Objectives:To discover a model for accurate risk stratification of intracerebral hemorrhage (ICH) patient outcomes across multiple time windows. Materials and methods:This retrospective study enrolled ICH patients with onset-to-imaging time (OIT) < 72 h. Patients were divided into three groups, 1-3: OIT< 6 h, 6-24 h, 24-72 h. Group 1 patients with preoperative reimaging within 72 h formed Group 4. The 90-day mRS score served as the endpoint (0-3: favorable prognosis; 4-6: poor prognosis). Binary logistic regression was used to build prognostic models for each group. The generalizability of the Group 1 model was validated across other subgroups. Results:A total of 2136 patients with ICH were included in the study. In the training， internal and external validation set, the AUC values achieved when the Group 1 radiomics model was assessed in Group 2 patients were 0.773, 0.759 and 0.706. The AUC values observed when the Group 1 combined model was evaluated in Group 3 patients were 0.811, 0.847 and 0.944. In the training set, the AUC obtained when the Group 1 radiomics model was tested in Group 4 patients was 0.779. The AUC of the radiomics model constructed by combining the key radiomic features of Groups 1 and 4 was 0.788. The AUC of the independent radiomics model for Group 4 was 0.815. Conclusion:Prognostic models for ICH patients with OIT < 6 h may also be generalizable to those with OIT < 72 h, enabling reliable early outcome assessment across different time windows.

To assess a deep learning (DL) model using portal-venous phase CT for discriminating colorectal cancer liver metastasis (CRLMs) and hemangiomas (HMs). Colorectal cancer (CRC) patients diagnosed with CRLMs or HMs at two medical centers from January 2018 and April 2024 were retrospectively included. Lesions were automatically segmented using TotalSegmentator. DL models, DenseNet-201 and ResNet-152, were trained to classify CRLMs and HMs. Their performance, measured by AUC, was evaluated on validation and test sets. Subgroup analyses were conducted for lesions ≤ 10 mm (subcentimeter) and 10–30 mm. Radiologists’ diagnostic performance with and without DL assistance was compared using a multi-reader multi-case analysis. 534 CRLMs (134 CRC-patients; median, 60 years) and 262 HMs (154 CRC-patients; median, 62 years) were divided into the training, validation and test set. The Dice coefficients of TotalSegmentor for automatically segmenting subcentimeter and 10–30 mm lesions were 0.692 ± 0.099 and 0.861 ± 0.033, respectively (p < 0.01). ResNet-152 model achieved AUCs of 0.875 (95

This study aimed to investigate the application of T2-based MRI delta-radiomics as a novel predictive tool for neoadjuvant chemotherapy (NACT) response in patients with osteosarcoma. We retrospectively analyzed data from 152 patients with pathologically confirmed osteosarcoma who underwent NACT at our institution. Axial T2-weighted MRI sequences were acquired both at baseline (pre-NACT) and after NACT (post-NACT). After image segmentation and preprocessing, 1158 radiomic features were extracted from the T2-weighted images. We developed and compared four models: the conventional quantitative imaging features-based model (CQIF model), the pre-NACT radiomics model, the post-NACT radiomics model, and the Delta-Radiomics model. Model performance was assessed using the area under the receiver operating characteristic curve (AUC) and accuracy (ACC). Based on histopathological assessment, patients were divided into two groups: good responders (n = 57) and poor responders (n = 95). Significant differences in change rates for tumor diameter and volume were observed between the two groups (P < 0.001). The Delta-Radiomics model demonstrated superior predictive performance compared to other models, achieving an AUC of 0.796, ACC of 0.756, sensitivity of 0.529, specificity of 0.893, PPV of 0.750, and NPV of 0.758 in the test set. However, the Delong test revealed no significant differences among these models, except between the Post-NACT and Delta-Radiomics models (P < 0.05). T2-based MRI delta-radiomics showed strong predictive value for NACT response in patients with osteosarcoma. This model holds potential for guiding clinical decision-making and improving patient management by identifying responders early in the treatment course.

Background: For sub-centimetre pure ground-glass nodules (pGGNs) in the lungs, accurately predicting their invasiveness remains a clinical challenge. This study aimed to assess the diagnostic value of peri-nodular radiomics features on enhanced computed tomography (CT) for predicting invasiveness, and develop a combined radiomics-clinical model to improve preoperative evaluation in early-stage lung adenocarcinoma (LUAD). Methods: This retrospective study analyzed patients with pathologically confirmed pGGNs from The Fourth Hospital of Hebei Medical University (training/internal validation: 309 nodules) and Xingtai People's Hospital (external validation: 38 nodules). Radiomics features were extracted from the nodule core and its surrounding 0-3 and 3-5 mm regions on CT scans. Feature selection was performed using the least absolute shrinkage and selection operator (LASSO) algorithm. Logistic regression was used to build predictive models for distinguishing non-invasive from invasive lesions. The dataset was split in a 7:3 ratio for training and internal validation. Model performance was assessed using the area under the receiver operating characteristic (ROC) curves (AUC) and decision curve analysis (DCA). A combined nomogram integrating radiomics and clinical features was also developed. Results: The combined intra-nodular and peri-nodular 0-3 mm radiomics model achieved the highest diagnostic performance in the validation set [AUC =0.847, 95% confidence interval (CI): 0.752-0.943], outperforming models based solely on intra-nodular (AUC =0.828) or peri-nodular features (AUC =0.800). The combined model further improved diagnostic accuracy (AUC =0.857), and DCA demonstrated its added clinical utility. A personalized nomogram incorporating RadScore and air bronchogram signs demonstrated potential clinical utility. Conclusions: Radiomics features from the peri-nodular 0-3 mm region significantly enhance the prediction of invasiveness in subcentimetric pGGNs. The combined radiomics-clinical model offers a promising tool for individualized decision-making in early-stage LUAD.

BACKGROUND:Epithelial ovarian cancer (EOC) can be broadly classified into type I and type II tumours, which exhibit distinct biological behaviours. Accurate preoperative differentiation between these subtypes is important for guiding treatment decisions and optimizing patient outcomes. PURPOSE:To evaluate the diagnostic value of quantitative parameters derived from synthetic magnetic resonance imaging (SyMRI) alone and in combination with clinico-morphological features for differentiating type I from type II EOCs. MATERIALS AND METHODS:This retrospective study included 92 patients with pathologically confirmed EOC, including 32 type I and 60 type II tumours, who underwent preoperative MRI. Quantitative parameters derived from SyMRI, including T1, T2, and proton density (PD), and DWI-derived apparent diffusion coefficient (ADC) values were measured from the solid components of the tumours. Clinico-morphological characteristics were also recorded. Differences between type I and type II EOCs were assessed using the independent Student's t, Mann-Whitney U test, or chi-squared tests. Multivariable logistic regression analysis was used to identify independent predictors and construct a combined model. The model's performance was evaluated using receiver operating characteristic (ROC) curve analysis. RESULTS:Type I EOCs exhibited significantly higher T1, T2, and ADC values than type II EOCs (all p < 0.05), whereas PD values did not differ significantly between the two groups (p = 0.746). Significant differences were also observed in patient age, serum CA125 levels, maximum tumour diameter, MRI enhancement characteristics, and texture (all p < 0.05). A multivariable analysis identified T1 value, CA125 level, maximum tumour diameter, and enhancement characteristics served as independent predictors. The combined model incorporating these variables achieved an AUC of 0.936, which was significantly higher than that of any individual parameter (all p < 0.05). CONCLUSIONS:SyMRI-derived quantitative parameters, particularly T1 values, may provide valuable biomarkers for differentiating EOC subtypes. A combined model integrating T1 values with clinico-morphological features showed high diagnostic performance and may support preoperative risk stratification and individualized treatment planning in patients with EOC.

OBJECTIVE:Coronary artery disease (CAD) progression is directly associated with major adverse cardiovascular events and death. This study aimed to construct a pericoronary adipose tissue (PCAT) radiomics model to predict subsequent progression in patients with CAD. METHODS:Data from 116 patients who had at least 2 coronary computed tomography angiography (CCTA) exams between March 1, 2020, and August 30, 2022, were collected at our institution. Obstructive stenosis, CAD-RADS classification, segment involvement score (SIS), and segment stenosis score (SSS) were noted. The radiomics features of the proximal to the left anterior descending artery, left circumflex artery, and right coronary artery were extracted on CCTA images using fully automated software. According to CAD-RADS, SIS, and SSS, non-progression was identified in 96, 80, and 72 patients and progression was identified in 20, 36, and 44 patients, respectively. All patients were randomly divided into the training and testing cohorts in a 7:3 ratio. Cox regression models were constructed based on PCAT radiomics signatures, and their predictive abilities were measured using receiver operating characteristic curves. RESULTS:We included 116 patients (age 58.00 [53.25, 64.00] years; 78 [67.20%] were male). After screening, 16 PCAT radiomics features were identified as being significantly related to CAD progression. The Cox regression models had area under the curve values of 0.841, 0.838, and 0.725 in the training cohort and 0.818, 0.817, and 0.851 in the testing cohort, respectively, to predict 2-year CAD-RADS, SIS, and SSS progression. CONCLUSIONS:PCAT-based radiomics models demonstrated promising performance in predicting subsequent CAD progression. ADVANCES IN KNOWLEDGE:PCAT-based radiomics signatures derived from coronary CT angiography provided incremental predictive value for CAD progression beyond conventional imaging markers (CAD-RADS, SIS, SSS), and may serve as noninvasive imaging biomarkers for individualized risk stratification.

BACKGROUND:Accurate preoperative assessment of lymphovascular invasion (LVI) remains challenging due to the high heterogeneity of gastric cancer (GC). PURPOSE:To evaluate the feasibility of a subregion-based radiomics model using multiparametric MRI (mpMRI) for preoperative evaluation of LVI and to further assess its prognostic value. STUDY TYPE:Retrospective. SUBJECTS:A total of 878 GC patients from four centers: 313 training, 133 internal test, and 432 external validation cases. FIELD STRENGTH/SEQUENCE:1.5 T and 3 T/mpMRI including T2-weighted imaging (FSE/TSE), diffusion-weighted imaging (SS-EPI), and contrast-enhanced T1-weighted imaging (FFE/VIBE). ASSESSMENT:The fuzzy c-means clustering was applied to subregion generation after manual segmentation. The subregional radiomics model was established using LVI-related features from a four-step extracted pipeline, with logistic regression, random forest, and support vector machine algorithms. The corresponding intra-tumoral subregion (ITS) index for each patient was obtained from the optimal subregional model. Subsequently, a combined model incorporating the ITS index and independent clinical characteristics was developed. Performance was further validated in test and validation cohorts. Additionally, the prognostic utility for overall survival (OS) and disease-free survival (DFS) was assessed in the follow-up cohort. STATISTICAL TESTS:Model area under the receiver operating characteristic curves (AUCs) was compared using net reclassification improvement (NRI) and integrated discrimination improvement (IDI). Kaplan-Meier survival analyses were conducted for prognostic evaluation. p < 0.05 was considered statistically significant. RESULTS:Pathological LVI-positive was detected in 448 (51.0%) patients. The combined model demonstrated satisfactory discrimination of LVI, achieving AUCs of 0.814 (training), 0.769 (test), and 0.758-0.783 (validation), outperforming the optimal subregional model with positive NRI and IDI across all cohorts. Furthermore, the ITS index maintained a significant association with OS (HR 33.50) and DFS (HR 30.00). DATA CONCLUSION:The combined model, which integrated the ITS index derived from subregional radiomics with clinical factors, demonstrated robust performance in evaluating both LVI and survival outcomes in GC patients. EVIDENCE LEVEL:3. TECHNICAL EFFICACY:Stage 3.

OBJECTIVES:Accurately predicting meningioma brain invasion preoperatively helps to select the appropriate surgical approach and predict prognosis, but there are few imaging features that are sufficient for discriminating it alone. We investigate the joint MR imaging features and apparent diffusion coefficient (ADC) to predict the risk of brain invasion of meningiomas preoperatively. METHODS:In this retrospective study, 143 patients (invasion group:51, non-invasion group: 92) diagnosed with meningioma by histopathology were included. The maximum (ADCmax), minimum (ADCmin) and mean (ADCmean) values of ADC and the mean ADC values of a comparative ROI in the normal appearing white matter (ADCNAWM) were calculated. Differences between clinical features, MRI morphological features, and all ADC values were assessed by Pearson's chi-square test and Kruskal-Wallis rank-sum test. Stepwise logistic regression analysis was used to select the optimal features and construct a prediction model. Furthermore, A nomogram was used to predict the risk of brain invasion, and a decision curve was used to verify the clinical utility of the nomogram. RESULTS:According to stepwise logistic regression analysis, we found that sex, maximum diameter, peritumoral edema and ADCmin were closely related to brain invasion in meningioma. The model of the above four variables has the optimal discriminative ability to predict brain invasion, with an AUC of 0.924 (95 % CI, 0.879-0.969) and a sensitivity of 92.2 % (95 % CI, 74.5%-98.0 %). CONCLUSIONS:Combining clinical features, MRI morphological characteristics and ADCmin, the model exhibits excellent discriminatory ability and high sensitivity, which can be used for predicting the risk of brain invasion of meningiomas.

Background:Small cell lung cancer (SCLC) comprises distinct molecular subtypes [neuroendocrine (NE) vs. non-NE] that have different prognoses, with NE tumors generally exhibiting a more aggressive clinical course. However, identifying these subtypes usually requires invasive tissue sampling. Radiomics-the extraction of quantitative features from medical images-offers a potential noninvasive alternative. This study aimed to predict the NE subtype of SCLC using radiomics analysis of contrast-enhanced computed tomography (CECT) images, and to compare a two-dimensional (2D) radiomics approach with a three-dimensional (3D) approach. Methods:In this single-center retrospective study, we included 51 patients with resected SCLC (NE subtype n=39, non-NE n=12) between 2005 and 2016, all with preoperative CECT scans and known molecular subtype confirmed by immunohistochemistry. Radiomics features were extracted from arterial-phase CECT images using both a 2D (single largest cross-sectional slice) and 3D (whole tumor volume) segmentation of the primary tumor. Radiomics-based logistic regression models were trained to classify NE vs. non-NE subtypes. Model performance was evaluated using receiver operating characteristic analysis [area under the curve (AUC)] with bootstrap 95% confidence intervals (CIs). A combined model incorporating radiomics and clinical factors was also tested. Additionally, we explored the association of the radiomics signature with recurrence-free survival (RFS) via Kaplan-Meier curves and Cox proportional-hazards analysis. Results:The 2D radiomics model achieved an AUC of 0.806 (95% CI: 0.666-0.945) for distinguishing NE vs. non-NE subtypes, comparable to the 3D model (AUC 0.784, 95% CI: 0.634-0.934; P=0.75 or 2D vs. 3D). At the optimal cutoff, the 2D model yielded 64.1% sensitivity and 83.3% specificity. The radiomics signature remained an independent predictor of NE subtype in a combined model [adjusted odds ratio (OR) 6.22, P=0.005], and the addition of radiomics improved the combined model's AUC to 0.861 (vs. 0.673 for clinical factors alone). No conventional clinical or CT features alone were significant predictors. Notably, the 2D radiomics score also stratified patients' outcomes: those predicted as NE subtype had a 5-year RFS of 48.1%, compared to 62.5% for non-NE (log-rank P=0.03). In multivariable Cox analysis, a higher radiomics score showed a trend toward shorter RFS [hazard ratios (HRs) 1.46 per SD increase, P=0.08]. Conclusions:Quantitative analysis of CECT images via radiomics can noninvasively distinguish NE and non-NE molecular subtypes of SCLC. A simplified 2D radiomics approach performed comparably to 3D volumetric analysis for subtype classification and also demonstrated prognostic relevance. Radiomics could serve as a valuable adjunct for SCLC subtype identification and risk stratification, potentially guiding more personalized treatment decisions.

Objective Exploring the construction of a fusion model that combines radiomics and deep learning (DL) features is of great significance for the precise preoperative diagnosis of meningioma sinus invasion. Materials and methods This study retrospectively collected data from 601 patients with meningioma confirmed by surgical pathology. For each patient, 3948 radiomics features, 12,288 VGG features, 6144 ResNet features, and 3072 DenseNet features were extracted from MRI images. Thus, univariate logistic regression, correlation analysis, and the Boruta algorithm were applied for further feature dimension reduction, selecting radiomics and DL features highly associated with meningioma sinus invasion. Finally, diagnosis models were constructed using the random forest (RF) algorithm. Additionally, the diagnostic performance of different models was evaluated using receiver operating characteristic (ROC) curves, and AUC values of different models were compared using the DeLong test. Results Ultimately, 21 features highly associated with meningioma sinus invasion were selected, including 6 radiomics features, 2 VGG features, 7 ResNet features, and 6 DenseNet features. Based on these features, five models were constructed: the radiomics model, VGG model, ResNet model, DenseNet model, and DL-radiomics (DLR) fusion model. This fusion model demonstrated superior diagnostic performance, with AUC values of 0.818, 0.814, and 0.769 in the training set, internal validation set, and independent external validation set, respectively. Furthermore, the results of the DeLong test indicated that there were significant differences between the fusion model and both the radiomics model and the VGG model (p < 0.05). Conclusions The fusion model combining radiomics and DL features exhibits superior diagnostic performance in preoperative diagnosis of meningioma sinus invasion. It is expected to become a powerful tool for clinical surgical plan selection and patient prognosis assessment.

Background: Pre-operative prediction of lymph node metastasis (LNM) in patients with colorectal cancer (CRC) is challenging, yet crucial for prognosis and treatment. This study aimed to develop and validate a computed tomography-based radiomics model to predict LNM preoperatively and to investigate its prognostic value. Methods: A total of 587 individuals with histologically confirmed CRC from two medical centers were retrospectively analyzed. From these, 257, 109, and 221 were allocated to training, internal validation, and external validation cohorts, respectively. A total of 1781 radiomics features were obtained from portal venous-phase computed tomography images. After feature selection, five machine learning classifiers were developed and compared. The optimal radiomics model was integrated with significant clinical predictors to develop a combined model whose performance was evaluated using receiver operating characteristics, calibration, and decision curves. The model's prognostic value was determined using Kaplan-Meier curve and Cox regression analyses. Results: Among the models, an extreme gradient boosting classifier demonstrated the best performance, achieving area under the receiver operating characteristic curves of 0.826, 0.807, and 0.752 in the training, internal validation, and external validation cohorts, respectively. The combined model integrating radiomics features and carcinoembryonic antigen levels showed higher predictive value, with area under the receiver operating characteristic curves of 0.842, 0.812, and 0.770 for the three cohorts. Risk stratification based on the combined model effectively identified patients with significantly different overall survival and disease-free survival (log-rank test, all p < 0.05). The model remained an independent predictor of both disease-free survival (hazard ratio = 2.857, 95% confidence interval: 1.694-4.818) and overall survival (hazard ratio = 1.975, 95% confidence interval: 1.001-3.919) in multivariable Cox analysis. Conclusions: Our proposed radiomics-based model demonstrated good performance in preoperative prediction of LNM and could provide valuable prognostic information for patients with CRC.

ObjectiveThis study aims to evaluate the clinical utility of computed tomography-guided percutaneous lung biopsy (CT-PLB) in the diagnosis of atypical pulmonary hamartoma (APH) and pulmonary sclerosing pneumocytoma (PSP).MethodsThis retrospective study analyzed 19 patients with pulmonary nodules who underwent CT-PLB at our hospital between October 2016 and August 2019. All patients underwent surgical excision within two weeks following CT-PLB, and the postoperative histopathological results were used as the reference standard for diagnosis. Among these patients, ten cases were confirmed as PSP and assigned to the PSP group, while nine cases were diagnosed as pulmonary hamartoma and assigned to the APH group. The diagnostic accuracy of CT-PLB for APH and PSP, as well as the incidence of procedure-related complications, were analyzed and compared.ResultsAmong the 19 patients, the overall diagnostic accuracy of CT-PLB was 89.5% (17/19). The diagnostic accuracy was 88.9% (8/9) in the APH group and 90.0% (9/10) in the PSP group. Complications included one case of minimal pneumothorax in the PSP group and one case of mild hemoptysis in the APH group. Additionally, 14 patients experienced mild to moderate pulmonary hemorrhage, including 8 cases (5 mild, 3 moderate) in the APH group and 6 cases (all mild) in the PSP group. No severe complications, such as pleural reactions, occurred in any patient.ConclusionCT-PLB demonstrates high clinical diagnostic value for both APH and PSP. The procedure is safe, reliable, and associated with few complications. Thus, it can serve as a valuable tool for preoperative clinical diagnosis.

Early prediction of chemotherapy efficacy continues to be a significant challenge in osteosarcoma patients. This study aims to investigate the impact of habitat analysis based on Intravoxel Incoherent Motion (IVIM) and Dynamic Contrast-Enhanced magnetic resonance imaging (DCE-MRI) for predicting neoadjuvant chemotherapy (NACT) response in osteosarcoma patients. We prospectively analyzed 71 patients diagnosed with osteosarcoma who underwent pre-treatment IVIM and DCE-MRI. IVIM metrics, including pseudo-diffusion coefficient (ADC_fast), diffusion coefficient (ADC_slow), and fractions of ADC_fast (f_ADC_fast), were evaluated alongside DCE-MRI parameters such as Ktrans, Ve, and Kep. We assessed the predictive value of habitat regions’ features from these functional images for NACT outcomes. Model performance was measured using the area under the receiver operating characteristic curve (AUC). Among the data, 39 patients were classified as poor responders and 32 as good responders. Demographic factors, including age, sex, and lesion location, showed no significant differences between groups (P > 0.05). Our findings indicated that an increase in volume in habitat 3, coupled with decreases in f_ADC_fast in habitat 1, ADC_slow in habitat 5, and f_ADC_fast in habitat 5 was valuable indicators for predicting the response of osteosarcoma to NACT. The Habitat Feature-Based IVIM-Bi (HAB-IVIM-Bi) model demonstrated the best performance, as indicated by the Delong test, which yielded all P values < 0.05. It achieved an AUC of 0.797, accuracy of 0.761, sensitivity of 0.719, and specificity of 0.795. Habitat analysis based on multiparametric MRI demonstrates promising potential for non-invasive prediction of NACT response in osteosarcoma. The HAB-IVIM-Bi model may serve as a valuable tool for treatment stratification and personalized therapy planning.

This study was aimed at differentiating brain metastases (BMs) from non-small cell lung cancer (NSCLC) vs. small cell lung cancer (SCLC), and the adenocarcinoma (AD) vs. non-adenocarcinoma (NAD) subtypes, according to radiomics features derived from multiparametric magnetic resonance imaging (MRI). A total of 276 patients with BMs, including 98 with SCLC and 178 with NSCLC, were randomly divided into training (193 cases) and test (83 cases) datasets in a 7:3 ratio. Of the 178 patients with NSCLC, 155 had primary AD, and 23 had NAD; those patients were also randomly divided into training (124 cases) and test (54 cases) datasets. Logistic regression analysis was used to construct classification models based on the radiomics features extracted from contrast-enhanced T1-weighted imaging (T1CE), T2-fluid-attenuated inversion recovery (T2-FLAIR), and diffusion-weighted imaging (DWI) images. Diagnostic efficiency was evaluated with the area under the receiver operating characteristic curve (AUC) through Delong's test, calibration curves through the Hosmer-Lemeshow test and Brier score, precision-recall curves, and decision curve analysis. Compared with radiomics features derived from a single sequence, multiparametric combined-sequence MRI radiomics features based on T1CE, T2-FLAIR, and DWI images exhibited greater specificity in distinguishing BMs originating from various lung cancer subtypes. In the training and test datasets, the AUCs of the model for the classification of SCLC and NSCLC BMs were 0.765 (95% CI 0.711, 0.822) and 0.762 (95% CI 0.671, 0.845), respectively, whereas the AUCs of the prediction models combining the three sequences in differentiating AD from NAD BMs were 0.861 (95% CI 0.756, 0.951) and 0.851 (95% CI 0.649, 0.984), respectively. The radiomics classification method based on the combination of multiple MRI sequences can be used for differentiating various lung cancer BMs.

To develop and validate a multimodal model that integrates radiomics features (RFs) and deep learning features (DFs) derived from preoperative multisequence magnetic resonance imaging (MRI) for the prediction of lymphovascular space invasion (LVSI) in patients with endometrial cancer (EC). This multicenter, retrospective study enrolled 892 patients with postoperative pathologically confirmed EC. Preoperative MRI comprised T2-weighted imaging, contrast-enhanced T1-weighted imaging, and apparent diffusion coefficient maps, were analyzed. Regions of interest (ROIs) were manually delineated for 2D and 3D analyses. RFs were extracted using PyRadiomics, and DFs were obtained using pretrained VGG 11, ResNet 101, and DenseNet 121 architectures. Five single-modality models (2D-RF, 3D-RF, VGG11-DF, ResNet101-DF, and DenseNet121-DF) were developed. In addition, the integration of RFs and DFs were explored to construct combined models. Models were trained in a training cohort (n = 378) and evaluated in both internal (n = 160) and external (n = 354) validation cohorts. Model performance was evaluated by the area under the receiver operating characteristic curve (AUC). In the training cohort, the 2D-RF and 3D-RF models showed comparable performance for LVSI prediction (AUC: 0.775 vs. 0.772, P = 0.89). Among the deep learning models, DenseNet121-DF achieved the highest AUC (0.757), which was significantly higher than ResNet-101-DF (AUC: 0.671; P = 0.01) and not statistically different from VGG11-DF (AUC: 0.720, P = 0.20). The optimal combined model, integrating features from 2D-RF and DenseNet121-DF, yielded the highest performance in the training cohort (AUC: 0.796). These findings were confirmed in both the internal and external validation cohorts. A multimodal MRI-based model integrating both RFs and DFs achieved superior performance for noninvasive prediction of LVSI in patients with EC. This approach holds potential to enhance preoperative risk stratification and guide personalized treatment planning.

BackgroundThe prognosis for hepatocellular carcinoma (HCC) is unfavorable, primarily attributable to the high incidence of recurrence.PurposeTo assess the prognostic value of multiparametric magnetic resonance imaging (mp-MRI) based on radiomic features for overall survival (OS) in patients with HCC.Material and MethodsPatients who underwent abdominal mp-MRI examination before hepatectomy in our hospital between January 2016 and December 2019 were retrospectively collected and divided into a training group and a verification group at a ratio of 7:3. The patients' images, clinical parameters, and semantic features were collected. A three-dimensional volume of interest was delineated and radiomics features were screened. Independent predictors of clinical imaging were screened and combined with radiomics features to construct a combinatorial model and draw a nomogram. The predictive efficacy of the model was evaluated.ResultsThe Harrell's C-index values were 0.737 and 0.711 for the clinical imaging model and 0.705 and 0.704 for the full sequence model in the training group and validation group, respectively. The combinatorial model had higher efficiency, and the C-index values in the training group and the validation group were 0.779 and 0.756, respectively. The survival curve showed that the low-risk group defined by the radiomics signature had significantly better OS than the high-risk group (3-year OS: 61.54% vs. 30.77%; P < 0.05).ConclusionThe combined model can predict the OS of patients with HCC non-invasively before surgical resection and can be used as a clinical tool to guide individualized treatment.

IntroductionThis study aimed to develop and validate a predictive model for tumor shrinkage patterns in hormone receptor-positive, HER2-negative (HR+/HER2-) breast cancer patients undergoing neoadjuvant chemotherapy (NAC).MethodsA retrospective analysis was conducted on 227 HR+/HER2- breast cancer patients with a desire for breast conservation, examining their clinicopathological characteristics, traditional MRI features, and radiomics features. Patients were divided into training and validation cohorts in a 7:3 ratio. Tumor shrinkage patterns were classified into Type I and Type II based on RECIST 1.1 criteria. A clinical model was established using Ki67 quantification and enhancement pattern. Radiomics features were extracted and analyzed using machine learning algorithms, including Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), and Random Forest (RF). A combined clinical-radiomics model was also developed.ResultsThe clinical model achieved an area under the curve (AUC) of 0.624 in the training cohort and 0.551 in the validation cohort. The RF radiomics model showed the highest predictive performance with an AUC of 0.826 in the training cohort and 0.808 in the validation cohort. The combined clinical-radiomics model further improved prediction accuracy, with an AUC of 0.831 in the training cohort and 0.810 in the validation cohort.ConclusionRadiomics features based on baseline MRI significantly enhance the prediction of tumor shrinkage patterns in HR+/HER2- breast cancer patients. This approach aids in the early identification of patients likely to benefit from breast-conserving surgery and facilitates timely treatment adjustments.

Purpose To establish an outcome prediction model for multiple timescales itracerebral hemorrhage (ICH). Materials and methods ICH patients with an onset-to-imaging time (OIT) of less than 72 h were retrospectively collected. Patients were divided into three groups according to their OIT. Group 1–3: OIT < 6h, 6 ≤ OIT<24h, 24 ≤ OIT<72h. The first preoperative review in Group 1 within 72 h were recorded for Group 4. A binary logistic regression classifier was used to construct outcome prediction models for each group. The predictive performance of each group’s model was compared with the application of the Group 1 model across different groups to explore its accuracy and applicability in predicting the outcomes for patients in various groups. Results A total of 2,136 patients with ICH were included in the study. In the training set, the AUC value obtained by directly applying the group 1 radiomics model to group 2 patients was 0.773. The AUC value for the direct application of the Group 1 combined model to Group 3 patients was 0.811. The AUC for applying the Group 1 radiomics model directly to Group 4 patients was 0.779. The AUC of the radiomics model, constructed by combining the key radiomics features of Groups 1 and 4, was 0.788. The AUC of the independent radiomics model for Group 4 was 0.815. Conclusion The radiomics and combined models for predicting outcomes in ICH patients with OIT < 6 h can be applied to all patients with OIT < 72 h, allowing for early and accurate outcome prediction across multiple timescales.

ObjectivesEsophageal sarcomatoid carcinoma (ESC) is a rare malignant tumor. This study aims to analyze the computed tomography (CT) features and clinicopathological characteristics of resectable ESC.MethodsThe CT and clinicopathological data of 25 patients with ESC, confirmed by postoperative pathology, were retrospectively analyzed. The preoperative CT feature analysis included the average tumor CT attenuation value (CTTumor), the average normal esophagus CT attenuation value (CTNormal), the tumor-to-normal esophagus attenuation ratio (TNR), enhancement pattern, tumor margin, and tumor morphology, which was classified into two types: mass-forming type and wall-thickening type. Additionally, CT-measured tumor thickness (cTT), CT-measured tumor length (cTL), and CT-measured tumor volume (cTV) were measured and recorded. The analysis of clinicopathological characteristics encompassed variables such as age, gender, clinical symptoms, pathological tumor thickness (pTT), pathological tumor length (pTL), T stage, N stage, lymphovascular invasion status, and neural invasion status. To assess the agreement and correlation between CT features and pathological results, Bland-Altman plots and Pearson correlation coefficient analyses were performed for pTT versus cTT and pTL versus cTL, respectively.ResultsAmong the 25 patients with ESC, 19 were male and 6 were female. The patients’ ages ranged from 47 to 73 years, with a mean age of 65.48 ± 6.85 years and a median age of 68 years. The pathological staging results showed that 14 cases were at stage T1, 5 cases at stage T2, 5 cases at stage T3, and 1 case at stage T4. Lymph node metastasis was identified in 12 cases, including 5 classified as N1, 4 as N2, and 3 as N3. The pTT ranged from 0.50 to 4.00 cm, with a mean of 1.92 ± 1.02 cm and a median of 1.50 cm; the cTT ranged from 0.60 to 4.10 cm, with a mean of 2.02 ± 0.90 cm and a median of 1.66 cm. Bland-Altman analysis demonstrated a mean difference of 0.10 cm between cTT and pTT, with 92.0% (23/25) of cases lying within the 95% limits of agreement (Mean ± 1.96 SD). The Pearson correlation coefficient between the two measurements was 0.980, indicating a strong positive correlation. The pTL ranged from 2.00 to 7.00 cm, with a mean of 4.24 ± 1.50 cm and a median of 4.00 cm; the cTL ranged from 2.57 to 7.50 cm, with a mean of 4.41 ± 1.48 cm and a median of 4.16 cm. Bland-Altman analysis demonstrated a mean difference of 0.18 cm between cTL and pTL, with 92.0% (23/25) of cases lying within the 95% limits of agreement. The Pearson correlation coefficient was 0.884. The cTV ranged from 1.56 cm³ to 41.49 cm³, with a median of 8.25 cm³. Regarding tumor morphology, 20 cases presented as mass-forming type, while 5 were classified as wall-thickening type. Significant differences in both pTT and cTT were observed between the two morphological types; however, no statistically significant differences were found in other CT features.ConclusionESC is a rare malignant tumor characterized by distinctive CT features. The majority of cases manifest as the mass-forming type, while a smaller proportion present as the wall-thickening type. The definitive diagnosis of ESC depends on pathological examination.

ObjectiveTo develop diagnostic models for differentiating gastric neuroendocrine carcinoma (g-NEC) and gastric mixed adeno-neuroendocrine carcinoma (g-MANEC) from gastric adenocarcinoma (g-ADC) based on traditional contrast enhanced CT imaging features and radiomics features.MethodsWe retrospectively analyzed 90 g-(MA)NEC (g-MANEC and g-NEC) patients matched 1:1 by T-stage with 90 g-ADC patients. Traditional CT features were analyzed using univariable and multivariable logistic regression. Tumor segmentation and radiomics features extraction were performed with Slicer and PyRadiomics. Feature selection was conducted through univariable analysis, correlation analysis, LASSO, and multivariable stepwise logistic. The combined model incorporated clinical and radiomics predictors. Diagnostic performance was assessed with ROC curves and DeLong’s test. The models’ diagnostic efficacy was further validated in subgroup of g-NEC vs. g-ADC and g-MANEC vs. g-ADC cases.ResultsTumor necrosis and lymph node metastasis were independent predictors for differentiating g-(MA)NEC from g-ADC (P < 0.05). The clinical model’s AUC was 0.700 (training) and 0.667(validation). Five radiomics features were retained, with the radiomics model showing AUC of 0.809 (training) and 0.802 (validation). The combined model’s AUCs were 0.853 (training) and 0.812 (validation), significantly outperforming the clinical model (P < 0.05). Subgroup analysis revealed that the combined model exhibited acceptable performance in differentiating g-NEC from g-ADC and g-MANEC from g-ADC, with AUC of 0.887 and 0.823 in the training cohort and 0.852 and 0.762 in the validation cohort.ConclusionA combined model based on traditional CT imaging and radiomic features provides a non-invasive and effective preoperative diagnostic method for differentiating g-(MA)NEC from g-ADC.