BMC Medical Imaging[JOURNAL]
Sun
200 papers
AMD-UPerNet: a tool for retinal layer and fluid assessment in age-related macular degeneration.
Ma Q, Liu X, Li J … +5 more , Bai Y, Mu W, Li N, Yan B, Wang Z
BMC Med Imaging · 2026 Apr · PMID 42015048 · Full text
OBJECTIVE: This study aims to develop a rapid and precise OCT segmentation model that simultaneously delineates retinal layers and fluid regions in age-related macular degeneration (AMD). METHODS: AMD-UPerNet was propose...
OBJECTIVE: This study aims to develop a rapid and precise OCT segmentation model that simultaneously delineates retinal layers and fluid regions in age-related macular degeneration (AMD). METHODS: AMD-UPerNet was proposed as the segmentation model designed to simultaneously delineate retinal layers and fluid compartments in OCT images. Key innovations include: (1) Swin Transformer was adopted as the backbone to extract multi-scale features and model global contextual relationships; (2) Content-Aware ReAssembly of FEatures (CARAFE) module was introduced into feature pyramid network to improve up-sampling and mitigate feature loss; (3) The combination of cross-entropy and Dice loss was used to handle class imbalance and reduce mis-segmentation of fluid regions. RESULT: Compared to the models such as PSPNet, DANet, OCNet, DenseASPP, SAM, MedSAM, and SegFormer, AMD-UPerNet achieved superior performance, with a Pixel Accuracy (PA) of 98.51%, Mean Pixel Accuracy (MPA) of 87.49%, Mean Precision (MPre) of 89.40%, and Mean Intersection over Union (MIoU) of 80.53%. CONCLUSION: Unlike previous methods that segment retinal layers and fluid regions separately, our model uses a joint framework that enables simultaneous optimization and contextual interaction. By using the anatomical continuity of retinal layers as structural priors, AMD-UPerNet improves fluid localization and enhances robustness against ambiguous boundaries and low-contrast regions. These results highlight AMD-UPerNet’s potential to improve OCT segmentation accuracy and efficiency, facilitating early diagnosis and optimized treatment for AMD.
Multicenter prospective study of dedicated breast positron emission tomography (dbPET) for breast cancer: examination in preoperative patients.
Satou Y, Nakagami Y, Suga K … +1 more , Yamaguchi Breast Cancer Study Group (YBC)
BMC Med Imaging · 2026 Apr · PMID 42010499 · Full text
BACKGROUND: Dedicated breast positron emission tomography (dbPET) was developed to detect breast cancers smaller than those detectable using whole-body PET (wbPET). Although several studies have explored the use of dbPET...
BACKGROUND: Dedicated breast positron emission tomography (dbPET) was developed to detect breast cancers smaller than those detectable using whole-body PET (wbPET). Although several studies have explored the use of dbPET, clear criteria for identifying which patients would benefit most from this modality are lacking. Our objective was to determine which patient groups would benefit most from dbPET and how it should be utilized. We conducted a multicenter, prospective exploratory study to investigate how the dbPET maximum standardized uptake values (SUVmax) correlate with patients' clinical characteristics, other imaging modalities, and pathological findings of the lesions. METHODS: In total, 219 patients with breast cancer (median age [range], 58.0 [30-83] years) were included in this study. The enrolled patients were divided into three groups (primary care, neoadjuvant therapy, postoperative follow-up patients). In this research, we examined the primary care group (n = 92). To investigate which patient groups benefit from dbPET, we examined which factors influence and correlate with dbPET SUVmax. Depending on the items being compared, correlation analysis, Wilcoxon signed rank test was used to examine the following items. Which factors (physical factors, pathological characteristics, etc.) correlate with dbPET SUVmax, differences between dbPET and other imaging examination (detection rate, etc.), and whether dbPET SUVmax L/H ratio (Lesion-to-Healthy (normal) site dbPET SUVmax ratio) incorporating dbPET SUVmax from healthy (normal) mammary gland tissue are necessary for evaluating dbPET SUVmax in lesion areas. RESULTS: dbPET SUVmax in healthy(normal) mammary gland tissue were strongly associated with background mammary density observed on mammography (MMG) examination (positive correlation, p<0.05). Ki-67 showed the strongest positive correlation with both the lesion-site dbPET SUVmax (r=0.56, R=0.31, p<0.05) and the dbPET SUVmax L/H ratio (lesion-to-healthy (normal) -site ratio) (r=0.47, R=0.22, p<0.05). Additionally, the tissue grade, MMG and ultrasonography categories were positively correlated with dbPET SUVmax. Regarding the lesion detection rate, dbPET identified 100% of the lesions, including benign findings. The dbPET SUVmax L/H ratio showed a trend nearly identical to that of the dbPET SUVmax. CONCLUSIONS: dbPET demonstrated higher detection capabilities than other imaging tests and showed a strong correlation with tissue malignancy. Therefore, they were suggested to be potentially useful for distinguishing benign findings from malignant lesions that are difficult to differentiate using other imaging tests. The dbPET SUVmax of healthy mammary gland tissue were presumed to correlate with the amount of mammary gland tissue within the breast. However, no significant differences were observed in the correlations between the dbPET values and the L/H ratio and the individual parameters. In this study, it remained unclear whether the dbPET values of normal breast tissue should be taken into account when evaluating dbPET values.
The diagnostic value of ⁶⁸Ga-FAPI PET/CT in patients with suspicious recurrent or metastatic differentiated thyroid cancer: a comparative study.
Liu J, Wang Z, Xu L … +1 more , Pang H
BMC Med Imaging · 2026 Apr · PMID 42010490 · Full text
OBJECTIVE: This study aimed to compare the detection rates and uptake values of gallium-68 (⁶⁸Ga)-labeled fibroblast activation protein inhibitor (FAPI) PET/CT with those of fluorine-18 (¹⁸F)- labeled fluorodeoxyglucose...
OBJECTIVE: This study aimed to compare the detection rates and uptake values of gallium-68 (⁶⁸Ga)-labeled fibroblast activation protein inhibitor (FAPI) PET/CT with those of fluorine-18 (¹⁸F)- labeled fluorodeoxyglucose (FDG) PET/CT in patients with suspicious recurrent or metastatic differentiated thyroid cancer (DTC), and to explore the association of these rates and values with diverse clinical-pathological characteristics and degrees of differentiation based on visual analysis and semi-quantitative parameters. METHODS: A retrospective cohort comprising 56 patients with elevated Tg, positive anti-Tg antibodies, or thyroid ultrasound suggesting recurrence or metastasis was enrolled. All patients had previously undergone ⁶⁸Ga-FAPI PET/CT and ¹⁸F-FDG PET/CT scans. Lesion detection and uptake values were analyzed visually and semi-quantitatively, with final diagnosis confirmed by histopathology or clinical follow-up. RESULTS: ¹⁸F-FDG PET/CT identified more positive patients (86% vs. 76%) and demonstrated a higher detection rate in lymph node metastases (71% vs. 59%, p=0.029). ⁶⁸Ga-FAPI showed higher uptake (SUVmax) in pulmonary metastases (3.31 vs. 1.74, p=0.017) but a lower tumor-to-background ratio (TBR) in lymph nodes (2.20 vs. 4.10, p<0.001). Among the clinical and pathological characteristics assessed, Tg levels, short axis diameter (SAD) of lymph node, and iodine uptake displayed significant differences between ⁶⁸Ga-FAPI uptake negative (FAPI-) and ⁶⁸Ga-FAPI uptake positive (FAPI+) groups (p=0.025, p<0.001, p=0.003). However, only the SAD of lymph node showed a moderate positive correlation with the SUVmax value of ⁶⁸Ga-FAPI in the lesion (r=0.393, p=0.042). CONCLUSION: Our findings indicate that ⁶⁸Ga-FAPI PET/CT does not demonstrate significantly superior diagnostic performance when compared to ¹⁸F-FDG PET/CT in patients with recurrent or metastatic DTC. Nevertheless, for patients with enlarged metastatic lymph nodes, elevated Tg levels, and iodine-affinity lesions, ⁶⁸Ga-FAPI PET/CT could represent a valuable diagnostic tool.
Quantitative assessment of intrapulmonary vessel volume by CTPA in vasculitis patients with pulmonary vascular involvement.
Ming Y, Piao S, Zhao R … +9 more , Wang J, Liao Z, Song L, Xiao R, Zhao R, Ma Z, Li B, Zheng F, Song W
BMC Med Imaging · 2026 Apr · PMID 42010396 · Full text
PURPOSE: To compare intrapulmonary vessel volume (IPVV) on computed tomography pulmonary angiography (CTPA) between vasculitis patients with pulmonary vascular involvement and CTPA-negative subjects. METHODS: This study...
PURPOSE: To compare intrapulmonary vessel volume (IPVV) on computed tomography pulmonary angiography (CTPA) between vasculitis patients with pulmonary vascular involvement and CTPA-negative subjects. METHODS: This study included 207 vasculitis patients with pulmonary vascular involvement between March 2019 and November 2024 and 202 CTPA-negative subjects between February 2019 and February 2025. A computer-aided pulmonary vascular segmentation algorithm was employed to automatically measure total intrapulmonary vessel volume (TIPVV), intrapulmonary arterial vessel volume (IPVVa) and intrapulmonary venous vessel volume (IPVVv). Additionally, IPVVs were analyzed and compared within five specific vessel diameter groups: 0.8–1.6 mm, 1.6–2.4 mm, 2.4–3.2 mm, 3.2–4.0 mm, and > 4.0 mm. RESULTS: TIPVV and IPVVv showed no significant differences between groups. The IPVVa measured in CTPA-negative subjects was 47.79 (42.48, 54.10) mL·m− 2, while that in vasculitis patients with pulmonary vascular involvement was 44.86 (39.33, 52.58) mL·m− 2. The IPVVa in vasculitis patients with pulmonary vascular involvement was significantly lower than that in CTPA-negative subjects (p < 0.01). In pulmonary arteries with diameters of 0.8–1.6 mm and 2.4–3.2 mm, the IPVVa in vasculitis patients with pulmonary vascular involvement was lower than that in CTPA-negative subjects (p < 0.05). In pulmonary veins with diameters of 1.6–2.4 mm and 3.2–4.0 mm, the IPVVv in vasculitis patients with pulmonary vascular involvement was higher than that in CTPA-negative subjects (p < 0.05). CONCLUSIONS: The computer-aided pulmonary vascular segmentation algorithm can automatically measure IPVV, enabling quantitative assessment of small pulmonary vessel involvement in vasculitis.
Bone age assessment: comparative analysis of Greulich-Pyle and Tanner-Whitehouse 3 by pediatric radiologists and endocrinologists.
Tewattanarat N, Chaisiwamongkol W, Wiratchotisatian P … +7 more , Paisarnsrisomsuk S, Hanpinitsak P, Tangprasert K, Poonpol C, Pansukrada C, Thaowandee W, Chaisiwamongkol R
BMC Med Imaging · 2026 Apr · PMID 42001038 · Full text
INTRODUCTION: Two widely used bone age assessment (BAA) methods—Greulich-Pyle (GP) and Tanner-Whitehouse 3 (TW3) methods—differ in complexity, accuracy, and clinical utility. OBJECTIVE: To compare GP and TW3 bone age est...
INTRODUCTION: Two widely used bone age assessment (BAA) methods—Greulich-Pyle (GP) and Tanner-Whitehouse 3 (TW3) methods—differ in complexity, accuracy, and clinical utility. OBJECTIVE: To compare GP and TW3 bone age estimations and assess inter- and intra-observer agreement between pediatric radiologists and endocrinologists. METHODS: This retrospective study analyzed 1,725 left-hand radiographs of children aged 0–19 years (2008–2022). Twelve experts (six radiologists, six endocrinologists) independently assessed bone age using GP and TW3 radius-ulnar-short bones (RUS) methods following standardized training. Assessment time and method preference were recorded. Statistical analyses included descriptive statistics, intra-class correlation coefficients (ICC), Bland–Altman analysis, and comparative tests (p < 0.05). RESULTS: GP estimates ranged from 6 to 228 months, whereas TW3-RUS estimates were slightly lower due to age-range limitations. Inter- and intra-observer reliability were excellent for both methods (ICC > 0.9). Mean bone age did not differ significantly between specialties using GP, whereas minor differences were observed with TW3-RUS for overall (p < 0.001) and when stratified by sex (males p = 0.016, female p = 0.005). Age-stratified analysis demonstrated low mean absolute differences (MAD) between specialties across most age groups, with slightly greater variability in 84.1–180 months. MAD between GP and TW3-RUS was modest in most age groups (approximately 2.5–3.3 months) but increased in older adolescents, particularly ≥ 180 months. GP assessments were significantly faster than TW3-RUS (p = 0.002). Experts preferred GP for routine use due to speed and ease, while TW3-RUS offered greater accuracy, specificity, and detail for complex and borderline cases. CONCLUSIONS: Both specialists demonstrated excellent agreement using both GP and TW3-RUS. We recommend GP as the primary method, reserving TW3-RUS for complex cases. Moreover, developing a GP-analog atlas with narrower, finely defined age intervals may improve clinical applicability.
The value of the "map sign" in shear wave elastography for differentiating seminomas from non-seminomas.
Liu S, Zhou Z, Xue N
BMC Med Imaging · 2026 Apr · PMID 41998592 · Full text
BACKGROUND: Diagnosing testicular tumors that are negative for tumor markers using ultrasound and MRI is currently challenging. Seminomas and non-seminomas of the testis have different clinical diagnoses and treatment pl...
BACKGROUND: Diagnosing testicular tumors that are negative for tumor markers using ultrasound and MRI is currently challenging. Seminomas and non-seminomas of the testis have different clinical diagnoses and treatment plans. Shear wave elastography (SWE) can reflect tissue stiffness, aiding in tumor type differentiation. This study aims to explore the diagnostic value of SWE color images in differentiating seminomas from non-seminomas of the testis by analyzing their characteristic features. METHODS: A retrospective analysis was conducted on the SWE features of 39 pathologically confirmed testicular tumors negative for tumor markers. Based on pathological results, the tumors were divided into seminoma and non-seminoma groups. Differences in conventional ultrasound and SWE features between the two groups were analyzed. RESULTS: There were no statistically significant differences in location, shape, echo type, echo uniformity, SWE maximum value and SWE average value between the seminoma and non-seminoma groups (all p > 0.05), with respective p-values of 0.256, 0.458, 0.847, 0.298, 0.394, and 0.245. However, 76% (16/21) of seminomas showed a “map sign” (predominantly blue with red streaks) on SWE, while 0% (0/18) of non-seminomas showed this sign. There was a statistically significant difference in SWE color image features between the seminoma and non-seminoma groups (p = 0.000). Using the “map sign” as a diagnostic criterion for seminomas, the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 76.2%, 100%, 100%, 78.2% and 87.2%, respectively, showing a statistically significant difference compared to conventional ultrasound (p = 0.030). CONCLUSION: The “map sign” on SWE can accurately diagnose testicular seminomas, improving the accuracy of conventional ultrasound diagnosis. This technique is non-invasive, simple, and repeatable, and it holds promise as a new method for diagnosing testicular seminomas.
Quantitative study and prognostic value of diffusion tensor imaging for lumbar spinal stenosis.
Qian G, Xia Z, Wang M … +3 more , Yang K, Li X, Miao C
BMC Med Imaging · 2026 Apr · PMID 41998586 · Full text
OBJECTIVE: To explore the correlation between diffusion tensor imaging (DTI) parameters and clinical symptoms of lumbar spinal stenosis (LSS) patients, as well as the predictive value of DTI parameter values for conserva...
OBJECTIVE: To explore the correlation between diffusion tensor imaging (DTI) parameters and clinical symptoms of lumbar spinal stenosis (LSS) patients, as well as the predictive value of DTI parameter values for conservative treatment prognosis of LSS patients. METHODS: Prospective collection of 33 patients with LSS accompanied by unilateral lower back and leg pain, matched with 33 volunteers based on clinical baseline data. All subjects underwent 3.0T MR scans and were independently measured by two physicians within the L4-S1 nerve roots. Spearman correlation analysis was used to test the correlation between symptomatic side DTI parameter values and clinical symptoms. Using firth penalized logistic regression analysis to explore independently associated with poor outcome in conservative treatment of LSS patients, and using receiver operating characteristic curve (ROC) to test its predictive value. RESULTS: DTI parameters can be used to evaluate the nerve roots in patients with LSS. There is varying degrees of correlation between DTI parameter values on the symptomatic side and VAS score, JOA score, RMDQ score, and ODI score. The difference in VAS score and the single FA value between the group with good prognosis and the group with poor prognosis is statistically significant, and the area under curve (AUC) of single root FA value is the largest. CONCLUSION: DTI is feasible for evaluating the nerve root symptoms of lumbar spinal stenosis, and there is a correlation between DTI parameter values and clinical symptoms. Single-root FA value is independently associated with poor outcome after conservative treatment of LSS patients. CLINICAL TRIAL NUMBER: Not applicable.
Independent component analysis of brain network alterations associated with cognitive impairment in coal workers' pneumoconiosis.
Lv X, Cao G, Ren L … +7 more , Xue B, Zhao Y, Xu Z, Man X, Hagiwara A, Liu Y, Han X
BMC Med Imaging · 2026 Apr · PMID 41998546 · Full text
OBJECTIVE: This study aimed to investigate alterations in brain functional networks in Coal Workers’ Pneumoconiosis (CWP) using independent component analysis (ICA), a data-driven method that separates resting-state netw...
OBJECTIVE: This study aimed to investigate alterations in brain functional networks in Coal Workers’ Pneumoconiosis (CWP) using independent component analysis (ICA), a data-driven method that separates resting-state networks from fMRI data without requiring predefined regions of interest. And we will explore their correlations with clinical indicators and cognitive functions. METHODS: A total of 91 subjects were finally included after strict inclusion and exclusion criteria, comprising 33 CWP patients with cognitive impairment (CWP-CI), 31 CWP patients without cognitive impairment (CWP-nonCI), and 27 demographically matched healthy controls (HCs). The resting-state fMRI and cognitive assessments using the Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Clock Drawing Test (CDT) were performed in all subjects. Following data preprocessing, ICA was employed to identify resting-state networks (RSNs), and both intra-network and inter-network functional connectivity (FC) analysis were conducted. The correlations were assessed between significantly altered FC, clinical indicators and cognitive scale scores. RESULTS: Eight RSNs were identified, including the default mode network (DMN), executive control network (ECN), salience network (SN), dorsal attention network (DAN), sensorimotor network (SMN), higher visual network (hVN), auditory network (AUN), and basal ganglia network (BG). Intra-network connectivity analysis revealed a significant difference within SMN. Inter-network connectivity analysis revealed specific alterations. Compared to the HC group, the CWP-nonCI group exhibited enhanced FC in multiple network pairs, including hVN-SN, hVN-DMN, ECN-SN, DAN-BG, DMN-BG and SMN-SN, while the CWP-CI group primarily showed enhanced FC between hVN-SN and hVN-DAN. Furthermore, compared to the CWP-nonCI group, the CWP-CI group demonstrated significantly reduced FC between BG-AUN and SN-DMN. Some of the altered inter-network FC was significantly correlated with pulmonary function, cognitive scale scores, and work duration. CONCLUSIONS: CWP patients exhibited specific reorganization of FC between brain networks. These findings suggested that disrupted inter-network functional communication may serve as a key neural mechanism in CWP-related cognitive impairment, providing new neuroimaging evidence for its basis.
Diagnostic performance of artificial intelligence versus conventional imaging for differentiating G2/G3 from G1 pancreatic neuroendocrine tumors: a systematic review and meta-analysis.
Zhang Z, Hu J, Lei P … +3 more , Wu Q, Huang C, Zhu X
BMC Med Imaging · 2026 Apr · PMID 41992154 · Full text
BACKGROUND: Accurate preoperative grading of pancreatic neuroendocrine tumors (PanNETs), specifically differentiating G2/G3 from G1, is pivotal for treatment planning but challenging with conventional biopsy. This study...
BACKGROUND: Accurate preoperative grading of pancreatic neuroendocrine tumors (PanNETs), specifically differentiating G2/G3 from G1, is pivotal for treatment planning but challenging with conventional biopsy. This study aims to evaluate the diagnostic performance of imaging modalities, particularly comparing Machine Learning/Deep Learning (ML/DL) algorithms against conventional expert interpretation. METHODS: We conducted a systematic review and meta-analysis of studies from PubMed, Embase, Web of Science, and Cochrane Library up to December 2025. Studies utilizing CT, MRI, or endoscopic ultrasound (EUS) for PanNET grading were included. A bivariate random-effects model was used to calculate pooled sensitivity, specificity, and area under the curve (AUC). Subgroup analyses were performed to investigate heterogeneity and the impact of validation strategies. RESULTS: Seventeen studies comprising 928 patients were included. The overall pooled sensitivity and specificity were 0.77 (95% CI: 0.71–0.83) and 0.83 (95% CI: 0.77–0.87), respectively, with an AUC of 0.87. Notably, ML/DL models demonstrated significantly higher sensitivity than conventional imaging (0.84 vs. 0.71, p < 0.01) but lower specificity (0.78 vs. 0.87, p < 0.01). Multicenter studies showed a trend toward higher diagnostic metrics compared to single-center cohorts. Interestingly, the performance gap between external and internal validation narrowed when restricted to AI subgroups, suggesting the robustness of modern algorithms. CONCLUSIONS: Imaging-based analysis offers high diagnostic accuracy for PanNET grading. A theoretical sequential diagnostic strategy is suggested: utilizing AI’s high sensitivity for initial screening, followed by expert radiological review to ensure specificity.
Integrating clinical, laboratory and quantitative CT features for predicting split renal function in urinary tract obstruction.
Ma Y, Zhan J, Feng J … +8 more , Zhang W, Liang F, Xu H, Hao Y, Qu L, He W, Lai S, Yang R
BMC Med Imaging · 2026 Apr · PMID 41992134 · Full text
BACKGROUND: Urinary tract obstruction (UTO) can lead to progressive renal impairment, making accurate evaluation of split renal function (SRF) essential for clinical decision-making. Although radionuclide renal dynamic s...
BACKGROUND: Urinary tract obstruction (UTO) can lead to progressive renal impairment, making accurate evaluation of split renal function (SRF) essential for clinical decision-making. Although radionuclide renal dynamic scintigraphy remains the gold standard for SRF assessment, its clinical application is constrained by procedural complexity, limited availability, and sensitivity to anatomical variations. Thus, there is a clinical need for simpler, reliable, and noninvasive alternative approaches. This study aimed to develop and validate a predictive model for SRF grading by integrating clinical variables, laboratory parameters, and quantitative contrast-enhanced computed tomography (CECT) features in patients with UTO. METHODS: A retrospective cohort of 78 patients with UTO (150 kidneys) was analyzed. Based on split renal glomerular filtration rate (GFR) determined using the Gates method, kidneys were categorized into normal, mild-to-moderate impairment, and severe impairment groups. Clinical variables, laboratory parameters, and quantitative CECT features were collected. Univariate and multivariate logistic regression analyses were performed to identify independent predictors and construct SRF grading models. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA). RESULTS: In Task 1 (normal vs. abnormal) and Task 2 (normal vs. mild-to-moderate), age was the key univariate predictor, while hemoglobin (Hb) and renal cortical thickness (Rc) were identified as independent predictors in the laboratory-based and CECT-based multivariate models, respectively. For Task 1, the Combined model [Clinical predictor (Age) + Laboratory model (Hb-based) + CECT model (Rc-based)] achieved the highest diagnostic performance (AUC = 0.890); for Task 2, the optimal model was [Clinical predictor (Age) + CECT model (Rc-based)] (AUC = 0.810). For Task 3 (mild-to-moderate vs. severe), Hb was the strongest univariate predictor, and Rc was the sole independent predictor in the CECT-based multivariate model. The highest diagnostic accuracy for Task 3 was achieved by the combined Laboratory predictor (Hb) + CECT model (Rc-based), with an AUC of 0.970. CONCLUSION: The CECT model (Rc-based) serves as a crucial imaging biomarker for evaluating SRF impairment in patients with UTO. Task-specific models combining the CECT model (Rc-based) with a clinical predictor (Age) and laboratory information—either as a univariable predictor (Hb) or as a multivariable laboratory model (Hb-based)—showed superior predictive performance. This integrated, noninvasive strategy may serve as a useful adjunct to radionuclide imaging for individualized SRF assessment, pending prospective validation.
Infant and toddler's preoperative chest imaging on photon-counting detector CT: a paired-sample comparison with energy-integrating detector CT on image quality and radiation dose.
Liu XY, Zhou W, Shi RY … +6 more , Xie WH, Qiao WD, Sun TX, Xu SY, Wang J, Wu LM
BMC Med Imaging · 2026 Apr · PMID 41992132 · Full text
BACKGROUND: Photon-counting CT (PCCT) may improve dose efficiency compared with energy-integrating detector CT (EID-CT), but evidence remains limited in very young children undergoing non-contrast preoperative chest CT u...
BACKGROUND: Photon-counting CT (PCCT) may improve dose efficiency compared with energy-integrating detector CT (EID-CT), but evidence remains limited in very young children undergoing non-contrast preoperative chest CT under real-world clinical protocols. OBJECTIVE: To provide a size-aware matched comparison of radiation dose and image quality between PCCT and EID-CT in infants and toddlers (≤ 36 months) undergoing non-contrast preoperative chest CT, and to examine whether the main findings remained directionally consistent across prespecified strata of age, BMI, and water-equivalent diameter (DW). METHODS: In this single-center matched cohort study, a prospective PCCT cohort and a retrospective EID-CT cohort were propensity score–matched 1:1 (49 pairs). Dose metrics (CTDIvol, DLP, SSDE) and objective and subjective image quality were evaluated, with subgroup analyses stratified by age (< 12 vs. ≥ 12 months), BMI tertiles, and DW tertiles. RESULTS: The PCCT group showed significantly lower radiation dose metrics (CTDIvol, DLP, and SSDE) compared to the EID-CT group (P < 0.001). In terms of objective image quality, PCCT significantly reduced image noise in all lung lobes and consistently demonstrated higher signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) (all P < 0.001). Subjective image quality assessments showed no significant difference between PCCT and EID-CT (P > 0.05). Across prespecified age, BMI, and DW strata, PCCT remained associated with lower dose metrics and favorable objective image-quality trends, although these subgroup findings should be interpreted as exploratory. CONCLUSION: In this matched cohort of infants and toddlers undergoing non-contrast preoperative chest CT, PCCT was associated with substantially lower radiation dose and improved objective image quality compared with EID-CT, and no statistically significant difference was detected in subjective image quality scores between PCCT and EID-CT.
Development and validation of an MRI-based clinical-radiomics nomogram for predicting lymph node metastasis in non-small cell lung cancer.
Li X, Chang N, Zhang S … +5 more , Hao H, Tian M, Lin X, Li N, Zhao P
BMC Med Imaging · 2026 Apr · PMID 41987095 · Full text
BACKGROUND: To develop and validate a nomogram that integrates clinical factors and multiparametric MRI-based radiomics features for the preoperative prediction of lymph node metastasis (LNM) in non-small cell lung cance...
BACKGROUND: To develop and validate a nomogram that integrates clinical factors and multiparametric MRI-based radiomics features for the preoperative prediction of lymph node metastasis (LNM) in non-small cell lung cancer (NSCLC). METHODS: This retrospective diagnostic accuracy study enrolled 220 patients with pathologically confirmed NSCLC (142 males; 60.77 ± 8.70 years) between September 2021 and October 2024. Patients were randomly divided into training and validation sets. A clinical model was constructed using independent predictors identified by univariable and multivariable logistic regression analysis. A radiomics signature was developed from T1WI, T2WI, and T1 mapping sequences using the least absolute shrinkage and selection operator logistic regression algorithm. A nomogram was developed by integrating the clinical model and the radiomics signature. Diagnostic performance was assessed by receiver operating characteristic analysis, calibration, and decision curve analysis. Two radiologists independently assessed LNM status in the validation set for comparison. RESULTS: Tumor maximum diameter and carcinoembryonic antigen level were identified as independent predictors for the clinical model. In the validation set, the nomogram achieved an area under the curve (AUC) of 0.847, significantly greater than the clinical model (AUC = 0.710, p = 0.033) and the radiomics signature alone (AUC = 0.802, p = 0.033). The AUC of the nomogram was significantly higher than two radiologists (0.847 vs. 0.682, p = 0.022; 0.847 vs. 0.698, p = 0.041, respectively). CONCLUSION: The nomogram could serve as a noninvasive tool for preoperative prediction of LNM in NSCLC, thereby aiding in clinical decision-making.
Development and validation of a combined model based on MRI radiomics for predicting placenta accreta spectrum and adverse pregnancy outcomes: a multicenter retrospective study.
Zhou Z, Ge H, Zou P … +8 more , Ye T, Zou T, Fang Q, Zeng Y, Wang Y, Zhang X, Xu Z, Huang X
BMC Med Imaging · 2026 Apr · PMID 41987062 · Full text
OBJECTIVE: To develop and validate a radiomics-clinical model of MRI fusion model for the prenatal prediction of placenta accreta spectrum (PAS) and pregnancy outcomes. METHODS: This multicentre retrospective study enrol...
OBJECTIVE: To develop and validate a radiomics-clinical model of MRI fusion model for the prenatal prediction of placenta accreta spectrum (PAS) and pregnancy outcomes. METHODS: This multicentre retrospective study enrolled 610 singleton pregnancies (292 with PAS vs.318 controls). After stratified randomization, the participants were allocated to training (n = 408), validation (n = 102), or external testing (n = 100) cohorts. Radiomics features were extracted from T2-weighted sagittal images, and MRI radiomics models, prenatal clinical models, and combined prediction models were developed. RESULTS: Prenatal clinical characteristics were significantly correlated with the presence of PAS in the training cohort (p < 0.05, top = 0.001). The newly constructed integrated prediction model incorporating radiomics features and the three prenatal clinical features had a greater area under the curve (AUC) for predicting PAS than the MRI radiomics and prenatal clinical models in the training cohort (0.767, 0.764, and 0.644; p < 0.05), validation cohort (0.795, 0.759, and 0.708; p < 0.05), and external testing cohort (0.792, 0.777, and 0.758; p < 0.05). Furthermore, in terms of pregnancy outcomes in pregnant women with PAS, the combined prediction model had a greater AUC than the MRI radiomics and prenatal clinical models in the training cohort (0.852, 0.834, and 0.673; p < 0.05), the validation cohort (0.879, 0.804, and 0.747; p < 0.05), and the external testing cohort (0.907, 0.839, and 0.700; p < 0.05). CONCLUSION: The MRI-based multimodal integrated framework significantly outperforms conventional models, achieving dual prediction of PAS diagnosis and pregnancy outcomes with clinical-grade accuracy. CRITICAL RELEVANCE STATEMENT: Our study introduces the models provide objective and efficient decision support for the management of pregnant women with high-risk placenta previa and are expected to become a core tool for the multidisciplinary management of PAS.
The role of radiological imaging in differentiating malignant and benign pulmonary nodules: a retrospective study.
Deng L, Qiu L, Li J … +4 more , Wu S, Li Y, Zhou W, Cheng G
BMC Med Imaging · 2026 Apr · PMID 41987051 · Full text
BACKGROUND: Accurate differentiation between malignant and benign pulmonary nodules remains a critical challenge in clinical practice. This study aimed to evaluate the diagnostic performance of computed tomography (CT) i...
BACKGROUND: Accurate differentiation between malignant and benign pulmonary nodules remains a critical challenge in clinical practice. This study aimed to evaluate the diagnostic performance of computed tomography (CT) imaging features in distinguishing malignant from benign pulmonary nodules. METHODS: A retrospective analysis was conducted on 200 patients with pulmonary nodules who underwent chest CT scanning and subsequent histopathological confirmation between January 2020 and December 2024. CT imaging features including nodule size, density, margin characteristics, internal characteristics, and relationship to adjacent structures were analyzed. To systematically evaluate the independent contribution of CT imaging features, multivariate logistic regression analysis was performed using a progressive modeling approach. Model stability was assessed using bootstrap internal validation (1,000 resamples) and verified by AIC-based backward stepwise selection. Receiver operating characteristic curve analysis was employed to assess the diagnostic performance of CT imaging features. RESULTS: Among 200 cases, 88 (44.00%) were malignant and 112 (56.00%) were benign. Malignant nodules demonstrated significantly larger mean diameter (18.76 ± 8.52 mm vs. 12.85 ± 5.94 mm, P < 0.001), higher prevalence of spiculated margins (70.45% vs. 28.57%, P < 0.001), lobulation (67.05% vs. 33.04%, P < 0.001), and pleural indentation (54.55% vs. 22.32%, P < 0.001). Calcification was more common in benign nodules (43.75% vs. 12.50%, P < 0.001). After adjustment for clinical confounders, nodule size (OR = 1.095, P < 0.001), spiculated margin (OR = 4.523, P < 0.001), lobulation (OR = 2.845, P < 0.001), and calcification (OR = 0.185, P < 0.001) remained independent predictors of malignancy. AIC-based backward stepwise selection independently confirmed the same four CT features, supporting the robustness of variable selection. The integrated model incorporating four CT imaging features achieved sensitivity of 71.59%, specificity of 88.39%, and area under the curve of 0.870(optimism-corrected AUC = 0.845). CONCLUSION: Chest CT imaging features, particularly nodule size, spiculated margin, lobulation, and calcification patterns, are independent predictors of malignancy and demonstrate good diagnostic performance in differentiating malignant from benign pulmonary nodules. These findings provide preliminary evidence supporting the integration of structured CT feature analysis into clinical decision-making; however, external validation in independent cohorts is needed before clinical implementation.
Differentiation of adrenal metastases and adenomas based on clinical characteristics, deep learning features, and radiomics features derived from ultrasound imaging.
Li Y, Song Y, Chang L … +1 more , Wei X
BMC Med Imaging · 2026 Apr · PMID 41981403 · Full text
BACKGROUND: Accurate diagnosis of adrenal incidentalomas is crucial in patients with extra-adrenal malignant tumors. This study aims to develop a nomogram integrating clinical features, deep learning-derived imaging feat...
BACKGROUND: Accurate diagnosis of adrenal incidentalomas is crucial in patients with extra-adrenal malignant tumors. This study aims to develop a nomogram integrating clinical features, deep learning-derived imaging features, and ultrasound radiomics characteristics to distinguish adrenal metastases from adrenal adenomas. METHODS: A retrospective analysis was conducted on 449 cases, including 228 cases of adrenal metastases and 221 cases of adrenal adenomas, divided into training and testing cohorts at a 7:3 ratio. Patient clinical data and ultrasonographic images were collected, with regions of interest (ROIs) delineated on ultrasound images. Feature extraction, selection, and radiomics model (Rad) construction were performed, followed by model evaluation. Multiple deep learning models were employed to identify the optimal architecture for deep feature extraction. These deep features were combined with radiomics features to establish a deep learning radiomics model (DLR) for the differentiation between adrenal metastases and adenomas. RESULTS: The study demonstrates that the Rad, DLR, and combined models exhibit superior diagnostic performance in differentiating adrenal metastases from adrenal adenomas. In the testing cohort, the combined model outperforms the Rad and DLR models. The area under the curve (AUC) in the testing set for Rad, DLR, and combined models were 0.839, 0.839, and 0.850, respectively. CONCLUSION: The nomogram integrating clinical features, deep learning features, and ultrasound radiomics features demonstrates robust performance in differentiating adrenal metastases from adrenal adenomas and can assist in preliminary diagnostic stratification of indeterminate adrenal nodules in patients with extrarenal tumors.
Automated scout-image-based estimation of contrast agent dosing: a deep learning approach.
Schirrmeister RT, Taleb L, Friemel P … +4 more , Reisert M, Bamberg F, Weiß J, Rau A
BMC Med Imaging · 2026 Apr · PMID 41965610 · Full text
BACKGROUND: Optimal contrast agent dosing in computed tomography (CT) depends on accurate patient weight, yet manual measurements increase workload and self-reporting can introduce bias. We developed and tested a deep-le...
BACKGROUND: Optimal contrast agent dosing in computed tomography (CT) depends on accurate patient weight, yet manual measurements increase workload and self-reporting can introduce bias. We developed and tested a deep-learning-based algorithm to automate the approximation of contrast agent dosage directly from CT scout images. METHODS: We retrospectively analyzed 817 patients undergoing thorax/abdomen CT. Prior to examination, patient weight was collected via manual scale measurements and self-reporting. We developed an EfficientNet convolutional neural network pipeline to estimate weight from scout images and used in-context learning and dataset distillation to analyze body-weight-informative CT features. The model was used in a browser-based user interface to provide dosing estimates for various contrast agent compounds. RESULTS: Self-reported patient weights were statistically significantly lower than manual scale measurements (75.13 kg vs. 77.06 kg; p < 10, Wilcoxon signed-rank test). In 5-fold cross-validation, the pipeline predicted patient weight with a mean absolute error (MAE) of 3.90 ± 0.20 kg. This error corresponds to a difference of roughly 4.48-11.70 ml of contrast agent, depending on the specific agent. Interpretability analysis confirmed that both larger anatomical shape and higher overall attenuation were the predictive features of body weight. CONCLUSIONS: This open-source deep learning pipeline enables automatic, accurate contrast agent dosing in routine CT workflows. The approach has the potential to improve patient safety and clinical efficiency by providing accurate weight estimates without requiring additional measurements or relying on potentially outdated records. Further validation on larger cohorts and across different clinical centers is required.
Pancreatic fat deposition as a mediator of adrenal gland volume and type 2 diabetes: evidence from MRI.
Zhang Q, Zhang Z, Wang S … +4 more , Liu Y, Lin L, Peng J, Liu A
BMC Med Imaging · 2026 Apr · PMID 41965599 · Full text
PURPOSE: To assess the role of body composition mediating the association between MRI-quantified adrenal gland volume (AGV) and type 2 diabetes mellitus (T2DM) and to clarify the underlying mechanisms involved. METHODS:...
PURPOSE: To assess the role of body composition mediating the association between MRI-quantified adrenal gland volume (AGV) and type 2 diabetes mellitus (T2DM) and to clarify the underlying mechanisms involved. METHODS: This retrospective study analyzed 239 hospitalized patients (54 T2DM, 185 non-T2DM) with abdominal MRI. AGV was manually segmented on T1-weighted images. Body composition parameters, including subcutaneous adipose tissue (SAT) area and fat fraction (FF), visceral adipose tissue (VAT) area and FF, hepatic FF (HFF), pancreatic FF (PFF), and abdominal muscle (AM) area and FF, were quantified using MRI fat fraction mapping. RESULTS: After adjusting for age, sex, hypertension, and BMI, mediation analysis showed that PFF partially mediated the association between AGV and T2DM (OR = 1.036, 95% CI: 1.009–1.101), accounting for 8.7% of the total effect. Multivariate analysis identified both AGV (OR: 1.475, 95% CI: 1.235–1.762) and PFF (OR: 1.069, 95% CI: 1.023–1.116) as independent factors associated with T2DM. ROC analysis showed AUCs of 0.792 for AGV and 0.803 for PFF in differentiating T2DM. The combined AGV + PFF model achieved an AUC of 0.836 (95% CI: 0.779–0.894). CONCLUSION: MRI-quantified AGV and PFF demonstrate potential as biomarkers for T2DM. Pancreatic fat deposition partially mediates the link between adrenal enlargement and T2DM, suggesting a novel pathway involving HPA axis dysregulation and ectopic fat deposition.
Evaluation of large language models for VI-RADS reports: a comparative analysis of zero-shot and few-shot prompting.
Halis A, Celiker D
BMC Med Imaging · 2026 Apr · PMID 41965564 · Full text
INTRODUCTION: Accurate preoperative staging is vital in bladder cancer management, particularly for assessing muscle invasion. Multiparametric MRI (mpMRI) combined with the Vesical Imaging–Reporting and Data System (VI-R...
INTRODUCTION: Accurate preoperative staging is vital in bladder cancer management, particularly for assessing muscle invasion. Multiparametric MRI (mpMRI) combined with the Vesical Imaging–Reporting and Data System (VI-RADS) offers a non-invasive and standardized approach for tumor stratification. However, inter-observer variability among radiologists remains a significant limitation. This study investigates the performance of three large language models (LLMs)—ChatGPT (OpenAI; GPT-5.2), Gemini (Google; Gemini 2.0), and Copilot (Microsoft; Copilot based on GPT-5 architecture)—in classifying bladder lesions according to VI-RADS using zero-shot and few-shot prompting strategies. MATERIALS AND METHODS: A synthetic dataset of 100 simulated bladder cancer cases was developed from expert-crafted synthetic radiology reports based on structured text descriptors and expert consensus, comprising 20 cases for each VI-RADS category (1–5). Each case included structured imaging descriptors aligned with VI-RADS scoring criteria. The LLMs were evaluated under zero-shot (no examples provided) and few-shot (with illustrative examples) prompting conditions. Performance metrics included accuracy, macro F1 scores, and Cohen’s kappa, with statistical significance assessed using McNemar’s test. RESULTS: Few-shot prompting significantly improved the classification performance of ChatGPT (OpenAI; GPT-5.2) (accuracy: 94%, F1: 0.939) and Gemini (Google; Gemini 2.0) (accuracy: 83%, F1: 0.823) compared to zero-shot. In contrast, Copilot (Microsoft; Copilot based on GPT-5 architecture)‘s accuracy (53%) and F1 score (0.503) declined under few-shot conditions. ChatGPT (OpenAI; GPT-5.2) demonstrated the highest consistency in identifying high-risk lesions (VI-RADS 4–5), followed by Gemini (Google; Gemini 2.0). CONCLUSION: Few-shot prompting enhances LLM performance in VI-RADS classification, particularly for ChatGPT (OpenAI; GPT-5.2) and Gemini (Google; Gemini 2.0). These findings highlight the potential of AI tools to support radiological decision-making in bladder cancer staging. Further studies using real imaging data and explainable AI are warranted.
AES-SDM meta-analysis of spontaneous brain activity in cerebral small vessel disease.
Li Y, Yang Z, Wang R … +1 more , Xu Z
BMC Med Imaging · 2026 Apr · PMID 41965563 · Full text
BACKGROUND: Cerebral small vessel disease (CSVD) is a major cause of vascular cognitive impairment and dementia, characterized by microvascular dysfunction and disruption of cortico-subcortical circuits. Although resting...
BACKGROUND: Cerebral small vessel disease (CSVD) is a major cause of vascular cognitive impairment and dementia, characterized by microvascular dysfunction and disruption of cortico-subcortical circuits. Although resting-state functional magnetic resonance imaging (rs-fMRI) studies have reported altered spontaneous brain activity in CSVD, their findings remain inconsistent across studies. METHODS: A coordinate-based meta-analysis was performed using Anisotropic Effect-Size Signed Differential Mapping (AES-SDM) to quantitatively synthesize rs-fMRI studies investigating spontaneous neural activity in patients with CSVD. Relevant literature published up to October 19, 2025 was systematically retrieved from PubMed, Embase, Web of Science, China National Knowledge Infrastructure, and Wanfang. Studies reporting between-group differences in local spontaneous brain activity were included. A random-effects model was applied with thresholds of uncorrected p < 0.005, peak height Z > 1, and cluster extent ≥ 5 voxels. Heterogeneity, sensitivity, and publication bias were assessed using Q/I² tests, Jackknife analysis, and Egger’s test, respectively. RESULTS: Nine studies comprising 277 CSVD patients and 287 healthy controls were included. Compared with controls, CSVD patients exhibited increased spontaneous brain activity in the bilateral postcentral gyri, bilateral precuneus, and left inferior frontal gyrus. No brain regions showed decreased activity. Egger’s test indicated no significant publication bias (all p > 0.05), and Jackknife analysis confirmed the stability of results across all iterations. CONCLUSION: This AES-SDM meta-analysis identified convergent increases in spontaneous brain activity in the somatosensory cortex, precuneus, and inferior frontal gyrus in CSVD. These results provide neuroimaging evidence of functional disruption in CSVD and help clarify the neural substrates potentially associated with its clinical manifestations.
Nano-radiomics of molecular MRI for early amyloid associated patterns in an Alzheimer's disease mouse model via an automatic pipeline.
Ngan E, Bhandari P, Parekh PA … +4 more , Badachhape AA, Annapragada AV, Ghaghada KB, Starosolski ZA
BMC Med Imaging · 2026 Apr · PMID 41963831 · Full text
BACKGROUND: The combination of nanoparticle contrast agents with advanced image analysis techniques like radiomics offers a powerful new approach for disease detection and characterization. This study presents a novel au...
BACKGROUND: The combination of nanoparticle contrast agents with advanced image analysis techniques like radiomics offers a powerful new approach for disease detection and characterization. This study presents a novel automated pipeline for segmentation and nano-radiomic analysis of nanoprobe-enhanced images. We demonstrate the effectiveness of this pipeline in an Alzheimer’s disease (AD) mouse model, showing improved detection of amyloid pathology compared to conventional methods. METHODS: The study used the magnetic resonance imaging (MRI) data collected from double transgenic (TG) AD mice, including 13 mice aged 6–8 month old with lower amyloid burden and 36 mice aged 11–18 month old with higher amyloid burden. Wild type (WT) mice served as controls. Three contrast doses were administered to older mice, while one dose was applied to younger mice. A UNet-based model was trained on mouse brain scans to automatically segment two specific regions, and the results were compared to those from semi-automatic segmentation (i.e. manual brain atlas registration to MRIs with automatic region extraction). 89 radiomic features (RFs) per region were computed, followed by genotype classification using machine learning and sensitivity analyses. Image-based conventional metric (percentage change between pre- and post-contrast MRI intensity signal) was compared to radiomics-based approaches. RESULTS: The UNet-based model with SWIN transformer performed the best, achieving Dice coefficients between 0.80 and 0.95. Non-parametric statistical tests showed no significant difference in segmentation performance based on amyloid burden, nanoparticle contrast, or contrast doses. RF-based classification performance was similar between automatic and semi-automatic approaches. In older mice with good-quality segmentation, the automatic approach outperformed the semi-automatic method across all dose-specific test sets (all 1.0) but not the validation set (0.90 vs. 1.0). The top classifier from automatic approach also required 2x less RFs. In younger mice, classification performance varied with segmentation approach and inclusion of fair-quality segmentation. Yet the top classifiers typically required only one or two RFs. CONCLUSIONS: Nano-radiomic analysis of targeted nanoparticle contrast-enhanced MRIs outperformed conventional imaging metrics for early detection of pathological amyloid deposition in an AD mouse model. Automated segmentation achieved performance comparable to the semi-automatic method, while improving efficiency.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe