Japanese Journal of Radiology最新文献

{"title":"Optimizing efficiency in CT services: a data-driven approach using action logs from the radiology information system.","authors":"Shingo Kayano, Hideki Ota, Yoshihiko Takubo, Hiromu Mashiko, Kosuke Usui, Kazuki Shimada, Minori Hoshika-Kayano, Kei Takase","doi":"10.1007/s11604-026-02034-1","DOIUrl":"10.1007/s11604-026-02034-1","url":null,"abstract":"<p><strong>Purpose: </strong>To evaluate whether Radiology Information System (RIS) action log-guided continuous quality improvement (KAIZEN) improved CT operational efficiency.</p><p><strong>Materials and methods: </strong>This single-center retrospective before-and-after quality improvement study included 150,540 weekday CT examinations from 2018 to 2022. Five workflow interventions were introduced sequentially from January through February 2021. Daily median examination time, daily median patient entry/exit time, and the end time of the final weekday CT examination were analyzed using segmented interrupted time-series regression with interruptions in January 2020 and January 2021 and heteroskedasticity- and autocorrelation-consistent standard errors. Key performance indicator (KPI) models were adjusted for daily volume, day of the week, and inpatient proportion; unadjusted models were used as sensitivity analyses.</p><p><strong>Results: </strong>Daily median examination time decreased from 351 s at baseline to 328 s after KAIZEN. KAIZEN initiation was associated with an immediate reduction in adjusted examination time (- 17.37 s; 95% CI, - 31.35 to - 3.40 s; p = 0.015), whereas the unadjusted sensitivity analysis was directionally consistent but non-significant (- 11.07 s; p = 0.060). Entry/exit time decreased from 216 to 198 s and showed a significant adjusted immediate reduction (- 15.99 s; 95% CI, - 23.93 to - 8.04 s; p < 0.001), consistent with the unadjusted analysis. The final weekday CT examination ended 12.49 min earlier (95% CI, - 22.11 to - 2.87 min; p = 0.011).</p><p><strong>Conclusion: </strong>RIS action log-guided KAIZEN was temporally associated with shorter turnover, earlier completion of weekday operations, and reduced adjusted examination time. Replication in diverse settings is warranted. Secondary Abstract. In this single-center study, RIS action log-guided KAIZEN was temporally associated with shorter CT patient entry/exit time, earlier weekday completion, and reduced adjusted examination time, supporting routine log-based workflow monitoring.</p>","PeriodicalId":14691,"journal":{"name":"Japanese Journal of Radiology","volume":" ","pages":"1874-1885"},"PeriodicalIF":2.1,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13623989/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148251603","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":"Computed tomography predictors of failure of conservative management in isolated superior mesenteric artery dissection.","authors":"Sayaka Shirai, Tatsuo Ueda, Fumie Sugihara, Hidemasa Saito, Ryutaro Fujitsuna, Taiga Matsumoto, Misa Iwasaki, Hiromitsu Hayashi, Shin-Ichiro Kumita","doi":"10.1007/s11604-026-02023-4","DOIUrl":"10.1007/s11604-026-02023-4","url":null,"abstract":"<p><strong>Purpose: </strong>To identify contrast-enhanced computed tomography (CECT) features associated with failure of conservative management in participants with acute-phase isolated superior mesenteric artery dissection (ISMAD).</p><p><strong>Materials and methods: </strong>This single-center, retrospective, observational study included consecutive patients with acute-phase ISMAD between January 2011 and September 2022. Failure of conservative management was defined as persistent abdominal pain and/or clinical or radiological deterioration despite standardized conservative treatment, prompting close monitoring or consideration of revascularization. The patients were classified into two groups: a failure group (F-G) and a successful group (S-G). The groups were compared with respect to clinical background, Yun classification, true lumen stenosis rate, the cutoff value for the true lumen stenosis rate, and dissection length. Additionally, to evaluate the degree of development of Riolan's arch (RA), computed tomography images were initially reviewed to confirm visualization, and the maximal vascular diameter was subsequently measured.</p><p><strong>Results: </strong>Thirty-two consecutive patients with acute-phase ISMAD (median age, 56.7 years; range, 41-81 years; men, n = 26, 81.3%) were included in this study. Nine and 23 patients were classified as F-G and S-G, respectively, according to the study criteria. The median true lumen stenosis rate was significantly higher in the F-G than in the S-G (100.0 vs. 55.0%, P < 0.05). Receiver operating characteristic analysis yielded an exploratory cutoff value of 66.8%, with an AUC of 0.855 (95% CI 0.71-0.99); sensitivity, 1.000; specificity, 0.652; positive predictive value, 0.50; and negative predictive value, 1.00. RA was more frequently detected in the F-G than in the S-G (9/9 vs. 12/23, P < 0.05).</p><p><strong>Conclusion: </strong>A high true lumen stenosis rate and the presence of RA on CECT are associated with failure of conservative management in ISMAD and may assist in identifying patients at risk of a clinical course requiring closer monitoring or consideration of revascularization rather than directly predicting intestinal ischemia.</p>","PeriodicalId":14691,"journal":{"name":"Japanese Journal of Radiology","volume":" ","pages":"1802-1809"},"PeriodicalIF":2.1,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13624004/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148251554","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":"Nationwide organ volume distributions and cross-sectional age-associated differences in abdominal CT from Japan.","authors":"Tomohiro Kikuchi, Kohei Yamamoto, Yosuke Yamagishi, Toshiaki Akashi, Shouhei Hanaoka, Takeharu Yoshikawa, Hiroyuki Fujii, Harushi Mori, Hisaki Makimoto, Takahide Kohro","doi":"10.1007/s11604-026-02028-z","DOIUrl":"10.1007/s11604-026-02028-z","url":null,"abstract":"<p><strong>Purpose: </strong>Large-scale CT-based reference distributions for abdominal organ volume that incorporate age, sex, and body size are limited. To establish sex- and age-specific reference distributions for major abdominal organ volumes on non-contrast abdominopelvic CT in a nationwide Japanese cohort to provide a foundation for automated clinical assessment.</p><p><strong>Materials and methods: </strong>In this retrospective, multicenter study, we used the Japan Medical Image Database to identify all non-contrast abdominopelvic CT examinations performed in 2024. Unique adults with available data on age, sex, height, and weight were included in this study. The final sample comprised 49,764 examinations (26,456 men and 23,308 women) conducted at nine different institutions. Automated segmentation (TotalSegmentator v2.10.0) was used to produce organ volumes, excluding the hollow viscera. Sex-specific 10th, 25th, 50th, 75th, and 90th percentiles were calculated. Age-volume relationships of body surface area (BSA)-normalized volumes (mL/m²) were modeled using natural cubic splines (four degrees of freedom) separately for each sex.</p><p><strong>Results: </strong>The median male vs. female volumes (mL) were as follows: liver, 1,194.7 vs 1,024.0; pancreas, 63.7 vs 52.3; spleen, 118.7 vs 95.4; kidneys (total), 268.3 vs 221.2; adrenals (total), 6.7 vs 4.2; iliopsoas (total), 483.4 vs 317.7; and prostate, 25.5 (men only). The age-volume relationships of BSA-normalized volumes showed convex patterns for the liver, pancreas, and kidneys in both sexes and for male adrenal glands; lower values in older age groups for the spleen and iliopsoas in both sexes; and higher values in older age groups for the prostate and female adrenal glands.</p><p><strong>Conclusion: </strong>This nationwide Japanese CT cohort provides sex- and age-resolved organ volume reference distributions. These distributions may serve as a quantitative reference for clinical interpretation and population-based imaging research.</p>","PeriodicalId":14691,"journal":{"name":"Japanese Journal of Radiology","volume":" ","pages":"1810-1817"},"PeriodicalIF":2.1,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13623781/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148217819","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":"Diffusion tensor image analysis along the perivascular space (DTI-ALPS): redefining its interpretation and role.","authors":"Toshiaki Taoka, Rintaro Ito, Rei Nakamichi, Toshiki Nakane, Shinji Naganawa","doi":"10.1007/s11604-026-02035-0","DOIUrl":"10.1007/s11604-026-02035-0","url":null,"abstract":"<p><p>Diffusion tensor image analysis along the perivascular space (DTI-ALPS) was originally developed and defined as a noninvasive diffusion MRI method intended to evaluate glymphatic function and is now widely used in neuroimaging research. However, accumulating evidence suggests that the biological meaning of the ALPS index is more complex than initially assumed as a simple marker of glymphatic function. The ALPS index does not directly measure whole-brain fluid transport but rather reflects localized directional diffusivity based on the Brownian motion of water molecules. Its value is strongly influenced by white matter microstructure, including fiber orientation, crossing fibers, extracellular geometry, and age-related diffusivity changes. In addition, because the ALPS index is a ratio-based measure derived from directional diffusivity components, its alterations may arise from different combinations of numerator and denominator diffusivity changes, indicating that similar ALPS index reductions may reflect distinct underlying microstructural mechanisms. White matter microstructure may represent not merely a confounding factor, but a structural substrate guiding interstitial fluid transport itself. White matter hydraulic permeability exhibits strong anisotropy, and brain fluid transport may occur preferentially along white matter tracts. Thus, the structure dependence of the ALPS index may reflect physiologically relevant interactions between white matter architecture and interstitial fluid dynamics rather than simple measurement bias. In this review, we propose redefining the ALPS index not as a direct marker of glymphatic function, but as a \"spatially fixed-point biomarker\" evaluating directional diffusivity within anatomically defined regions. Within this framework, the ALPS index can be understood as a composite biomarker reflecting interactions among white matter microstructure, extracellular environment, vascular geometry, and neurofluid-related tissue environment. Because the ALPS index has been associated with aging, sleep, and various diseases, it may function less as a disease-specific marker and more as an adjunctive imaging marker relevant to brain health assessment and the brain tissue environment.</p>","PeriodicalId":14691,"journal":{"name":"Japanese Journal of Radiology","volume":" ","pages":"1707-1725"},"PeriodicalIF":2.1,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13624023/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148258123","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":"Assessing and mitigating demographic bias in large language models for diagnostic radiology.","authors":"Daisuke Horiuchi, Hiroyuki Tatekawa, Hirotaka Takita, Yasuhito Mitsuyama, Shannon L Walston, Kenichi Saito, Taro Shimono, Yukio Miki, Daiju Ueda","doi":"10.1007/s11604-026-02021-6","DOIUrl":"10.1007/s11604-026-02021-6","url":null,"abstract":"<p><strong>Purpose: </strong>Large language models (LLMs) are increasingly integrated into radiology workflows, but their demographic biases have not been evaluated in diagnostic radiology. This study aimed to investigate racial and sex biases in the diagnostic performance of LLMs (text-only and vision models) in radiology and to evaluate whether prompting strategies mitigate these biases.</p><p><strong>Materials and methods: </strong>This retrospective study included consecutive Diagnosis Please cases published in Radiology from April 1998 to October 2024, excluding cases with sex-specific diseases. For each case, eight race-sex scenarios were generated by altering four race/ethnicity categories (Asian, Black, Hispanic, White) and two sex categories (male, female). Three LLMs (GPT-5, Claude Sonnet 4.5, Gemini 2.5 Flash) were evaluated as text-only models (medical history and imaging findings) and vision models (medical history and images) using three prompting strategies (basic, self-consistency, chain-of-thought prompting). Generalized estimating equations were used to compare diagnostic accuracy across race/ethnicity, sex, and prompting strategies.</p><p><strong>Results: </strong>A total of 286 cases were included. Across models and conditions, ten significant race-related and four significant sex-related differences in diagnostic accuracy were observed. Among the four race/ethnicity groups, Black patients were most likely to have significantly lower accuracy (four of the ten statistically significant race-related comparisons [40%]) and least likely to have significantly higher accuracy (1/10 [10%]). Vision models with female patients under basic prompting showed a higher number of significant race-related differences (6/10 [60%]) than vision models with male patients and all text-only models. Text-only and vision models showed no statistically significant differences in diagnostic accuracy across prompting strategies (p = 0.78 and 0.95, respectively); basic and self-consistency prompting produced ten and four significant race- or sex-related differences, respectively, whereas no significant differences were observed with chain-of-thought prompting.</p><p><strong>Conclusion: </strong>Large language models exhibited racial and sex biases in diagnostic radiology, and chain-of-thought prompting may help mitigate these biases.</p>","PeriodicalId":14691,"journal":{"name":"Japanese Journal of Radiology","volume":" ","pages":"1863-1873"},"PeriodicalIF":2.1,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13623782/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148163227","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":"Updates on molecular classification and treatment in endometrial cancer.","authors":"Kensuke Sakai, Takuma Yoshimura, Wataru Yamagami","doi":"10.1007/s11604-026-02031-4","DOIUrl":"10.1007/s11604-026-02031-4","url":null,"abstract":"<p><p>Endometrial cancer (EC) is the most common gynecological malignancy in developed countries, with a steadily rising global incidence. The Cancer Genome Atlas (TCGA) study identified four molecular subtypes- POLE (ultramutated), MSI (hypermutated), copy number low, and copy number high-each with distinct prognostic and therapeutic implications. Integration of molecular classification into the WHO 2020 classification and the FIGO 2023 staging system has transformed clinical practice. Although major European ESGO/ESTRO/ESP guidelines now incorporate molecular subtyping into treatment algorithms, the Japanese JSGO 2023 guidelines have not yet formally integrated molecular classification into adjuvant treatment decisions. Notable differences exist between Japanese and Western approaches, particularly regarding the role of adjuvant radiation therapy versus chemotherapy. Recent landmark trials have provided compelling evidence for biomarker-driven treatment based on molecular classification. Furthermore, some immune checkpoint inhibitors (ICIs) and antibody-drug conjugates (ADCs) are emerging as promising therapeutic options. This review provides an updated overview of the molecular classification of EC, its impact on current guidelines with emphasis on radiation therapy, and the latest therapeutic advances including immunotherapy and ADCs.</p>","PeriodicalId":14691,"journal":{"name":"Japanese Journal of Radiology","volume":" ","pages":"1758-1767"},"PeriodicalIF":2.1,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148271515","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":"In reply: Moving from maximum uptake to volumetric heterogeneous dose coverage in BNCT.","authors":"Yuki Yoshino, Satoshi Takeno, Teruhito Aihara, Naonori Hu, Yasukazu Kanai, Akinori Sasaki, Mai Nojiri, Keiji Nihei, Koji Ono","doi":"10.1007/s11604-026-02057-8","DOIUrl":"10.1007/s11604-026-02057-8","url":null,"abstract":"","PeriodicalId":14691,"journal":{"name":"Japanese Journal of Radiology","volume":" ","pages":"1889-1890"},"PeriodicalIF":2.1,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148548875","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":"MYC upregulation as candidate predictive biomarker for radioresistance in cervical cancer.","authors":"Kento Tomizawa, Soichiro Sakamoto, Yuhei Miyasaka, Takahiro Oike, Hidenari Hirata, Yuya Yoshimoto, Yasushi Sasaki, Takashi Tokino, Ken Ando, Atsushi Shibata, Tatsuya Ohno","doi":"10.1007/s11604-026-02036-z","DOIUrl":"10.1007/s11604-026-02036-z","url":null,"abstract":"<p><strong>Background: </strong>A subset of uterine cervical cancers recurs locally after definitive radiotherapy, highlighting the need for establishing biomarkers to predict radioresistance for dose optimisation. However, low local recurrence rate constitutes a barrier to research progress. This study aimed at exploring candidate predictive biomarker of radioresistance of uterine cervical cancer by analyzing multiple omics layers using nested case-control approach.</p><p><strong>Methods: </strong>A discovery cohort (n = 8) was created from a series of 192 patients (all with squamous cell carcinoma of the uterine cervix underwent definitive radiotherapy) by 1:1 random case-control matching on pelvic recurrence (PR) for RNA sequencing. A validation cohort (n = 54) was created from the same series by 1:2 propensity score matching on PR for panel-based DNA sequencing and pelvic recurrence-free survival (PRFS) analysis. by using the Kaplan-Meier method and cause-specific proportional hazards models.</p><p><strong>Results: </strong>In the discovery cohort, Gene set enrichment analysis revealed that the genes targeted by MYC were upregulated in patients with PR. In the validation cohort, the prevalence of MYC amplification was significantly greater in patients with PR than in those without (72% vs. 36%, respectively). The 5-year PRFS of patients with MYC amplification was significantly worse than that of those without (47.4% vs. 80.6%, respectively). Multivariate analysis identified MYC amplification as the significant predictor of PR.</p><p><strong>Conclusions: </strong>These data suggest that upregulation of MYC plays a role in radioresistance of uterine cervical cancer, making MYC amplification a potential predictive biomarker for radioresistance.</p>","PeriodicalId":14691,"journal":{"name":"Japanese Journal of Radiology","volume":" ","pages":"1854-1862"},"PeriodicalIF":2.1,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13623779/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148470677","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":"White matter microstructural abnormalities in children with Tourette syndrome using tract-based spatial statistics analysis.","authors":"Jirui Wang, Xianbin Wang, Wenyan Zhang, Zhongyi Liu, Anyi Zhang, Weiwei Men, Guojun Zhang, Xu Hong, Yonghua Cui, Tianyuan Lei","doi":"10.1007/s11604-026-02024-3","DOIUrl":"10.1007/s11604-026-02024-3","url":null,"abstract":"<p><strong>Objectives: </strong>This study aimed to investigate alterations in white matter microstructure in children with Tourette syndrome (TS) and to explore its potential role in pathophysiology.</p><p><strong>Methods: </strong>Diffusion tensor imaging data were collected from 53 children with TS and 91 typically developing controls. White matter integrity was assessed and analyzed using fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD). Correlations with tic severity and quality of life (QOL) were examined, followed by mediation analysis to assess whether motor tic severity mediated the relationship between FA values in the right anterior corona radiata (ACR_R) and QOL.</p><p><strong>Results: </strong>Children with TS exhibited widespread white matter alterations, lower FA in ACR_R and left anterior limb of the internal capsule, lower MD in the inferior fronto-occipital fasciculus, and lower AD in ACR_R, alongside higher MD in the genu of the corpus callosum, higher AD in the inferior fronto-occipital fasciculus, and higher RD in ACR_R. The FA values in ACR_R showed a significant negative correlation with motor tic severity (r = - 0.302, P = 0.031) and a significant negative correlation with physical/activities of daily living (ADL) subscale of QOL (r = - 0.468, P < 0.001). Motor tic severity was positively correlated with physical/ADL subscale of QOL (r = 0.430, P = 0.008). Motor tic severity partially mediated the relationship between FA values in ACR_R and physical/ADL scores (β = - 0.091, 95% CI [- 0.208, - 0.016]).</p><p><strong>Conclusions: </strong>These findings suggest that impaired white matter microstructure may be associated with the pathophysiology of TS, and that future interventions may benefit from simultaneously addressing neural circuit integrity and symptom management.</p>","PeriodicalId":14691,"journal":{"name":"Japanese Journal of Radiology","volume":" ","pages":"1818-1828"},"PeriodicalIF":2.1,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148163142","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":"Verification of the impact of differences between objective and subjective evaluation methods on the interpretation of artificial intelligence systems generating mammogram reports using vision-language model.","authors":"Chiharu Kai, Hideaki Tamori, Yuta Hirono, Sachi Ishizuka, Satoshi Kondo, Tsunehiro Ohtsuka, Satoshi Kasai","doi":"10.1007/s11604-026-02032-3","DOIUrl":"10.1007/s11604-026-02032-3","url":null,"abstract":"<p><strong>Purpose: </strong>Research on vision-language models (VLMs) in the medical field has recently increased. However, while multifaceted evaluation is necessary to avoid the high risks associated with misdiagnosis, artificial intelligence (AI)-assisted mammogram report generation remains insufficient, with no studies on objective and subjective generation. We aimed to develop an AI system that generates mammogram reports and to verify the impact of differences between objective and subjective evaluation methods on the interpretation of this AI system.</p><p><strong>Materials and methods: </strong>We used a public dataset consisting of mammograms and their reports, preparing question prompts and performing low-rank adaptation tuning on Qwen2.5(7B). We analyzed the Breast Imaging Reporting and Data System (BI-RADS) and findings agreement rate, Recall-Oriented Understudy for Gisting Evaluation (ROUGE), and Bilingual Evaluation Understudy (BLEU) for the objective evaluation. A breast clinician performed score-based evaluations of generated reports as subjective assessments. Finally, we analyzed samples of inconsistent objective and subjective results.</p><p><strong>Results: </strong>The BI-RADS agreement rate was 58.1%. Findings were accurately included in generated reports at 76.7% for mass and 81.4% for calcification. ROUGE-L F1 and overall BLEU were 0.672 and 0.542, respectively. Although ROUGE-L F1 or overall BLEU was above average, two samples received low scores in the generated report evaluation; these were considered over- and under-estimation.</p><p><strong>Conclusion: </strong>We developed an AI model that generates mammogram reports. In assessing this model, we found that multifaceted objective and subjective medical VLM evaluations are necessary for determining over- and under-estimation.</p>","PeriodicalId":14691,"journal":{"name":"Japanese Journal of Radiology","volume":" ","pages":"1791-1801"},"PeriodicalIF":2.1,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148549084","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}