wilson method] agreement with the reference standard and between sessions and models was assessed with the Cohen kappa (κ) coefficient; and differences in classification were assessed with the McNemar test.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;The radiologist achieved an accuracy of 96.0% (95% CI: 90.3-98.4) with almost perfect agreement with the reference standard (κ: 0.921). In S1, GPT-5 and Gemini 2.5 Pro achieved accuracies of 43.6% and 49.5%, and in S2, 46.5% and 47.5%, respectively. Both models combined high sensitivity (83.3%-97.9%) with low specificity (1.9%-13.2%), and balanced accuracy ranged from 0.45 to 0.52. Between-session reproducibility was fair for GPT-5 (κ: 0.360) and moderate for Gemini 2.5 Pro (κ: 0.485), and inter-model agreement was slight at both sessions (κ: 0.035 at S1 and κ: 0.086 at S2). Accuracy did not differ significantly between the two models (&lt;i&gt;P&lt;/i&gt; = 0.211).&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusion: &lt;/strong&gt;In this single-center study, GPT-5 and Gemini 2.5 Pro, in the versions evaluated, combined high sensitivity with low specificity and showed only fair-to-moderate between-session reproducibility and exhibited slight inter-model agreement on temporal bone key-image HRCT. These findings do not support their use as independent second readers, and broader generalization to other multimodal LLMs would require the evaluation of additional models.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Clinical significance: &lt;/strong&gt;The evaluated models lacked the spatial precision and consistency required for the accurate assessment of complex middle ear structures. This finding underscores th","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":2.7,"publicationDate":"2026-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148705806","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}