Diagnostic and interventional radiology最新文献

{"title":"A diagnostic model based on magnetic resonance imaging for Menière's disease: a multicentre study.","authors":"Xinyi Chen, Yanfeng Zhao, Yunchong Han, Kai Wei, Shufang Cheng, Yongjun Ye, Jie Feng, Xinchen Huang, Jingjing Xu","doi":"10.4274/dir.2025.253293","DOIUrl":"https://doi.org/10.4274/dir.2025.253293","url":null,"abstract":"<p><strong>Purpose: </strong>To evaluate the diagnostic performance of delayed post-gadolinium enhancement magnetic resonance imaging (DEMRI) in diagnosing Menière's disease (MD) and to establish an effective MRI-based diagnostic model.</p><p><strong>Methods: </strong>This retrospective multicenter study assessed DEMRI descriptors in patients presenting with Ménièriform symptoms who were examined consecutively between May 2022 and May 2024. A total of 162 ears (95 with MD, 67 controls) were included. Each ear was randomly assigned to either a training set (n = 98) or a validation set (n = 64). In the training cohort, diagnostic models for MD were developed using logistic regression. The area under the curve (AUC) was used to evaluate the diagnostic performance of the different models. The Delong test was applied to compare AUC estimates between models.</p><p><strong>Results: </strong>The proposed DEMRI diagnostic model demonstrated strong diagnostic performance in both the training cohort (AUC: 0.907) and the validation cohort (AUC: 0.887), outperforming the clinical diagnostic model (<i>P</i> = 0.01231; 95% confidence interval: 0.033-0.269) in the validation cohort. The AUC of the DEMRI model was also higher than that of the combined DEMRI-clinical model (AUC: 0.796), although the difference was not statistically significant (<i>P</i> = 0.054). In the training set, the sensitivity and specificity of the DEMRI model were 78.9% and 88.5%, respectively.</p><p><strong>Conclusion: </strong>A diagnostic model based on DEMRI features for MD is more effective than one based solely on clinical variables. DEMRI should, therefore, be recommended when MD is suspected, given its significant diagnostic potential.</p><p><strong>Clinical significance: </strong>This model may improve the accuracy and timeliness of MD diagnosis, as it is less influenced by the attending physician's level of inquiry or the patient's self-reporting ability. It may also contribute to more effective disease management in patients with MD.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144173051","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":"The AI Act: responsibilities and obligations for healthcare professionals and organizations.","authors":"Kicky Gerhilde van Leeuwen, Leon Doorn, Erik Gelderblom","doi":"10.4274/dir.2025.252851","DOIUrl":"https://doi.org/10.4274/dir.2025.252851","url":null,"abstract":"","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2025-05-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144173055","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":"Reply: evaluating text and visual diagnostic capabilities of large language models on questions related to the Breast Imaging Reporting and Data System (BI-RADS) Atlas 5^th edition.","authors":"Yasin Celal Güneş, Turay Cesur, Eren Çamur, Leman Günbey Karabekmez","doi":"10.4274/dir.2025.253360","DOIUrl":"https://doi.org/10.4274/dir.2025.253360","url":null,"abstract":"","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144110124","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":"Retrieval-augmented generation for answering Breast Imaging Reporting and Data System (BI-RADS)-related questions with large language models.","authors":"Esat Kaba","doi":"10.4274/dir.2025.253272","DOIUrl":"https://doi.org/10.4274/dir.2025.253272","url":null,"abstract":"","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144110128","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":"Amide proton transfer-weighted magnetic resonance imaging for the evaluation of testicular spermatogenic function: a preliminary study.","authors":"Guanglei Tang, Shulin Ma, Wenhao Fu, Weijian Yun, Yang Peng, Jian Guan","doi":"10.4274/dir.2025.253248","DOIUrl":"https://doi.org/10.4274/dir.2025.253248","url":null,"abstract":"<p><strong>Purpose: </strong>To determine the amide proton transfer-weighted (APTw) imaging features in testes with age, and to assess the feasibility of APTw magnetic resonance imaging (MRI) in assessing testicular spermatogenic function.</p><p><strong>Methods: </strong>A total of 23 male patients with clinically confirmed hypospermatogenesis caused by epididymo-orchitis were included in the case group (group A) and another 93 men (age range, 20-80 years) were included in the control group. The control group was divided into four subgroups: group B1 (20-34 years, n = 25), group B2 (35-49 years, n = 23), group B3 (50-64 years, n = 21), and group B4 (65-80 years, n = 24). All participants underwent 3.0T MRI scan, and the APT signal intensity (SI) and apparent diffusion coefficient (ADC) value of each testis were examined. The ADC and APT SI were independently measured by two radiologists blinded to clinical data, and average values were calculated. A power analysis was conducted to determine the required sample size.</p><p><strong>Results: </strong>APT SI was negatively correlated with age (r = -0.510, <i>P</i> < 0.001), whereas ADC was positively correlated with age (r = 0.317, <i>P</i> = 0.006). The APT SI was significantly higher in group A (1.77 ± 0.41) than in group B1 (1.43 ± 0.21), group B2 (1.37 ± 0.31), group B3 (1.30 ± 0.35), and group B4 (1.20 ± 0.35) (all <i>P</i> < 0.01). The ADC was significantly higher in group A [(0.549 ± 0.091) × 10^-3 mm²/s] compared with group B1 [(0.449 ± 0.047) × 10^-3 mm²/s], group B2 [(0.475 ± 0.022) × 10^-3 mm²/s], and group B3 [(0.488 ± 0.051) × 10^-3 mm²/s] (all <i>P</i> < 0.01), whereas no statistically significant difference was found between group A and group B4 (<i>P</i> > 0.05).</p><p><strong>Conclusion: </strong>The APT SI of the normal testes decreased with age, whereas a significant elevation of APT SI was detected in patients with hypospermatogenesis caused by epididymo-orchitis.</p><p><strong>Clinical significance: </strong>Hypospermatogenesis caused by degeneration or inflammation can be differentiated by APT quantity combined with ADC value.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144110116","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":"The role of T1 hyperintensity in differentiating granulomatous prostatitis from prostate cancer: a retrospective analysis of 31 lesions.","authors":"Selahattin Durmaz, Mert Kılıç, Bilgen Coşkun, Sergin Akpek, Barış Bakır, Tarık Esen, Metin Vural, Emre Altınmakas","doi":"10.4274/dir.2025.253242","DOIUrl":"https://doi.org/10.4274/dir.2025.253242","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;To investigate the multiparametric magnetic resonance imaging (mpMRI) characteristics of granulomatous prostatitis (GP) and share our experience with 31 pathologically confirmed GP lesions in 19 patients.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;This two-center retrospective study reviewed the pathological and imaging data of 856 patients who underwent prostate biopsy between January 2012 and April 2024. Of these, 19 patients with available prebiopsy mpMRI and a pathologically confirmed diagnosis of GP were included. Additionally, 280 biopsy-naïve patients diagnosed with clinically significant prostate cancer (csPCa) were included as a control group for comparative analysis. Prebiopsy mpMR images of patients with GP were assessed by consensus between two of three radiologists (M.V., B.C., S.D.), evaluating lesion location, size, shape, multifocality, extraprostatic extension (EPE), signal characteristics on T1-, T2-, and diffusion-weighted imaging (DWI), the mean apparent diffusion coefficient (ADC&lt;sub&gt;mean&lt;/sub&gt;) value, enhancement patterns, and prostate imaging reporting and data system (PI-RADS) scores. Statistical analyses were conducted using SPSS version 30.0.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;In 19 patients, 31 pathologically confirmed GP lesions were identified on prebiopsy mpMRI. Twenty-six lesions were located in the peripheral zone and five in the transitional zone. Multifocal involvement was observed in nine patients (47.3%). Thirty of 31 lesions were hypointense on T2-WI, and seven showed capsular bulging and/or irregularity, suggesting EPE. DWI revealed markedly impeded diffusion in all lesions. The median ADC&lt;sub&gt;mean&lt;/sub&gt; value was 825 × 10&lt;sup&gt;-3&lt;/sup&gt; mm&lt;sup&gt;2&lt;/sup&gt;/s (IQR: 230 × 10&lt;sup&gt;-3&lt;/sup&gt; mm&lt;sup&gt;2&lt;/sup&gt;/s). On dynamic contrast-enhanced sequences, 25 lesions showed early enhancement, five showed prolonged enhancement, and one showed prolonged ring enhancement. Based on mpMRI findings, 17 lesions were assigned a PI-RADS score of 4, and 13 lesions were assigned a PI-RADS score of 5. Notably, 22 lesions (71%) in 14 patients with GP (73.7%) exhibited hyperintensity on T1-WI despite no prior prostate biopsy history. Statistical analysis comparing the GP and csPCa groups revealed that hyperintensity on T1-WI was significantly more frequent in GP, both on a per-patient basis (73.7% vs. 3.2%) and a per-lesion basis (71.0% vs. 3.1%) (&lt;i&gt;P&lt;/i&gt; &lt; 0.0001 for both).&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusion: &lt;/strong&gt;GP shares overlapping imaging features with prostate cancer on mpMRI. However, hyperintensity on T1-WI may serve as a distinguishing feature, potentially reducing unnecessary prostate interventions. Radiologists should consider GP in PI-RADS ≥4 lesions exhibiting T1-WI hyperintensity. Furthermore, given the high incidence of GP following intravesical Bacillus Calmette-Guérin (BCG) therapy, a thorough history of BCG treatment should be obtained.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Clinical significance: &lt;/strong&gt;GP is recognized for its tenden","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144110166","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":"New lens on breast health: harnessing high-b-value synthetic diffusion-weighted imaging for breast lesion characterization.","authors":"Serap Karabiyik, Saime Ramadan, Emil Settarzade, Ali Ilker Filiz, Hatice Ozturkmen Akay","doi":"10.4274/dir.2025.253190","DOIUrl":"https://doi.org/10.4274/dir.2025.253190","url":null,"abstract":"<p><strong>Purpose: </strong>This study aims to evaluate the diagnostic efficacy of synthetic diffusion-weighted imaging (sDWI) at various high b-values in distinguishing malignant from benign breast lesions and to compare its performance with that of conventional DWI (cDWI).</p><p><strong>Methods: </strong>After the exclusion of 22 lesions, 63 women (age range, 24-99 years; mean age, 53.7 ± 15.1 years) with 68 suspicious breast lesions on ultrasound who underwent multiparametric breast magnetic resonance imaging before biopsy between January 2021 and April 2023 were included in this retrospective study. According to the pathological results, lesions were classified as malignant or benign. Volumetric mask images were defined. The lesion signal/normal breast signal ratio [relative signal intensity (rSI)] was measured on different diffusion-weighted images (cDWI at b = 800 and 1500 s/mm²; sDWI at b = 1500-5000 s/mm²), and lesion SI on apparent diffusion coefficient (ADC) 0-800 and ADC0-1500 maps (mADC) was calculated. The diagnostic performances of these parameters were evaluated using a receiver operating characteristic curve analysis and the DeLong test in both the mass and non-mass lesion groups.</p><p><strong>Results: </strong>A total of 32 (47.06%) benign and 36 (52.94%) malignant lesions were identified. Malignant lesions exhibited significantly higher rSI values on cDWI800, cDWI1500, sDWI1500, sDWI2000, and sDWI3000 (<i>P</i> values: <0.001, <0.001, <0.001, <0.001, <0.001, 0.03) and lower mADC800 and mADC1500 values (<i>P</i> values: 0.01 and 0.03). In mass lesions, synthetic b1500 and conventional b1500 demonstrated diagnostic accuracy comparable with that of routine mADC800 and mADC1500. However, in non-mass lesions, high-b-value DWI maps (b ≥ 2000 s/mm²) significantly outperformed mADC and cDWI in differentiating malignant from benign lesions. The highest diagnostic accuracy in non-mass lesions was observed with rSIC4000 [area under the curve (AUC) = 0.87], whereas in mass lesions, rSIC1500 exhibited the highest diagnostic performance (AUC = 0.79).</p><p><strong>Conclusion: </strong>The optimal b-value for DWI differs between mass and non-mass breast lesions, emphasizing the need for separate evaluation protocols. Although high-b-value sDWI provides limited added diagnostic value in mass lesions, it significantly improves malignancy detection in non-mass lesions, outperforming cDWI and ADC mapping.</p><p><strong>Clinical significance: </strong>This study underscores the need for a tailored DWI protocol for optimal breast lesion characterization, particularly for non-mass lesions, where high-b-value synthetic imaging enhances diagnostic accuracy.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144110121","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":"Evaluating the reference accuracy of large language models in radiology: a comparative study across subspecialties.","authors":"Yasin Celal Güneş, Turay Cesur, Eren Çamur","doi":"10.4274/dir.2025.253101","DOIUrl":"https://doi.org/10.4274/dir.2025.253101","url":null,"abstract":"&lt;p&gt;&lt;strong&gt;Purpose: &lt;/strong&gt;This study aimed to compare six large language models (LLMs) [Chat Generative Pre-trained Transformer (ChatGPT)o1-preview, ChatGPT-4o, ChatGPT-4o with canvas, Google Gemini 1.5 Pro, Claude 3.5 Sonnet, and Claude 3 Opus] in generating radiology references, assessing accuracy, fabrication, and bibliographic completeness.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Methods: &lt;/strong&gt;In this cross-sectional observational study, 120 open-ended questions were administered across eight radiology subspecialties (neuroradiology, abdominal, musculoskeletal, thoracic, pediatric, cardiac, head and neck, and interventional radiology), with 15 questions per subspecialty. Each question prompted the LLMs to provide responses containing four references with in-text citations and complete bibliographic details (authors, title, journal, publication year/month, volume, issue, page numbers, and PubMed Identifier). References were verified using Medline, Google Scholar, the Directory of Open Access Journals, and web searches. Each bibliographic element was scored for correctness, and a composite final score [(FS): 0-36] was calculated by summing the correct elements and multiplying this by a 5-point verification score for content relevance. The FS values were then categorized into a 5-point Likert scale reference accuracy score (RAS: 0 = fabricated; 4 = fully accurate). Non-parametric tests (Kruskal-Wallis, Tamhane's T2, Wilcoxon signed-rank test with Bonferroni correction) were used for statistical comparisons.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Results: &lt;/strong&gt;Claude 3.5 Sonnet demonstrated the highest reference accuracy, with 80.8% fully accurate references (RAS 4) and a fabrication rate of 3.1%, significantly outperforming all other models (&lt;i&gt;P&lt;/i&gt; &lt; 0.001). Claude 3 Opus ranked second, achieving 59.6% fully accurate references and a fabrication rate of 18.3% (&lt;i&gt;P&lt;/i&gt; &lt; 0.001). ChatGPT-based models (ChatGPT-4o, ChatGPT-4o with canvas, and ChatGPT o1-preview) exhibited moderate accuracy, with fabrication rates ranging from 27.7% to 52.9% and &lt;8% fully accurate references. Google Gemini 1.5 Pro had the lowest performance, achieving only 2.7% fully accurate references and the highest fabrication rate of 60.6% (&lt;i&gt;P&lt;/i&gt; &lt; 0.001). Reference accuracy also varied by subspecialty, with neuroradiology and cardiac radiology outperforming pediatric and head and neck radiology.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Conclusion: &lt;/strong&gt;Claude 3.5 Sonnet significantly outperformed all other models in generating verifiable radiology references, and Claude 3 Opus showed moderate performance. In contrast, ChatGPT models and Google Gemini 1.5 Pro delivered substantially lower accuracy with higher rates of fabricated references, highlighting current limitations in automated academic citation generation.&lt;/p&gt;&lt;p&gt;&lt;strong&gt;Clinical significance: &lt;/strong&gt;The high accuracy of Claude 3.5 Sonnet can improve radiology literature reviews, research, and education with dependable references. The poor performance of other models, with h","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2025-05-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143987792","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":"Comparison of the novel simultaneous biplane versus in-plane imaging technique in ultrasound-guided biopsy: a prospective randomized multi-operator cross-over phantom study.","authors":"Baki Akca, Florian Vafai-Tabrizi, Michel Bielecki, Georg-Christian Funk","doi":"10.4274/dir.2025.253191","DOIUrl":"https://doi.org/10.4274/dir.2025.253191","url":null,"abstract":"<p><strong>Purpose: </strong>To evaluate and compare the in-plane and novel biplane imaging techniques in ultrasound-guided biopsies (USBx). USBx are effective for obtaining tissue samples in suspected malignancy or infection. The in-plane technique is the gold standard, offering continuous needle visualization. The biplane technique enables simultaneous in-plane and out-of-plane visualization, potentially improving biopsy outcomes. A study was conducted using gel phantoms to simulate USBx, with the goal of determining whether one technique offers distinct advantages over the other.</p><p><strong>Methods: </strong>A total of 30 participants (mean age: 30 ± 7 years; 20 men) were recruited, primarily consisting of physicians in training with varying levels of experience. Each participant performed biopsies on gel phantoms using both the in-plane and biplane techniques in a randomized order after watching a standardized tutorial video. Procedure-related parameters were analyzed, and post-intervention questionnaires, including the NASA task load index (NASA-TLX), were collected to assess cognitive workload and personal preferences.</p><p><strong>Results: </strong>All participants achieved successful biopsies with both techniques. The first-puncture success rate was significantly higher with the biplane technique (83% vs. 63%; <i>P</i> = 0.01). The biplane technique required significantly fewer biopsy attempts than the in-plane approach (37 vs. 43; <i>P</i> = 0.03). Although the biplane technique had a longer \"mean time to first successful biopsy\" (120 seconds vs. 72 seconds), this difference was not statistically significant (<i>P</i> = 0.09), likely due to high variability. No significant differences were found in safety-related parameters, including the number of skin punctures, needle retractions, percentage of time the needle tip was visible, and the number of biopsy attempts without needle tip visualization. The NASA-TLX indicated higher mental demand with the biplane technique (<i>P</i> = 0.013), but other dimensions showed no significant differences. Overall, 83% of participants, including 88% of more experienced operators, preferred the biplane technique, citing enhanced visualization and perceived safety.</p><p><strong>Conclusion: </strong>In this study, the biplane technique in USBx was substantially superior in terms of total biopsy attempts and first-puncture success rate compared with the in-plane approach. It may offer safety and efficiency advantages, particularly for less-experienced operators. Further studies with larger sample sizes and experienced operators, especially in clinical settings, are needed to determine clear superiority.</p><p><strong>Clinical significance: </strong>These findings suggest that biplane imaging may be especially beneficial for training less-experienced operators and in cases with elevated complication risk.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2025-05-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143962513","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":"Pix2Pix generative-adversarial network in improving the quality of T2-weighted prostate magnetic resonance imaging: a multi-reader study.","authors":"Yeliz Başar, Mustafa Said Kartal, Mustafa Ege Seker, Deniz Alis, Delal Seker, Müjgan Orman, Sabri Şirolu, Serpil Kurtcan, Aydan Arslan, Nurper Denizoğlu, İlkay Öksüz, Ercan Karaarslan","doi":"10.4274/dir.2025.243102","DOIUrl":"https://doi.org/10.4274/dir.2025.243102","url":null,"abstract":"<p><strong>Purpose: </strong>To assess the performance and feasibility of generative deep learning in enhancing the image quality of T2-weighted (T2W) prostate magnetic resonance imaging (MRI).</p><p><strong>Methods: </strong>Axial T2W images from the prostate imaging: cancer artificial intelligence dataset (n = 1,476, biologically males; n = 1,500 scans) were used, partitioned into training (n = 1300), validation (n = 100), and testing (n = 100) sets. A Pix2Pix model was trained on original and synthetically degraded images, generated using operations such as motion, Gaussian noise, blur, ghosting, spikes, and bias field inhomogeneities to enhance image quality. The efficacy of the model was evaluated by seven radiologists using the prostate imaging quality criteria to assess original, degraded, and improved images. The evaluation also included tests to determine whether the images were original or synthetically improved. Additionally, the model's performance was tested on the in-house external testing dataset of 33 patients. The statistical significance was assessed using the Wilcoxon signedrank test.</p><p><strong>Results: </strong>Results showed that synthetically improved images [median score (interquartile range) 4.71 (1)] were of higher quality than degraded images [3.36 (3), <i>P</i> = 0.0001], with no significant difference from original images [5 (1.14), <i>P</i> > 0.05]. Observers equally identified original and synthetically improved images as original (52% and 53%), proving the model's ability to retain realistic attributes. External testing on a dataset of 33 patients confirmed a significant improvement (<i>P</i> = 0.001) in image quality, from a median score of 4 (2.286)-4.71 (1.715).</p><p><strong>Conclusion: </strong>The Pix2Pix model, trained on synthetically degraded data, effectively improved prostate MRI image quality while maintaining realism and demonstrating both applicability to real data and generalizability across various datasets.</p><p><strong>Clinical significance: </strong>This study critically assesses the efficacy of the Pix2Pix generative-adversarial network in enhancing T2W prostate MRI quality, demonstrating its potential to produce high-quality, realistic images indistinguishable from originals, thereby potentially advancing radiology practice by improving diagnostic accuracy and image reliability.</p>","PeriodicalId":11341,"journal":{"name":"Diagnostic and interventional radiology","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2025-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143977174","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}