BMC Medical Imaging[JOURNAL]
Sun
200 papers
Diagnostic performance of saline-aided ultrasound in the detection of colorectal polyps in children.
Wang T, Zhou H, Ding T … +7 more , Li J, Hu L, Yu W, Chen M, Lin Z, Deng J, Zhou L
BMC Med Imaging · 2026 May · PMID 42143287 · Full text
OBJECTIVES: This study aimed to evaluate the diagnostic performance of saline-aided gastroenteric ultrasound (US) for detecting colorectal polyps in pediatric patients. METHODS: In this single-center, retrospective study...
OBJECTIVES: This study aimed to evaluate the diagnostic performance of saline-aided gastroenteric ultrasound (US) for detecting colorectal polyps in pediatric patients. METHODS: In this single-center, retrospective study (April 2018 to November 2023), we analyzed 466 patients with suspected colorectal polyps. All patients underwent conventional US followed by diagnostic colonoscopy. A consecutive paired cohort (n = 292), enrolled after the introduction of a new protocol, additionally received saline-aided US. Diagnostic metrics (accuracy, sensitivity, specificity) and Cohen's kappa were calculated to compare conventional and saline-aided US against histopathological reference standards. Multivariate logistic regression was used to assess clinical parameters affecting diagnostic outcomes. RESULTS: Among 466 recruited patients (mean age 4.83 ± 3.12 years, 276 males), 391 were colonoscopy-positive. The most common pathological type was juvenile intestinal polyp (93.1%); the most frequent location was the rectum (59.3%). Saline-aided US was associated with improved diagnostic performance (accuracy: 81.8%, sensitivity: 81.0%, specificity: 85.0%) compared to conventional US (accuracy: 51.4%, sensitivity: 39.6%). Multivariate analysis identified polyp size and location as factors significantly associated with diagnostic outcome (both p < 0.001). For saline-aided US, smaller polyps (< 10 mm) and those located in the rectum (missed diagnosis rate: 22.8%) were associated with higher rates of missed diagnosis. CONCLUSION: Saline-aided US significantly enhances the detection of colorectal polyps in children, demonstrating superior diagnostic accuracy and sensitivity over conventional US.
Incidence, timing, and risk factors of severe complications after percutaneous ultrasound-guided tumor biopsy: a proposal for safe aftercare based on 30 years´ experience.
Weilert H, Schröder CC, Wohlmuth P … +3 more , König A, Mahnken AH, Stang A
BMC Med Imaging · 2026 May · PMID 42143271 · Full text
BACKGROUND: Percutaneous ultrasound (US)-guided core-needle biopsies (CNBs) is increasingly used to obtain tumor tissue across all anatomic regions to guide personalized cancer treatment; however, data with particular at...
BACKGROUND: Percutaneous ultrasound (US)-guided core-needle biopsies (CNBs) is increasingly used to obtain tumor tissue across all anatomic regions to guide personalized cancer treatment; however, data with particular attention on serious adverse events (SAEs) and appropriate observation times to detect clinically-relevant SAEs are limited. We aimed to assess the incidence, influencing factors, and temporal manifestation of SAEs after US-guided CNB of tumors in the thorax and abdomen to identify patient safety measures with a focus on observation time that will allow safe post-biopsy patient discharge. METHODS: This retrospective review of a prospectively collected single-institutional database included 7.494 patients who underwent percutaneous US-guided automated CNBs of 522 thoracic and 6.972 abdominal tumors between 1992 and 2022. The database was reviewed to identify biopsy-related SAEs of Clavien-Dindo (CD) grade ≥ 2, which were classified as acute (< 6 h), subacute (6-24 h), or delayed (> 24 h). CD and CIRSE ≥ 2 scales were compared for counting biopsy-related SAEs. RESULTS: Among a total of 29 (0.39%) SAEs CD grade ≥ 2, 28% (8/29) were acute, 62% (18/29) subacute, 10% (3/29) delayed, 90% (26/29) bleeding-related, and 10% (3/29) pneumothorax-related. Treatments were erythrocyte transfusion (12x), surgical bleeding control (10x), transarterial coiling (4x), and chest tube placement (3x). Incidences of ≥ 1% were observed for bleeding after spleen (1.6%) and kidney (1.3%) biopsy (other sites: 0-0.4%) and for pneumothorax after transpleural biopsy (1.6%) of intrapulmonal nodules. Bleeding risk associated significantly with INR, number of needle passes, double antiplatelet medication, and, in liver biopsy, with a hypervascular tumor type (p < 0.05). CIRSE classification added 7 prolonged observational stays without treatment (CIRSE grade 2/3), raising the SAE rate from 0.39% to 0.48%. No patient death occurred. CONCLUSIONS: Our results support the safety of the procedure but also highlight potential risks. SAE incidences vary by biopsy site and site-independent factors, and most SAEs occur within 6-24-hours after biopsy, suggesting need for risk-adjusted observation times, either for < 6-hours or 24-hours, to allow safe patient discharge. Further prospective studies are needed to validate our findings and to define reasonable thresholds for low and high SAE risks that could tailor aftercare time in a pragmatic and clinically-relevant sense. CLINICAL TRIAL NUMBER: Not applicable.
MTRFU-Net: a lung nodule segmentation model based on improved U-Net architecture with spatial-frequency fusion.
Wu Y, Li B
BMC Med Imaging · 2026 May · PMID 42143263 · Full text
Accurate pulmonary nodule segmentation is a key step for the early diagnosis of lung cancer, yet existing deep learning methods still have limitations in addressing challenges such as nodule heterogeneity, blurred bounda...
Accurate pulmonary nodule segmentation is a key step for the early diagnosis of lung cancer, yet existing deep learning methods still have limitations in addressing challenges such as nodule heterogeneity, blurred boundaries, and multi-scale variations. To tackle these issues, this study proposes the MTRFU-Net model, a pulmonary nodule segmentation model based on an improved U-Net architecture integrating spatial-frequency features and multi-module collaboration, which adopts a three-tier progressive module collaboration strategy to achieve dynamic spatial-frequency fusion. The encoder of the model is built on ResNet50 and incorporates a Spatial-Frequency Fusion (SFF) module, enabling the parallel extraction and dynamic fusion of dual-domain features. The bottleneck layer combines a Transformer encoder with an optimized Atrous Spatial Pyramid Pooling (ASPP) module, effectively capturing long-range dependencies and multi-scale contextual information. For the decoder, residual connections are paired with a dynamically weighted scSE attention mechanism to enhance the response capability to critical features. Extensive experiments on the LIDC-IDRI dataset demonstrate that MTRFU-Net exhibits excellent performance in terms of the Dice Similarity Coefficient (DSC), mean Intersection over Union (mIoU). This research validates the effectiveness of frequency-domain information in pulmonary nodule segmentation tasks, providing valuable references for the development of robust clinically oriented segmentation models.
A nomogram based on body composition and the prognostic nutritional index to predict early postoperative complications of colorectal cancer.
Zhu N, Liu Y, Yu H … +3 more , Xu J, Wei Y, Zhai J
BMC Med Imaging · 2026 May · PMID 42143259 · Full text
OBJECTIVE: This study aimed to construct a nomogram based on body composition parameters and the prognostic nutritional index (PNI) using quantitative computed tomography (QCT) to predict early postoperative complication...
OBJECTIVE: This study aimed to construct a nomogram based on body composition parameters and the prognostic nutritional index (PNI) using quantitative computed tomography (QCT) to predict early postoperative complications in patients with colorectal cancer (CRC). MATERIALS AND METHODS: We retrospectively analyzed the data of 157 patients who underwent radical resection for CRC between January 2019 and April 2024. All patients underwent QCT 1 month prior to surgery. Body composition was assessed at the level of the third lumbar vertebra, including measurements of the visceral fat area, subcutaneous fat area, and intramuscular fat infiltration (MFI) of the posterior vertebral muscles. The visceral-to-subcutaneous fat ratio (VSR) was calculated. RESULTS: Among the 157 patients, 31 (19.7%) experienced early postoperative complications. Univariate analysis revealed that the PNI, albumin level, VSR, and MFI were significantly associated with these complications. Multivariate logistic regression analysis identified the PNI (odds ratio [OR] = 0.801; 95% confidence interval (CI): 0.653-0.983), VSR (OR = 3.084; 95% CI: 1.365-6.968), and MFI (OR = 1.074; 95% CI: 1.009-1.145) as independent risk factors for early postoperative complications in CRC. The areas under the receiver operating characteristic curves for the PNI, VSR, MFI, and nomogram model for predicting postoperative complications were 0.796, 0.798, 0.648, and 0.879, respectively. Based on these three independent risk factors, the nomogram demonstrated good discrimination, calibration, goodness of fit, and clinical utility. CONCLUSIONS: The nomogram model utilizing QCT-based body composition metrics and the PNI exhibited strong predictive capability for early postoperative complications in patients with CRC.
A multitask Transformer-VAE framework for robust PPG denoising and atrial fibrillation screening.
Zhang L, Yuan L
BMC Med Imaging · 2026 May · PMID 42143257 · Full text
BACKGROUND: Photoplethysmography (PPG) is a noninvasive biosignal widely used for atrial fibrillation (AF) screening via wearable devices. However, reliable PPG-based detection remains challenging because motion artifact...
BACKGROUND: Photoplethysmography (PPG) is a noninvasive biosignal widely used for atrial fibrillation (AF) screening via wearable devices. However, reliable PPG-based detection remains challenging because motion artifacts, noise contamination, and intersubject variability distort waveform morphology and compromise diagnostic consistency. METHODS: We propose an end-to-end dual-task model that denoises PPG signals and detects AF simultaneously. The model uses a transformer-based encoder with two task branches and an additional lightweight alignment constraint to encourage the two tasks to learn consistent representations under motion artifacts. We trained and evaluated the framework on public PPG datasets, including an internal cohort of 30,773 segments from 91 patients and an external dataset derived from the MIMIC-III waveform database. Performance was evaluated via the AUC and accuracy, together with robustness tests across different noise conditions. We also applied an A-Test procedure that repeats balanced k-fold validation with different k values to assess error stability across data splits. RESULTS: The proposed model achieved an AUC of 0.9097 and an accuracy of 88.4%, outperforming conventional convolutional and single-task baselines. Robustness experiments demonstrated stable performance across varying signal‒to‒noise ratios. The dual-task architecture preserved physiologically relevant rhythm features, including heart rate variability. A-Test evaluation indicated stable error behavior across data splits, suggesting reduced partition sensitivity and improved learning consistency. CONCLUSIONS: This Transformer-VAE framework integrates signal reconstruction and diagnostic learning to enable accurate and noise-resilient AF detection from PPG signals. Within a healthcare workflow, the model supports wearable screening by generating AF risk indicators and signal-quality cues that can trigger confirmatory single-lead ECG assessment and appropriate clinical follow-up. TRIAL REGISTRATION: Not applicable.
Evaluating the image quality and brain microstructural exploration of mean apparent propagator-MRI-a pilot study.
Shi Q, Gu Y, Shi D … +9 more , Zhou M, Li H, Zhang H, Li H, Hong J, Fu J, Wang R, Wan X, Liu Z
BMC Med Imaging · 2026 May · PMID 42143250 · Full text
BACKGROUND: To investigate the impact of positivity constraints and Laplacian regularization on Mean Apparent Propagator (MAP)-MRI image quality and effectiveness in analyzing brain microstructural changes, validated usi...
BACKGROUND: To investigate the impact of positivity constraints and Laplacian regularization on Mean Apparent Propagator (MAP)-MRI image quality and effectiveness in analyzing brain microstructural changes, validated using an Intermittent Exotropia (IXT) cohort. METHODS: A total of 31 patients with IXT (16 males, 15 females; mean age 10.94 ± 1.69 years) and 37 age- and sex-matched healthy controls (23 males, 14 females; mean age 11.14 ± 2.06 years) underwent MRI scans. Imaging protocols included simultaneous multi-slice echo planar imaging (SMS-EPI) for q-space sampling and 3D magnetization-prepared rapid acquisition gradient echo (MP-RAGE) sequences. Three MAP-MRI-based reconstruction algorithms were evaluated: MAPL (with Laplacian regularization), CMAP (positivity-constrained), and CMAPL (combining both constraints). Image quality was assessed objectively using wavelet-based signal-to-noise ratio (WSNR), and subjectively by two radiologists using a 5-point scale rating signal uniformity, contrast, noise level, and anatomical completeness. Statistical analyses included one-way ANOVA, Kruskal-Wallis tests, and Cohen's κ for inter-rater agreement. Diagnostic performance was evaluated using voxel-wise t-tests with Gaussian Random Field (GRF) correction at P = 0.01. RESULTS: MAPL outperformed CMAP and CMAPL in both objective and subjective image quality, particularly in RTOP, RTAP, MSD, and QIV maps (all P < 0.05). It also provided a more comprehensive visualization of changes in the bilateral dorsal visual pathways, including BA17, BA18, BA19, BA39, BA40, BA8, BA6, and the supplementary motor area (SMA). Moreover, MAPL demonstrated superior sensitivity in detecting alterations in brain regions associated with language, emotion, and decision-making (BA10, BA11, BA22, BA37, and BA47), which may reflect secondary changes induced by visual impairment. CONCLUSIONS: MAPL surpasses CMAP and CMAPL in both image quality and the ability to display brain microstructural changes, establishing it as the superior method for exploring brain structural alterations. CLINICAL TRIAL INFORMATION: It was registered on July 18, 2021 (ChiCTR2100048852) at www.chictr.org.cn and commenced on September 1, 2021.
Evaluation of cartilaginous endplate degeneration with histogram features of multiple parameters in UTE MRI.
Miao J, Li C, Huang B … +1 more , Ren YX
BMC Med Imaging · 2026 May · PMID 42141391 · Full text
OBJECTIVE: The cartilaginous endplate (CEP) exhibits an intrinsically short [Formula: see text] model based on ultrashort echo time (UTE) sequence, and to explore the diagnostic utility of bi-exponential [Formula: see te...
OBJECTIVE: The cartilaginous endplate (CEP) exhibits an intrinsically short [Formula: see text] model based on ultrashort echo time (UTE) sequence, and to explore the diagnostic utility of bi-exponential [Formula: see text]-derived parameters and monoexponential UTE-[Formula: see text] mapping values in grading CEP damage. METHODS: This study retrospectively collected 46 patients who underwent lumber UTE MRI, of whom 43 met the inclusion criteria. CEP was graded into three groups according to morphological features on multiple MRI sequences: healthy (structurally intact), mild damage (localized thinning or concavity with preserved continuity), and moderately damage (defects < 50% with disrupted continuity). Multiple quantities were evaluated using the UTE sequence on the CEP manually drawn by an experienced radiologist. One-way Kruskal-Wallis test was used to inspect the distribution differences among groups. Logistic regression and support vector machine models were applied to predict the level of degeneration with a considerably good precision, and receiver operating characteristic curves suggests a distinguishable performance among those models. RESULTS: [Formula: see text] values showed significant differences among the groups (p < 0.05), with Tukey's test indicating the most significant difference between the moderately damaged group and the healthy controls. [Formula: see text] values were non-normally distributed but statistically different between the moderate and mild damage groups (p < 0.05), a trend also observed in monoexponential UTE-[Formula: see text] mapping values (p < 0.05). The logistic regression and SVM models performed well in identifying moderate damage (AUC of 0.878 and 0.858, respectively), but had limited ability to detect mild damage (AUC of 0.718 and 0.729). CONCLUSION: The UTE bi-exponential [Formula: see text] model enable effective separation and quantification of distinct water components within the CEP. Both [Formula: see text] and UTE-[Formula: see text] mapping show promise as imaging biomarkers for grading CEP degeneration even in the early stage of CEP degeneration.
Multiparametric MRI findings, PI-RADS scores, and gleason scores in prostate cancer patients at Kumasi, Ghana: a cross-sectional analysis.
Ampah BA, Badu-Peprah A, Amankwa AT … +7 more , Otoo OK, Twum KA, Anyitey-Kokor IC, Afreh YA, Ayisi-Boateng NK, Addae M, Quarshie F
BMC Med Imaging · 2026 May · PMID 42135698 · Full text
BACKGROUND: Prostate cancer is a major cause of cancer-related morbidity and mortality in men, affecting about 2.3 million individuals every five years globally. Evaluating the diagnostic performance of multiparametric M...
BACKGROUND: Prostate cancer is a major cause of cancer-related morbidity and mortality in men, affecting about 2.3 million individuals every five years globally. Evaluating the diagnostic performance of multiparametric MRI (mpMRI) and the Prostate Imaging Reporting and Data System (PI-RADS) against histopathological grading is crucial for optimal management. This study examined the relationship between mpMRI findings, PI-RADS scores, and Gleason scores in patients with prostate cancer. METHODS: In this cross-sectional study, 76 men with prostate cancer underwent mpMRI. Data were analyzed using STATA and R software. Spearman's rank correlation was employed to assess associations between PI-RADS and Gleason scores. Receiver operating characteristic (ROC) analysis was performed to evaluate the ability of imaging parameters to stratify tumor aggressiveness within a cancer-positive cohort. One-way ANOVA tested differences in mean apparent diffusion coefficient (ADC) values across Gleason score and International Society of Urological Pathology (ISUP) grade groups, with p ≤ 0.05 considered statistically significant. RESULTS: A weak but statistically significant positive correlation was observed between PI-RADS and Gleason scores (r = 0.31, p = 0.007). PI-RADS 5 was predominantly associated with Gleason scores 7, 8, and 9, with proportions of 78%, 76%, and 100%, respectively (p < 0.001). For high-grade disease (Gleason score > 7), PI-RADS 5 demonstrated high sensitivity (86.2%) but low specificity (27.7%), consistent with its strength in capturing aggressive tumors within a biopsy-confirmed cohort. ADC values showed a moderate inverse correlation with Gleason grades (r = - 0.47, p < 0.001) and ISUP grades (r = - 0.43, p = 0.003), with progressively lower ADC values in higher-grade tumors. CONCLUSIONS: Higher PI-RADS categories and lower ADC values were associated with increasing histopathological aggressiveness in this cohort of biopsy-proven prostate cancer patients. These findings support the potential role of mpMRI in disease stratification, but should be interpreted cautiously given the cancer-only study population and skewed PI-RADS distribution. Larger prospective studies are required to validate these findings in mixed-suspicion cohorts.
Development and validation of a novel deep learning coronary artery plaque quantification model.
Malmqvist J, Jernberg T, Vedad R … +1 more , Wang C
BMC Med Imaging · 2026 May · PMID 42135689 · Full text
BACKGROUND: Coronary Computed Tomography Angiography (CCTA) is an established tool for assessing coronary artery disease. CCTA determined total coronary artery plaque volume predicts cardiovascular events, but manual qua...
BACKGROUND: Coronary Computed Tomography Angiography (CCTA) is an established tool for assessing coronary artery disease. CCTA determined total coronary artery plaque volume predicts cardiovascular events, but manual quantification is impractical for routine use. Deep learning-based methods offer a promising solution for automated plaque volume assessment. METHODS: We developed and validated a novel software, QuantiPlaque, powered by deep learning models, for automated segmentation of total and calcified coronary plaque volume. Expert manual annotation served as the reference standard. Correlations between deep learning and expert segmentation were evaluated using intraclass correlation coefficient (ICC), Pearson's r, and Spearman's rho, and agreements were evaluated with Bland-Altman analyses. RESULTS: A total of 115 CCTA scans were included. Mean (range) age was 58 (51-64) and 51% were females. The mean coronary artery calcium score was 56 Agatston units, ranging from 0 to 601. The model demonstrated strong correlation and agreement with expert annotation for per-patient total plaque volume (ICC 0.95, mean difference - 8.35 mm, 95% limit of agreement - 102 to 85 mm, Spearman's rho 0.87, Pearson's r 0.95) and calcified plaque volume (ICC 0.90, mean difference - 0.28 mm, 95% limits of agreement - 21.97 to 21.42 mm, Spearman's rho 0.93, Pearson's r 0.90). Per-vessel analysis showed strong correlation in the left anterior descending artery and right coronary artery but was weaker in the left circumflex artery territory. CONCLUSION: The evaluated deep learning model provides accurate quantification of total and calcified plaque burden, with strong correlation and agreement to expert annotation.
Perilesional radiomics enhances 90-day outcome prediction in intracerebral hemorrhage: development and validation of a combined model with hematoma radiomics and clinical features.
Zhang K, Yang B, Yan T … +4 more , Wang K, Tan M, Wang P, Wang Z
BMC Med Imaging · 2026 May · PMID 42135677 · Full text
OBJECTIVES: This study aimed to develop and validate a prognostic model integrating hematoma (R1), perilesional (R2), and clinical features to predict 90-day outcomes. METHODS: A total of 759 ICH patients from two center...
OBJECTIVES: This study aimed to develop and validate a prognostic model integrating hematoma (R1), perilesional (R2), and clinical features to predict 90-day outcomes. METHODS: A total of 759 ICH patients from two centers were enrolled and allocated to training, internal validation, and external test sets. The primary endpoint was a poor 90-day outcome, defined as a modified Rankin Scale (mRS) score > 3. Independent clinical risk factors were identified via univariate and multivariate logistic regression analyses. Subsequently, seven prognostic models were constructed using R1, R2, clinical features, and their combinations. Model discrimination was compared using the DeLong test for the Area Under the Curve (AUC). Calibration and clinical utility were evaluated using calibration curves and Decision Curve Analysis (DCA). RESULTS: Multivariate analysis identified four independent risk factors for poor outcome: hematoma volume (OR 1.042; 95% CI 1.025-1.058; p < 0.001), mean hematoma density (OR 0.916; 95% CI 0.863-0.973; p = 0.004), age (OR 1.078; 95% CI 1.054-1.103; p < 0.001), and admission Glasgow Coma Scale (GCS) score (OR 0.777; 95% CI 0.708-0.853; p < 0.001). Among the seven models constructed, the tri-combined model (R1 + R2+Clinical) demonstrated the most stable and relatively better performance across all datasets, with an AUC of 0.791 (95% CI: 0.716-0.867) in the external test set. This model exhibited good calibration and favorable statistical net benefit on DCA. CONCLUSION: The integrated prognostic model combining hematoma and perilesional radiomic features with clinical data provides stable and incremental prognostic value for 90-day functional outcomes in patients with ICH.
MEP-Net: a PIDNet-based model with median-enhanced spatial-channel attention for segmentation of hepatocellular carcinoma in CEUS images.
Yang SH, Wen JB, Li YR … +3 more , Zhang FF, Li WQ, Cheng SQ
BMC Med Imaging · 2026 May · PMID 42129688 · Full text
PURPOSE: Hepatocellular carcinoma (HCC) remains a major global health concern due to its high incidence and mortality. Contrast-enhanced ultrasound (CEUS) offers notable advantages for HCC diagnosis, including real-time...
PURPOSE: Hepatocellular carcinoma (HCC) remains a major global health concern due to its high incidence and mortality. Contrast-enhanced ultrasound (CEUS) offers notable advantages for HCC diagnosis, including real-time imaging and non-invasiveness. However, CEUS images are often affected by noise, uneven signal distribution, and unclear boundaries between lesions and surrounding tissues, which pose challenges for automatic lesion segmentation. METHODS: To improve segmentation performance, this study proposes an improved PIDNet-based model termed MEP-Net. The model integrates a median-enhanced spatial-channel attention mechanism (MECS) and an efficient channel attention (ECA) module to enhance lesion-related feature representation and multi-branch feature fusion. We evaluate the model on a self-built CEUS dataset and the publicly available BUSI breast ultrasound dataset. It is also compared with several mainstream semantic segmentation methods, and ablation studies are conducted to analyze the contribution of each module. RESULTS: The results show that MEP-Net outperforms the baseline PIDNet by 1.95%, 1.25%, and 2.51% in Dice, MIoU, and Recall, respectively, on the CEUS dataset, and by 1.37%, 1.06%, and 2.41% on the BUSI dataset. In addition, MEP-Net is compared with eight semantic segmentation methods and demonstrates superior overall segmentation performance and improved lesion representation. Ablation studies further confirm the complementary benefits of the MECS and ECA modules in improving segmentation accuracy. CONCLUSION: The proposed MEP-Net achieves improved performance in CEUS image segmentation. By introducing attention mechanisms tailored to ultrasound image characteristics, it provides an effective approach for automatic HCC lesion segmentation.
Application of a multimodal MRI model integrating radiomics and habitat features for predicting glioma pathology and prognosis.
Sun L, Yang Y, Cao Z … +10 more , Yang D, Du N, Lu Y, Yan M, Li J, Shi F, Zhou X, Fang X, Lin G, Li S
BMC Med Imaging · 2026 May · PMID 42121116 · Full text
BACKGROUND: Accurate grading and prognostic assessment of glioma requires integrating key molecular biomarkers, including IDH mutation status and the Ki-67 proliferation index. However, current radiomics studies often fo...
BACKGROUND: Accurate grading and prognostic assessment of glioma requires integrating key molecular biomarkers, including IDH mutation status and the Ki-67 proliferation index. However, current radiomics studies often focus on single-task predictions and rely on manual tumor segmentation, which fails to capture intratumoral spatial heterogeneity. This study proposes an automated whole-tumor segmentation-based multimodal MRI approach integrating habitat radiomics to achieve noninvasive, multitask prediction of WHO grade, IDH mutation, Ki-67 labeling index (LI), and 2-year postoperative survival in glioma. METHODS: This retrospective study enrolled 185 patients with pathologically confirmed glioma. Preoperative multimodal MRI - including T1-weighted imaging (T1WI), T2-weighted fluid-attenuated inversion recovery (T2W-FLAIR), and T1-weighted contrast-enhanced imaging (T1W CE) - was acquired for analysis. Using the uAI Research Portal platform, we performed automated whole-tumor segmentation and subsequent feature extraction, deriving 2,264 radiomics features and 61 habitat-based features. Predictive models were developed using multiple machine learning algorithms, and feature selection was rigorously performed within the training folds of a five-fold cross-validation to prevent overfitting. Model performance was evaluated using AUC, accuracy, sensitivity, and specificity, with statistical comparisons conducted performed DeLong's test. RESULTS: The habitat model exhibited superior sensitivity in capturing tumor heterogeneity across all four prediction tasks. Building on this, the integrated model combining habitat and conventional radiomics features, achieved the highest overall predictive performance, with AUCs of 0.916 (95% CIs: 0.858-0.975) for glioma grading, 0.877 (95% CIs: 0.828-0.926) for IDH mutation status, 0.859 (95% CIs: 0.788-0.930) for Ki-67 LI, and 0.906 (95% CIs: 0.837-0.974) for 2-year survival prediction, consistently outperforming single-modality models. SHAP interpretability analysis revealed that patient age exhibited strong correlation with tumor grade, IDH mutation status, and Ki-67 LI. Furthermore, tumor grade, IDH status, and Ki-67 LI demonstrated potential predictive value for 2-year postoperative survival. CONCLUSIONS: The automated habitat radiomics framework effectively quantified intratumoral spatial heterogeneity in glioma. When combined with conventional radiomics, it significantly enhanced accuracy in predicting key molecular and clinical endpoints.
Deep residual network fusing CT images and clinical variables to predict lung adenocarcinoma aggressiveness.
Peng J, Zhong W, Li K … +7 more , Zhang L, Huang D, Hong J, Liu X, Zou Y, Liu X, Tang B
BMC Med Imaging · 2026 May · PMID 42121095 · Full text
BACKGROUND: Lung adenocarcinoma presenting as ground-glass nodules (GGNs) comprises three invasive subtypes (adenocarcinoma in situ [AIS], minimally invasive adenocarcinoma [MIA], invasive adenocarcinoma [IAC]) with dist...
BACKGROUND: Lung adenocarcinoma presenting as ground-glass nodules (GGNs) comprises three invasive subtypes (adenocarcinoma in situ [AIS], minimally invasive adenocarcinoma [MIA], invasive adenocarcinoma [IAC]) with distinct prognoses and management strategies. Preoperative discrimination of these subtypes remains challenging for radiologists, and existing deep learning models rarely integrate multi-modal data for reliable prediction. PURPOSE: This study aimed to develop and internally validate a multi-modal fusion framework based on the standard ResNet50 architecture, integrating CT images, clinical variables, and tumor markers, to improve the preoperative prediction of ground-glass nodule invasiveness. METHODS: A retrospective study was conducted including 431 patients with pathologically confirmed ground-glass nodules. All patients underwent standard chest computed tomography before surgery. A multi-modal deep learning model was constructed based on the ResNet50 network, combined with clinical characteristics and laboratory indicators. Model performance was evaluated using accuracy, area under the receiver operating characteristic curve, precision, recall, and F1-score with five-fold cross-validation. RESULTS: The proposed multi-modal model achieved an overall accuracy of 72.2%, precision of 95.6%, negative predictive value of 96.0%, weighted F1-score of 40.0%, and multiclass Matthews correlation coefficient of 73.1% in the three-class classification of AIS, MIA, and IAC. Per-class analysis showed precision of 84.6%, 35.7%, and 84.4% and recall of 57.9%, 29.4%, and 81.8% for AIS, MIA, and IAC, respectively. The fusion model yielded a macro-average AUC of 0.87, which was higher than the CT-only model (0.79) and both the senior (0.67) and junior radiologists (0.57). The model demonstrated superior diagnostic performance compared to human readers, particularly for the challenging MIA subtype. CONCLUSION: This multi-modal deep learning model combining CT images, clinical variables, and serum tumor markers enables accurate and robust three-class classification of AIS, MIA, and IAC in ground-glass nodules. The proposed model outperforms both human radiologists and the imaging-only model, suggesting its potential as a reliable auxiliary tool to improve preoperative prediction of lung adenocarcinoma invasiveness and assist clinical decision-making.
Radiomics model for risk stratification of intracranial aneurysm: a high-resolution vessel wall imaging-based study.
Zhou Z, Wu Q, Deng Y … +6 more , Wang Y, Jiang S, Yan C, Yuan W, Wang Z, Chen T
BMC Med Imaging · 2026 May · PMID 42120992 · Full text
BACKGROUND: High-resolution vessel wall imaging (HR-VWI) enables in vivo assessment of aneurysm wall pathology, but conventional evaluation remains largely qualitative. This study aimed to develop and validate an HR-VWI-...
BACKGROUND: High-resolution vessel wall imaging (HR-VWI) enables in vivo assessment of aneurysm wall pathology, but conventional evaluation remains largely qualitative. This study aimed to develop and validate an HR-VWI-based radiomics model using aneurysm wall and parent artery wall features to identify symptomatic intracranial aneurysms (SIAs) and improve risk stratification. METHODS: A total of 410 patients with 446 intracranial aneurysms (IAs), comprising symptomatic (n = 112) and asymptomatic (n = 334) intracranial aneurysms from two centers, were included in this retrospective study. HR-VWI images were preprocessed to extract regions of interest from both the aneurysm wall and the parent artery (PA). Radiomic features were subsequently extracted using Pyradiomics, yielding a comprehensive set of 851 features per ROI. Feature selection was performed through a multi-stage process involving variance analysis, independent t-tests, and ElasticNet regularization. Based on these selected features, three imaging models were developed, including Radscore_IA (for the IA), Radscore_PA (for the PA), and Radscore_IA_PA (a combined model). Moreover, aneurysm location was incorporated as a morphological parameter into the refined models: Radscore_LOC_IA, Radscore_LOC_PA, and Radscore_LOC_IA_PA. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), with the PHASES score serving as a comparative benchmark. RESULTS: Radiomics features derived from the IA (n = 7) and PA (n = 3) associated with SIAs were identified. In the validation cohort, the AUC values for SIA identification were as follows: PHASES (0.679), Radscore_IA (0.837), Radscore_PA (0.820), Radscore_IA_PA (0.878), Radscore_LOC_IA (0.852), Radscore_LOC_PA (0.842), and Radscore_LOC_IA_PA (0.888). The Radscore_LOC_IA_PA model exhibited the best performance, outperforming other models. Calibration and decision curve analyses confirmed the robustness and clinical application of all developed models. CONCLUSIONS: This study presents an innovative HR-VWI radiomics-based model for identifying high-risk SIA, Radscore_LOC_IA_PA, which integrates radiomics features from the IA wall and PA wall along with aneurysm location. Compared to traditional stratification methods, the model exhibits showed improved discrimination to identify high-risk SIA, enabling more accurate risk stratification and clinical management strategies for this patient population.
Styloid process morphology in extracranial internal carotid artery dissection: a CTA case-control study.
Zhang X, Shen X, Hu S … +9 more , Shen Y, Cao Z, Wu Y, Hu D, Cui M, Zhai D, Shi D, Cai W, Ju S
BMC Med Imaging · 2026 May · PMID 42115977 · Full text
BACKGROUND: To evaluate the association between styloid process (SP) morphology derived from CTA and extracranial internal carotid artery dissection (ICAD). MATERIALS AND METHODS: This retrospective case-control study in...
BACKGROUND: To evaluate the association between styloid process (SP) morphology derived from CTA and extracranial internal carotid artery dissection (ICAD). MATERIALS AND METHODS: This retrospective case-control study included 85 patients with unilateral extracranial ICAD and 85 frequency-matched controls who underwent head and neck CTA. Two neuroradiologists independently measured SP length, the distance from the SP tip to the internal carotid artery (SPT-ICA distance), and SP orientation (medial and anterior inclination angles). Group comparisons were performed between the dissection side in the ICAD group and the corresponding matched side in controls. Multivariable logistic regression was used to identify independent morphologic factors associated with ICAD, and ROC analysis was performed to evaluate the diagnostic performance of each individual parameter and of the combined model. RESULTS: Compared with controls, ICAD showed a longer SP, a shorter SPT-ICA distance, and a smaller anterior inclination angle (all p < 0.001). Multivariable logistic regression identified SP length, SPT-ICA distance, and anterior inclination angle as independent factors associated with ICAD. ROC analysis showed that the combined model achieved the best performance (AUC = 0.836), compared with SP length (AUC = 0.752), SPT-ICA distance (AUC = 0.742), and anterior inclination angle (AUC = 0.717). CONCLUSION: The combined model based on SP length, SPT-ICA distance, and anterior inclination angle showed the best performance, suggesting that these CTA-derived SP morphologic features may help identify an anatomic pattern associated with extracranial ICAD.
A longitudinal analysis of T1-weighted MRI features associated with progression from mild cognitive impairment to Alzheimer's disease.
Wang Y, Song W, Yang X … +9 more , Meng W, Ma Y, Wang A, Guo G, Zhang Z, Li Z, Han H, Wang S, Shi F
BMC Med Imaging · 2026 May · PMID 42115951 · Full text
OBJECTIVE: Our study aimed to systematically identify T1-weighted MRI-derived brain structural features associated with progression from mild cognitive impairment (MCI) to Alzheimer's disease (AD). METHODS: We utilized t...
OBJECTIVE: Our study aimed to systematically identify T1-weighted MRI-derived brain structural features associated with progression from mild cognitive impairment (MCI) to Alzheimer's disease (AD). METHODS: We utilized the data from the Alzheimer's Disease Neuroimaging Initiative (ADNI). A total of 947 participants with MCI at baseline were included. All participants underwent a neuropsychological assessment and clinical diagnosis every 6 months. The longest follow-up period was 15.5 years, with a median follow-up time of 3.0 years (range: 0-15.5 years). During the follow-up period, 314 (33.16%) individuals progressed to AD, while 633 (66.84%) remained stable or reverted to normal cognition. After ComBat-based harmonization to reduce scanner-related batch effects, 192 high-dimensional T1-magnetic resonance imaging (MRI)-derived morphometric and volumetric measures were analyzed. To identify the characteristics associated with AD progression from MCI, we employed a comprehensive analysis framework. Firstly, we applied the penalized generalized estimating equations (PGEE) and mixed effects random forest (MERF) models for feature screening. Based on the union features obtained from these two methods, a high-dimensional joint model (HDJM) was further used to select the key brain structural features. Lastly, a multivariate joint model was employed to capture the influence of the longitudinal MRI trajectories on the MCI-to-AD conversion. RESULTS: We identified 8 brain structural features from 192 MRI features that were associated with the risk of MCI progressing to AD, including: left hippocampus, left inferior lateral ventricle, left amygdala, right middle temporal gyrus, left fusiform gyrus, right amygdala, left cortical total volume, and brain parenchyma total volume. In the multivariate joint model, the atrophy of left hippocampus volume (α = -0.0009, P = 0.0010), the expansion of left lateral inferior ventricle volume (α = 0.0003, P = 0.0238), the atrophy of right middle temporal gyrus volume (α = -0.0002, P = 0.0036), and the accelerated atrophy of brain parenchyma total volume (α = 0.000005, P = 0.0008) were all significantly associated with the risk of disease transformation. Additionally, the covariate APOE ε4 allele remained a significant independent risk factor (γ = 0.6930, P < 0.0001). CONCLUSION: Left hippocampal atrophy, left inferior lateral ventricular enlargement, right middle temporal gyrus atrophy, and brain parenchyma total volume atrophy were independently associated with the risk of progression from MCI to AD, alongside the established genetic risk factor APOE ε4.
Anatomical variations of lamina papyracea, internal carotid artery, and optic canal dehiscence and protrusion: a computed tomography retrospective cohort study.
Tuysuz O, Tekin S, Dablan A … +2 more , Mutlu SZ, Karagulle M
BMC Med Imaging · 2026 May · PMID 42106695 · Full text
BACKGROUND: Anatomical variations of the paranasal sinuses, particularly those involving the lamina papyracea (LP), internal carotid artery (ICA), and optic canal (OC), may significantly increase the risk of complication...
BACKGROUND: Anatomical variations of the paranasal sinuses, particularly those involving the lamina papyracea (LP), internal carotid artery (ICA), and optic canal (OC), may significantly increase the risk of complications during functional endoscopic sinus surgery (FESS). High-resolution computed tomography (CT) plays a critical role in identifying these variations preoperatively. In this context, the objective of the present study is to determine the prevalence and radiological characteristics of LP dehiscence, ICA protrusion/dehiscence, and optic canal protrusion/dehiscence in a large CT cohort. METHODS: This retrospective study included 3052 patients who underwent paranasal CT between June 2020 and October 2025. Patients younger than 18 years, those with maxillofacial trauma, or with prior sinonasal surgery were excluded. All CT scans were acquired using a 128-slice multidetector system. Two blinded radiologists evaluated LP, ICA, and OC variations; LP dehiscence was classified into three grades. ICA and optic canal protrusion were graded as < 50% or ≥ 50% based on the extent of canal invagination into the sphenoid sinus cavity. Statistical analyses were performed with a significance level of p < 0.05. RESULTS: LP dehiscence was identified in 70 patients (2.3%), more common in males (p = 0.046). Most cases were unilateral (95.7%) and located in the anterior ethmoid cells (75.7%). Grade 1 was the most common type (70%). ICA protrusion was detected in 241 patients (7.9%), with 50.6% being bilateral. ICA dehiscence occurred in 24 patients (0.8%), with no significant right-left difference. OC protrusion was present in 544 patients (17.8%), nearly half bilateral (46.3%). OC dehiscence was identified in 101 patients (3.3%). Sex-related differences were observed in the laterality distribution of lamina papyracea dehiscence and optic canal variants, whereas ICA laterality did not differ significantly. CONCLUSIONS: LP dehiscence, ICA protrusion/dehiscence, and OC protrusion/dehiscence are clinically relevant variations that must be recognized prior to FESS. Their detection on CT is essential for surgical planning, as failure to identify these high-risk anatomical features may lead to serious complications such as optic nerve injury or ICA hemorrhage. Routine and systematic CT evaluation should therefore be considered mandatory in all sinonasal surgical candidates. TRIAL REGISTRATION: Not applicable.
A nomogram combined substantia nigra hyperechogenicity with third ventricular width assessed by transcranial sonography in prediction Parkinson's disease-related cognitive impairment.
Yang QH, Hou C, Li MX
BMC Med Imaging · 2026 May · PMID 42106687 · Full text
OBJECTIVE: To develop a nomogram that incorporating clinical data and transcranial sonography (TCS) markers for predicting Parkinson's disease patients with cognitive impairment (PD-CI). METHODS: 149 PD with normal cogni...
OBJECTIVE: To develop a nomogram that incorporating clinical data and transcranial sonography (TCS) markers for predicting Parkinson's disease patients with cognitive impairment (PD-CI). METHODS: 149 PD with normal cognition (PD-NCI), 117 PD with mild CI (PD-MCI), and 79 PD with dementia (PDD) were included as the training set, 145 PD patients and 154 age- and gender-matched volunteers were enrolled as the test set and control group, respectively. Logistic regression was utilized to screen risk factors for predicting PD-CI, and a nomogram was generated. RESULTS: A predictive model was developed using age, education level, homocysteine, substantia nigra hyperechogenicity (SNH), and third ventricle (V3) width. Receiver operating characteristic curves indicated that V3 width, homocysteine, and the model effectively distinguished between PD-NCI and PD-CI, PDD and non-PDD, as well as PDD and PD-MCI, with the predictive model yielding the highest area under the curve. Calibration curves showed that the predictions of the final model in both the training and test sets closely matched the actual probabilities, while clinical decision curves suggested that the nomogram provided a substantial clinical net benefit. CONCLUSION: A predictive model for PD-CI that incorporates age, education level, plasma homocysteine levels, SNH, and V3 width has been developed and validated.
A human-AI collaborative workflow for improved gross tumor volume delineation in esophageal cancer radiotherapy.
Sun H, Li J, Liu Y … +11 more , Meng F, An Z, Cui Y, Lv B, Chai G, Jia L, Shi Z, Liu M, Zhu B, Gong J, Zhao L
BMC Med Imaging · 2026 May · PMID 42106675 · Full text
BACKGROUND: Esophageal cancer tumors exhibit complex and variable distribution. Due to differences in clinical experience, junior oncologists often show less accuracy in gross tumor volume (GTV) delineation compared to t...
BACKGROUND: Esophageal cancer tumors exhibit complex and variable distribution. Due to differences in clinical experience, junior oncologists often show less accuracy in gross tumor volume (GTV) delineation compared to their senior counterparts. Deep learning methods have the potential to assist junior oncologists in improving the accuracy of GTV delineation. This study proposes a human-AI collaborative workflow for esophageal cancer radiotherapy and explores its clinical feasibility. METHODS: A retrospective collection of 730 esophageal cancer radiotherapy cases from our center was divided into training and testing sets at a 4:1 ratio. The human-AI collaborative workflow involves the following steps: (1) developing an automatic GTV delineation model based on planning CT images and obtaining an AI-prompt; (2) junior oncologists manually refining the GTVs generated by the model to serve as human-guided prompt; (3) using both prompts in combination with planning CT images to build two automatic refinement models; and (4) junior oncologists comparing the refined GTVs obtained by the above models, ultimately submitting the finalized GTVs for expert review by senior oncologists. Three junior oncologists from our center independently completed the manual tasks in steps 2 and 4, and the GTVs derived from the new approach were compared for accuracy against the ground truth delineated by senior oncologists. RESULTS: The results from the testing set indicated that after manual refinement by the three junior oncologists, the Dice similarity coefficient (DSC) between the automatic GTV delineation (step 1) and the ground truth increased from an initial value of 0.6538 ± 0.2022 to 0.7109 ± 0.1958, 0.7253 ± 0.1632, and 0.7236 ± 0.1965. Furthermore, by combining manual refinement, automatic refinement, and decision-making, the accuracy of the GTV delineation further improved to 0.7552 ± 0.1366, 0.7671 ± 0.1190, and 0.7757 ± 0.1354. These results demonstrate that the new approach significantly enhances the accuracy and stability of esophageal cancer GTV delineation. CONCLUSION: With the support of the human-AI collaborative workflow, junior oncologists can delineate esophageal cancer GTV contours with improved accuracy, providing a reliable foundation for the integration of AI in clinical radiotherapy. The study also demonstrates that, in the current trend toward radiotherapy automation, clinical oncologists remain an indispensable and vital part of the process.
Can incorporating preoperative arterial enhancement ratio map into habitat analysis predict TACE refractoriness in unresectable hepatocellular carcinoma?
Xiao Y, Song W, Liu H … +3 more , Yang Q, Zhou J, Guo D
BMC Med Imaging · 2026 May · PMID 42106669 · Full text
PURPOSE: To investigate the value of applying preoperative magnetic resonance imaging (MRI)-based arterial enhancement ratio (AER) map to habitat analysis for predicting transarterial chemoembolization (TACE) refractorin...
PURPOSE: To investigate the value of applying preoperative magnetic resonance imaging (MRI)-based arterial enhancement ratio (AER) map to habitat analysis for predicting transarterial chemoembolization (TACE) refractoriness in patients with unresectable hepatocellular carcinoma (HCC). METHODS: 176 patients with unresectable HCC who underwent preoperative contrast-enhanced MRI and received consecutive TACE treatments were randomly allocated to a training cohort (n = 124) and a validation cohort (n = 52). A dual-mode encoding strategy (mode 1: based on T1-pre; mode 2: based on T1-pre with the AER map) was employed to define tumor habitats. Signal intensity, entropy, and volume fraction were quantified for each habitat. Combined with laboratory findings and imaging features, logistic regression analyses were performed to identify independent risk factors for TACE refractoriness and a predictive model was constructed. The performance of the model was evaluated using the area under the receiver operating characteristic curve (AUC). RESULTS: Each tumor lesion was segmented into 5 habitats using mode 1 and 3 habitats using mode 2. Regression analysis identified Barcelona Clinic Liver Cancer stage (OR = 2.099, P = 0.027), blood products in mass (OR = 11.063, P = 0.028), and AER entropy value of habitat 2 derived from mode 2 (OR = 4.586, P = 0.033) as independent factors for predicting TACE refractoriness. AUC value of the nomogram on validation cohort was 0.721 (95% CI = 0.577-0.866). Calibration curve demonstrated favorable clinical applicability of the model. CONCLUSION: Incorporation of the AER map into habitat analysis enables preoperative prediction of TACE refractoriness risk in patients with unresectable HCC. CLINICAL TRIAL NUMBER: Not applicable.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe