Objective To investigate the predictive value of arterial enhancement fraction（AEF）in primary lesions measured by dual-layer detector spectral CT for cervical lymph node metastasis（LNM）in papillary thyroid carcinoma（PTC）. Methods This study included 145 PTC patients（17 male,128 female）.Patients were divided into the non-LNM group（n=65）and the LNM group（n=80）based on the presence or absence of cervical LNM.To evaluate the morphological characteristics of thyroid papillary carcinoma lesions,measure the iodine density of the lesion in the arteriovenous phase and the internal carotid artery,calculate the AEF value,and measure the AEFmapping value on the generated AEF fusion map.Spearman correlation analysis assessed the correlation between AEF and AEFmapping.The Bland-Altman plot was used to describe the deviation between the average differences of AEF and AEFmapping and to estimate the consensus interval.Multivariate Logistic regression analysis was used to construct three models：conventional imaging features model,AEF+conventional imaging features model,and AEFmapping+conventional imaging features model.Receiver operating characteristic（ROC）curves were plotted to evaluate diagnostic efficiency. Results Significant differences were found between the LNM and the non-LNM groups regarding lesion long diameter,short diameter,aspect ratio,and contact extent with the capsule（P<0.05）.Both AEF and AEFmapping of the lesions were significantly higher in the LNM group compared to the non-LNM group（P<0.05）.AEF and AEFmapping showed good correlation（r=0.815,P<0.001）and good agreement,with a mean difference of 0.6 and 95% limits of agreement of -7.2 to 8.3（P=0.084）.Multivariate Logistic regression analysis showed that when the conventional imaging feature model included short diameter,aspect ratio,and contact extent with the capsule（≥50%）,the AUC for predicting cervical LNM in patients with papillary thyroid carcinoma was 0.810,with a sensitivity of 76.25% and a specificity of 73.85%.Combining AEF or AEFmapping with conventional features significantly improved diagnostic efficiency（AUC：0.875 vs. 0.810,P=0.015 for AEF model;0.869 vs. 0.810,P=0.019 for AEFmapping model）.The diagnostic efficiency of the AEF+conventional imaging features model and the AEFmapping+conventional imaging features model was comparable（P=0.45）. Conclusion AEF and AEFmapping demonstrate good correlation and consistency.Combining AEF or AEFmapping with conventional imaging features improves the diagnostic efficacy for predicting cervical lymph node metastasis in papillary thyroid carcinoma.

Objective To investigate the value of spectral CT multi-parameter imaging in the preoperative prediction of sentinel lymph node metastatic burden in breast cancer. Methods Clinical data and spectral CT multi-parameter imaging features of breast cancer patients admitted to the First Affiliated Hospital of Kunming Medical University from September 2023 to December 2024 were retrospectively collected. Spectral CT parameters and clinical characteristics were analyzed to identify independent influencing factors for sentinel lymph node metastatic burden, and predictive models were constructed. Results A total of 99 patients were included, with 74 in the low-burden group and 25 in the high-burden group. Based on the analysis, clinicopathological-spectral CT conventional imaging model, a spectral CT quantitative parameter model, and a combined nomogram model were developed, all of which demonstrated good predictive performance. Conclusion Spectral CT-based multi-parameter imaging can be effectively used for preoperative prediction of sentinel lymph node metastatic burden in breast cancer, providing a reliable imaging basis for early and accurate clinical assessment of axillary lymph node status.

Objective To explore the diagnostic performance of dual-layer spectral detector CT (DLCT) for the rupture of anterior cruciate ligament (ACL). Methods 44 patients with unilateral ACL rupture were prospectively enrolled, and the bilateral knees were scanned by DLCT and MRI. The conventional 120 kVp CT value, 40-80 keV virtual monoenergetic CT value, electron density and effective atomic number (Z_eff) were measured to quantitatively differentiate rupture ACLs from normal ACLs. Receiver operating characteristic (ROC) curve was used to analyze the efficacy of different energy spectrum quantitative parameters, and area under the curve (AUC) values were compared using DeLong's test. Two doctors assessed the integrity of ACLs through MRI and DLCT images independently, and then the diagnostic ratings were recorded. The diagnostic performances of DLCT and MRI for ACL integrity were compared using McNemar’s test. Results Significant differences were found in all of the energy spectrum parameters. Besides Z_eff, all the other parameters exhibited excellent diagnostic efficacy in detecting ACL ruptures (AUC=0.985-0.998). McNemar’s test showed no evidence of a difference between DLCT and MRI for the detection of ACL rupture (P=0.125). Conclusion DLCT has excellent diagnostic performance in qualitatively and quantitatively diagnosing the rupture of ACL.

Objective To investigate the value of dual-layer detector spectral CT (DECT) multi-parameter imaging for the preoperative prediction of lymphovascular invasion (LVI) and perineural invasion (PNI) in esophageal squamous cell carcinoma (ESCC). Methods A retrospective analysis was conducted on the clinicopathological and DECT imaging data of 118 patients with pathologically confirmed ESCC, who were divided into LVI/PNI-positive and negative groups. All patients underwent preoperative dual-phase enhanced DECT scans. Spectral parameters, including iodine density (ID), normalized iodine density (NID), effective atomic number (Zeff), electron density (ED), and spectral curve slope(λ), were measured and calculated. Univariate and multivariate Logistic regression analyses were employed to identify independent predictors of LVI/PNI. A predictive model was constructed using the balanced random forest algorithm with five-fold cross-validation. Receiver operating characteristic (ROC) curves were plotted to evaluate model performance. Results Among the 118 patients, 51 were in the LVI/PNI-positive group and 67 were in the negative group. Multivariate Logistic regression analysis identified clinical N stage and normalized iodine density in the venous phase (NIDV) as independent predictors of LVI/PNI. The combined predictive model based on these two factors demonstrated excellent diagnostic performance, achieving a mean area under the curve (AUC) of 0.897 with five-fold cross-validation. The mean accuracy, recall, F1-score, and precision were 78.9%, 91.2%, 0.787, and 69.9%, respectively. Conclusion Multiparameter imaging of DECT holds significant value for the preoperative prediction of LVI and PNI in esophageal cancer. The model combining clinical N stage and NIDV demonstrates optimal performance and is expected to provide imaging evidence for clinical decision-making.

Objective To investigate the effect of dual-layer spectral CT (DLCT) combined with motion-compensated reconstruction (MCR) algorithm on the improvement of image quality in prospective electrocardiogram-gated coronary CTA (CCTA). Methods A total of 121 patients (61 male, 60 female, mean age 53.1±10.91 years) suspected or diagnosed with coronary artery disease were prospectively enrolled. Coronary CT angiography (CCTA) was performed using a second-generation DLCT scanner with prospective ECG gating. Patients were divided into two groups based on heart rate: Group A (heart rate ≤ 75 beats/min, n=61) with reconstruction at （75±5）% R-R interval, and Group B (heart rate > 75 beats/min, n=60) with reconstruction at （45±5）% R-R interval. Both groups underwent reconstruction of standard (STD) images and MCR algorithm images. Objectiveimage quality parameters, including CT values, standard deviation (SD), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR), were evaluated for the aortic root and coronary artery segments (LM, LAD, LCX, RCA). Two radiologists with over five years of experience scored the subjective image quality. Results The effective radiation dose was （3.33±0.56） mSv in Group A and （3.51±0.72） mSv in Group B. In Group A, the CT values of coronary artery segments (LMA, LAD, LCX, RCA) on MCR images were higher than those on STD images (all P<0.05). For the RCA, the SNR and CNR of MCR images were higher than those of STD images (all P<0.05), while the SD of MCR images was lower than that of STD images (P<0.05). In Group B, the CT values, SNR, and CNR of coronary artery segments on MCR images were all higher than those on STD images (all P<0.05), and the SD of MCR images was lower than that of STD images (P<0.05). Subjective image quality assessments showed good consistency (STD Kappa=0.86, MCR Kappa=0.85), and all evaluations indicated that the MCR algorithm provided better visualization of coronary branches (P<0.05). In Group A, 56.96% of STD images and 97.95% of MCR images were rated as “excellent or good”; in Group B, the proportions were 35.84% for STD images and 95.42% for MCR images. Conclusion sThe MCR algorithm of DLCT can be applied to prospectively ECG-gated CCTA and effectively improves image quality, especially in patients with heart rates greater than 75 bpm.

Objective This study aimed to assess the hemodynamic features and biological activity of cerebral alveolar echinococcosis (CAE) using perfusion-weighted imaging (PWI). Methods Seventeen patients diagnosed with alveolar echinococcosis in the First Affiliated Hospital of Xinjiang Medical University from June 2013 to January 2020 were collected. Conventional enhanced 3.0 T MR and PWI examinations were performed to determine rCBF, rCBV, MTT and TTP in the parenchymal area, marginal area, edema area, and contralateral normal area of the lesion. Ten patients underwent PET-CT simultaneously. A total of 16 lesions matched with perfusion examination were selected. The correlation between PWI parameters in the marginal area and SUV_max was analyzed, and the ROC curve was drawn to evaluate the diagnostic efficacy. Results The rCBF and rCBV in the marginal area of the lesion were significantly higher than those in the parenchymal area (P<0.05), and the TTP decreased, suggesting abundant blood supply in the marginal area and insufficient blood supply in the parenchymal area. The TTP in the marginal area was negatively correlated with SUV_max (r=-0.319, P<0.05). ROC analysis showed that the AUC of rCBF, rCBV, and SUV_max was 0.845, 0.828, and 0.856 respectively (P<0.05). Conclusion PWI can effectively evaluate the activity of CAE lesions. The TTP in the marginal area is related to SUVmax and can be used as an alternative method to PET/CT in clinical practice.

Objective To explore the value of the quantitative model constructed based on the radiomics features of the infarct area on CT in the personalized prediction of the onset time window (≥ 4.5 h or < 4.5 h) of acute ischemic stroke (AIS). Methods A total of 191 AIS patients were retrospectively enrolled from two centers (134 cases in the training set and 57 cases in the testing set). Radiomics features of the infarct area were extracted from non-contrast enhanced CT, and a radiomics score (Rad-score) was then established. In the training set, correlations between Rad-score, clinical and imaging characteristics, and the time window were assessed using Logistic regression analysis. An quantitative time window prediction model was established based on the Shapley Additive Explanation (SHAP) method. Subsequently, the model was externally validated in the test set. The diagnostic performance of the model was assessed using the receiver operating characteristic (ROC) curve. Results The Rad-score, established by seven radiomics features, was negatively correlated with stroke onset time (r = - 0.38, P < 0.05). In multivariate analysis, age and Rad-score were independent predictors of stroke time window. The model based on Rad-score quantitatively assessed the likelihood of onset time greater or less than 4.5 h for each AIS patient, and the AUCs in the training and test cohorts were 0.87 and 0.79, respectively. When the variable of age was added, the AUCs in the training and test cohorts were 0.88 and 0.83, respectively. Conclusion The quantitative model based on radiomics features was dependable to make a personalized prediction for the time window in AIS patients and holds potential as a novel method for evaluating the time window in wake-up stroke patients.

Objective To evaluate whether the combined use of DWI and MRS can improve the diagnostic performance of the Kaiser score for breast lesions. Methods A retrospective analysis was conducted on 107 female patients (111 lesions) who underwent 3.0 T breast MRI (including non⁃contrast, dynamic contrast⁃enhanced scans) and MRS at Jinan Eighth People's Hospital between January 2023 and February 2025. Each lesion was assessed using the Kaiser score, ADC value measurement, and tCho peak detection. Pathological results served as the gold standard. Receiver operating characteristic (ROC) curves and Kappa consistency analysis were employed to evaluate the diagnostic efficacy of the Kaiser score and the combined Kaiser+ADC+MRS (Kaiser+ score). The area under the curve (AUC) was compared using Delong’s test. Results The Kaiser score achieved an AUC of 0.938, with an optimal diagnostic cutoff of 4 points, demonstrating a sensitivity of 96.67% (95% CI: 88.5%-99.6%) and specificity of 80.39% (95% CI: 66.9%-90.2%). The Kaiser+ score yielded an AUC of 0.969 (optimal cutoff: 4 points), with a sensitivity of 93.33% (95% CI: 83.8%-98.2%) and specificity of 90.20% (95% CI: 78.6%-96.7%). The AUC of the Kaiser+ score was significantly higher than that of the Kaiser score alone (P=0.026). Kappa consistency analysis revealed agreement coefficients of 0.779 (95% CI: 0.663-0.896) for the Kaiser score and 0.837 (95% CI: 0.663-0.896) for the Kaiser+ score against pathological results. Conclusion The integration of DWI and MRS with the Kaiser score enhances diagnostic performance, maintaining high sensitivity while improving specificity and potentially reducing unnecessary biopsy rates.

Objective To identify independent risk factors for postoperative recurrence/metastasis by constructing a LASSO regression model integrating multimodal MRI radiomic features and clinicopathological indicators. Methods A total of 500 pathologically confirmed esophageal cancer patients who underwent MRI examination before receiving neoadjuvant chemotherapy combined with immunotherapy (NAC + immunotherapy) were enrolled, including 73 in the favorable group (no recurrence/metastasis) and 427 in the unfavorable group (with recurrence/metastasis). Multimodal MRI parameters and clinicopathological characteristics were collected. A LASSO-Logistic model was constructed to screen for risk factors influencing recurrence/metastasis after NAC + immunotherapy. The model underwent internal and external validation. Receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA) were used to evaluate model performance and clinical utility. Results The proportion of stage Ⅲ disease, poorly differentiated tumors, lymphovascular invasion positivity, and lymph node metastasis were significantly higher in the unfavorable group (P<0.05). Maximum tumor diameter, tumor volume, apparent diffusion entropy, high b-value signal intensity, ADC kurtosis, peritumoral 3-mm region mean ADC, K_trans, K_ep, AUC, TTP, and the proportion of high-perfusion areas were higher in the unfavorable group (P<0.05). Mean ADC value and the proportions of medium- and low-perfusion areas were lower in the unfavorable group (P<0.05). The optimal model was obtained at λ = 0.041. LASSO-Logistic regression identified Ktrans (high), lymphovascular invasion (positive), lymph node metastasis (yes), peritumoral mean ADC (low), proportion of high-perfusion area (high), TNM stage (Ⅲ), and mean ADC (low) as risk factors for recurrence/metastasis after NAC + immunotherapy (P<0.05). A nomogram prediction model was constructed based on Logistic results. K_trans (high), lymphovascular invasion (positive), proportion of high-perfusion area (high), peritumoral mean ADC (low), and mean ADC (low) were the strongest predictors, followed by TNM stage (Ⅲ) and lymph node metastasis (yes). ROC curve analysis showed AUCs of 0.841 (95% CI: 0.774-0.897) and 0.862 (95% CI: 0.787-0.934) for the model and validation groups, respectively. Internal and external validation via bootstrap resampling, Hosmer-Lemeshow calibration curves, and DCA curves indicated no statistically significant difference between predicted and observed values (P>0.05). Conclusion The LASSO model based on multimodal MRI and pathological features can accurately identify high-risk patients for postoperative recurrence in esophageal cancer, providing an imaging tool for individualized prevention and treatment.

Objective This study aims to investigate the correlation between abdominal visceral fat area, measured by quantitative computed tomography (QCT), and coronary artery calcification (CAC) in patients with type 2 diabetes mellitus (T2DM). Methods A total of 449 patients (228 males, 221 females) who underwent coronary artery computed tomography angiography (CTA) and concurrent chest, abdominal, or lumbar spine CT scans at our hospital from January 2023 to March 2025 were enrolled. Visceral fat area (VFA) at the L₂/L₃ level was measured using quantitative computed tomography (QCT). The Agatston score was calculated using an artificial intelligence-assisted diagnostic system. Patients were stratified into low-risk group (coronary artery calcification score, CACS ≤ 100) and high-risk group(CACS > 100) groups. According to the fasting blood glucose(FBG) levels, Patients were divided into a normal blood glucose group (FBG < 7mmol/L) and hyperglycemia group (FBG ≥ 7mmol/L). Partial correlation analysis was performed to assess the association between VFA and CACS. Multivariate regression analysis was conducted to identify the risk factors for CAC. A clinical model was constructed based on clinical indicators, and a combined model was developed by integrating VFA with the clinical model. The diagnostic efficacy of the models in predicting high-risk CAC in patients with T2DM was analyzed using ROC. Results There were 241 cases in the low-risk group and 208 cases in the high-risk group. The high-risk group had higher VFA, age, and prevalence of hypertension than the low-risk group, and the differences were statistically significant (P < 0.05). Partial correlation analysis showed no correlation between VFA and CACS (P = 0.100). Multivariate regression analysis indicated that VFA, age, and hypertension were independent risk factors for high-risk CAC in T2DM patients.In the male subgroup, the high-risk group had significantly higher age and hypertension prevalence compared to the low-risk group,and the differences were statistically significant (P < 0.05). Although VFA was higher in the high-risk group, the difference was not statistically significant(P < 0.05). VFA was higher in the high-risk group than in the low-risk group, but the difference was not statistically significant (P > 0.05).In the female subgroup, the high-risk group showed significantly higher age, VFA, menopausal history, and prevalence of hypertension and hyperglycemia. and the differences were statistically significant (P < 0.05). Partial correlation analysis showed no correlation between VFA and CACS (P = 0.277). Multivariate regression analysis indicated that VFA, age, and hypertension were independent risk factors for high-risk CAC in female T2DM patients. In the hyperglycemia subgroup, the high-risk group had higher age and prevalence of hypertension than the low-risk group, and the differences were statistically significant (P < 0.05). VFA was higher in the high-risk group than in the low-risk group, but the difference was not statistically significant (P > 0.05). In the normal blood glucose subgroup, the high-risk group had higher age, VFA, and prevalence of hypertension than the low-risk group, and the differences were statistically significant (P < 0.05). ROC analysis showed that in the female subgroup, the predictive efficacy of the conjunctive model (AUC = 0.764) was higher than that of VFA (AUC = 0.731) and the clinical model (AUC = 0.712), and the difference was statistically significant (P = 0.027). Conclusion sQCT-measured VFA positively correlates with CACS in T2DM patients with normal blood glucose control. VFA has certain incremental value in predicting CAC risk stratification in female T2DM patients.

Objective The present study aims to investigate the relationship between solitary solid pulmonary nodules (SPN) and pulmonary vasculature, with the Objective of providing additional valuable insights for the differential diagnosis between benign and malignant solitary pulmonary nodules. Methods A retrospective analysis was conducted on the clinical and CT data of 365 patients with SPN who were confirmed by surgical pathology from January 2018 to June 2023. The patients were divided into the malignant group (152 cases) and the benign group (213 cases). The latter was further divided into the benign tumor group (103 cases) and the inflammatory nodule group (110 cases). The CT vascular signs were classified according to the positional relationship between the pulmonary vessels and the SPN and their morphological characteristics. The differences in CT vascular signs between the malignant group and the benign group, as well as between the inflammatory nodule group and the benign tumor group were compared, and their diagnostic efficacy for benign and malignant SPN was evaluated. Results (1)CT vascular signs can be classified into four types: TypeⅠ, where the pulmonary vessels run closely along the edge of the nodule, with normal morphology or being flattened due to compression; Type Ⅱ, where the pulmonary vessels run within the nodule and have normal morphology; Type Ⅲ, where the pulmonary vessels terminate at the edge of the nodule; Type Ⅳ, where the pulmonary vessels run within the nodule and have abnormal morphology. (2)Benign lesions exhibited significantly higher rates of Type I and Ⅱ CT vascular signs than malignant counterparts, whereas the latter showed markedly elevated incidences of Type Ⅲ and Ⅳ signs. All intergroup differences were statistically significant (P < 0.05 for all).The sensitivity of CT vascular signs for differentiating benign and malignant SPN was 80.3%, the specificity was 89.7%, and the accuracy was 85.8%. (3)In the benign tumor group, the incidence of type Ⅰ was significantly higher than that in the inflammatory nodule group, while the incidence of type Ⅱ in the inflammatory nodule group was significantly higher than that in the benign tumor group. All differences were statistically significant (all P < 0.05). There were no statistically significant differences in the incidence of Type Ⅲ and Type Ⅳ between the two groups (all P > 0.05). Conclusion :CT vascular signs have certain value in differentiating benign and malignant SPN. When the pulmonary vessels terminate at the edge of the SPN or run within it with abnormal morphology, one should be highly vigilant of the possibility of lung cancer.

Objective To explore the value of clinical and flexible subtraction CT(SCT) features for predicting microvascular invasion (MVI) in hepatocellular carcinoma (HCC). Methods A retrospective review was conducted of 90 patients with pathologically confirmed HCC who underwent abdominal CT with contrast enhancement. Based on postoperative pathological results, the patients were divided into an MVI-positive group (43 cases) and an MVI-negative group (47 cases), and clinical and imaging characteristics were analyzed. Flexible attenuation reduction technology was used to obtain contrast-enhanced images and iodine maps. CT values were measured in three phases of contrast-enhanced images, and iodine concentration (IC) was measured in the tumor lesions and the same level of the abdominal aorta during the arterial and venous phases in the iodine maps. Normalized iodine concentration（NIC）and portal vein phase iodine uptake reduction rates (ICrr) were calculated. Statistical methods were used to analyze the differences in clinical characteristics and quantitative SCT parameters between the two groups of patients. The diagnostic performance of each indicator, clinical and conventional CT feature model, flexible SCT parameter model, and combined model was evaluated using the receiver operating characteristic (ROC) curve. Results The differences in alpha-fetoprotein (AFP) levels and tumor maximum diameter between the two groups were statistically significant (both P < 0.001); HCC patients with peritumoral abnormal enhancement, irregular tumor morphology, and invasion of major vessels were more likely to develop MVI (both P < 0.05); The CT values in the portal vein phase, NIC-AP, and ICrr were significantly higher in the MVI-positive group than in the MVI-negative group (all P < 0.05). The combined model had the highest diagnostic performance, with an AUC of 0.920 (95% CI: 0.868-0.973), sensitivity of 79.1%, and specificity of 95.7%; followed by the flexible SCT parameter model with an AUC of 0.893 (95% CI: 0.825-0.960), sensitivity of 69.8%, and specificity of 93.6%; both models demonstrated superior predictive performance compared to clinical and conventional CT feature models (all P < 0.05). There was no statistically significant difference between the flexible SCT parameter model and the combined model (P = 0.092). Conclusion Compared with clinical and imaging features, flexible SCT features are more helpful for preoperative noninvasive assessment of MVI in hepatocellular carcinoma, among which NIC-AP has the best diagnostic performance, and the combined model can improve predictive value.

Objective To explore the performance and clinical utility of a nomogram integrating radiological and clinical features for preoperative prediction of resection margin status in pancreatic body and tail adenocarcinoma. Methods A total of 221 patients with pancreatic body and tail adenocarcinoma who underwent radical resection at our center were retrospectively included. According to pathological assessment based on the “1 mm” margin standard, patients were classified into R0 resection and margin-positive (R1/R2) resection. Patients were randomly divided into a training cohort and a validation cohort at a 7:3 ratio. Clinical and radiological features were compared between R0 and R1/R2 groups. Univariate and multivariate Logistic regression analyses were performed on variables in the training cohort to identify independent risk factors for R1/R2 resection, based on which predictive models were constructed. Receiver operating characteristic (ROC) curves were applied to evaluate the predictive performance of each model. Finally, a nomogram based on the optimal model was established to assist individualized clinical risk assessment. Results In the training cohort, tumor maximum diameter, CA19-9 level, resectability according to NCCN guidelines, peripancreatic fat stranding, rim enhancement, adjacent organ invasion, splenic artery and/or vein encasement, and the minimum distance between the tumor and the anterior and posterior surfaces of the pancreas were significantly different between the R0 and R1/R2 groups (P < 0.05). Univariate and multivariate Logistic regression analyses identified resectability, tumor maximum diameter, rim enhancement, peripancreatic fat stranding, and the minimum distance between the tumor and the anterior surface of the pancreas as independent predictors of R1/R2 resection (P < 0.05). Based on these factors, clinical, radiological, and combined models were constructed. ROC curve analysis demonstrated that the combined model integrating clinical and CT imaging features outperformed both the clinical and radiological models in predicting the resection margin status of pancreatic body and tail cancer. The AUCs of the combined model were 0.874 in the training cohort and 0.851 in the validation cohort, both higher than those of the individual models. Conclusion The combined model integrating radiological and clinical features demonstrates higher discriminative ability and clinical utility for predicting resection margin status, outperforming models based on clinical or radiological features alone.

Objective To analyze the application value of combining serum urokinase type plasminogen activator (uPAR), alpha methylacyl CoA racemase (AMACR) and transabdominal MRI imaging features in the diagnosis of prostate cancer (PCa). Methods A retrospective study was conducted on 100 suspected PCa patients admitted. After surgical and pathological confirmation, they were divided into PCa group (n=64) and benign prostatic hyperplasia (BPH) group (n=36). The serum levels of uPAR and AMACR, as well as the imaging characteristics of transabdominal MRI, were compared between the two groups. Binary Logistic regression was used to analyze the independent risk factors of PCa and a model was constructed. Receiver operating characteristic (ROC) curves were used to analyze the diagnostic value of serum levels of uPAR and AMACR for PCa. Results Age, BMI and comorbidities did not differ significantly between the two groups (P>0.05). The TIC curve of PCa group was mainly type III, while the TIC curve of BPH group was mainly type I. The comparison of TIC curve types between two groups of patients showed a statistically significant difference (P<0.05). The levels of uPAR and AMACR in the PCa group were higher than those in the BPH group, and the difference was statistically significant (P<0.05). As the degree of differentiation increases, the levels of uPAR and AMACR in PCa patients also increase; as TNM staging increases, the levels of uPAR and AMACR in PCa patients also increase, and the difference is statistically significant (P<0.05). UPAR, AMACR, and MRI are independent risk factors for PCa, and patients with high levels of uPAR, AMACR, and MRI positivity have a higher risk of PCa (P<0.05). Based on binary logistic regression analysis, a model was constructed with the formula: 0.040 * serum uPAR+0.978 * serum AMACR-0.760 * MRI-0.941. uPAR,AMACR,MRI. The AUC values for joint prediction were 0.771, 0.827, 0.732, and 0.900, with sensitivities of 0.766, 0.750, 0.667, and 0.833, and specificities of 0.694, 0.761, 0.797, and 0.844, respectively. Paired Z-test showed that the combined AUC was higher than that of individual indicators, and the sensitivity and specificity of joint prediction were optimal (P<0.05). Conclusion The levels of uPAR and AMACR in the serum of PCa patients are significantly different from those of BPH patients, and they increase significantly with the deepening of tumor differentiation and TNM staging progression. The combined application of serological uPAR, AMACR markers, and transabdominal MRI diagnostic techniques can significantly improve the sensitivity and specificity of diagnosis, and the optimal AUC value for combined diagnosis provides a new effective strategy for the precise diagnosis of PCa.

Objective To evaluate the predictive value of multiparameter MRI combined with the prostate health index (PHI) for identifying extracapsular extension (ECE) in prostate cancer. Methods Prostate cancer patients admitted were enrolled and randomly divided into training and validation sets at a 7∶3 ratio. Patients were stratified into ECE and non-ECE groups based on postoperative pathological findings following radical prostatectomy. Logistic regression was performed to identify independent predictors of ECE and to develop predictive models. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, precision-recall (PR) curves, F1_max scores, and integrated Brier scores (IBS). Results A total of 329 patients with prostate cancer were included, with a median age of 66 years (interquartile range: 64-68). ECE was observed in 21.28% (70/329) of patients. Logistic regression revealed that elevated PHI (OR=1.040, 95% CI: 1.019-1.061, P<0.001), higher extraprostatic extension (EPE) grade (OR=2.524, 95% CI: 1.454-4.381, P=0.001), and increased volume transfer constant (K^trans) (OR=31.997, 95% CI: 9.383-109.117, P<0.001) were independent risk factors for ECE. Conversely, higher apparent diffusion coefficient (ADC) values (OR=0.091, 95% CI: 0.026-0.322, P<0.001) were independently associated with a reduced risk of ECE. Model 3, incorporating EPE grade, PHI, ADC, and K^trans, demonstrated superior performance in the training set [ROC-AUC: 0.964 (95% CI: 0.943-0.985), PR-AUC: 0.886 (95% CI: 0.768-0.948), F1_max: 0.815, IBS: 0.064] and validation set [ROC-AUC: 0.962 (95% CI: 0.924-1.000), PR-AUC: 0.859 (95% CI: 0.618-0.958), F1_max: 0.857, IBS: 0.053], outperforming model 1 (EPE grade + PHI) and model 2 (EPE grade + ADC + K^trans). Conclusion The predictive model integrating EPE grade, PHI, ADC, and K^trans demonstrates excellent performance in predicting ECE and holds promise as a valuable tool for preoperative risk stratification in prostate cancer.

Objective This study aims to develop an Objective, accurate diagnostic model for pernicious placenta previa (PPP) with placenta accreta spectrum (PAS) using deep learning-based MRI radiomics. Methods This retrospective study included 124 hospitalized pregnant women from two centers (Center A and B). T₂-weighted MRI data were used to extract deep learning (DL) features via DenseNet-121 and radiomics features via PyRadiomics. After mutual information-based feature selection, a combined DL-radiomics model (Rad-Score) was constructed. Clinical features were screened using LASSO and logistic regression. The performance of the DL-radiomics model, clinical model, and combined model was evaluated. The Rad-Score cutoff was used to stratify total hospital stay using Kaplan-Meier analysis. Results Among the 124 patients, 43 had PPP with PAS; 16 had PPP without PAS. Two DL and five radiomics features formed the Rad-Score model. Ultrasound was the only independent risk factor in the clinical model. The DL-radiomics model achieved AUCs of 0.877 (95% CI: 0.814-0.939), 0.916 (95% CI: 0.844-0.989), and 0.824 (95% CI: 0.711-0.937) for the full cohort, Center A, and Center B, respectively. The ultrasound model yielded an AUC of 0.767 (95% CI: 0.690-0.844). No maternal or neonatal deaths occurred. Kaplan-Meier curves based on a Rad-Score cutoff of -0.229 showed that patients with low scores had shorter hospital stays (median 12 days) compared to high-score patients (median 16 days, P _Log-rank<0.001). Conclusion A DL-based MRI radiomics model enables accurate, Objective prediction of PPP with PAS, showing strong clinical applicability.

Objective To explore the diagnostic value of CT lymphangiography for Gorham-Stout disease. Method The clinical and imaging data of 30 patients diagnosed with Gorham-Stout disease were retrospectively collected. All patients underwent CT lymphangiography. The observation indicators mainly included: (1) The location and morphology of bone lesions; (2) Abnormal lipiodol deposition sites and ranges throughout the body such as the neck, chest, abdomen and pelvis; (3) Abnormal manifestations of CT in the lungs; (4) Other manifestations related to lymphatic vessel abnormalities. Statistical description was performed using the composition ratio of qualitative data. Results Skeletal abnormalities were observed in all 30 patients, including 28 cases of vertebrae, 26 cases of pelvic bone, 19 cases of femur, 15 cases of rib, 13 cases of humerus, 11 cases of scapula, 8 cases of sternum, and 5 cases of clavicle. According to the lesion morphology, there were 16 cases of cystic manifestation, 19 cases of tubular manifestation, 12 cases of loofah manifestation, and 6 cases of osteosclerosis manifestation. Among them, 19 patients presented with mixed manifestation. Abnormal lipiodol distribution was observed in CTL in all 30 patients. Among them, abnormal lipiodol deposition was seen in bone in 21 patients. There were 15 cases at the end of the thoracic duct, 4 cases at the end of the right lymphatic duct, 5 cases at the hilum of the lung, 1 case around the bronchovascular tract, 7 cases in the mediastinum, 3 cases in the pericardium, 5 cases around the trachea, 9 cases in the pleural region, 1 case in the axilla, 1 case around the pancreas, 1 case around the spleen, 1 case at the hilum of the liver , 1 case in the kidney, 1 case around the small intestine, 11 cases behind the peritoneum, 15 cases in the pelvic cavity, 10 cases on the abdominal and pelvic wall, 7 cases in the perineum, and 6 cases in the iliac fossa. Abnormal CT changes in the lungs included: 6 cases of ground-glass density shadows in the lungs, 2 cases of consolidation shadows, 2 cases of thickening of lobular septa, 6 cases of thickening of bronchovascular tracts, and 11 cases of atelectasis. Other lymphatic-related abnormalities: lymphatic malformation occurred in the mediastinum in 9 cases, the chest wall in 5 cases, the neck in 1 case, the abdominal and pelvic cavity in 10 cases, the retroperitoneal region in 8 cases, the abdominal and pelvic wall in 8 cases, and the spleen in 12 cases. Conclusion The imaging manifestations of CT lymphangiography in Gorham-Stout disease have certain characteristics, and are often accompanied by abnormalities of extra bone lymphatic vessels and lymphatic vessels in other organs, which can provide important imaging evidence for the diagnosis and differentiation of this disease.

Objective To investigate the MRI features of fibroma of tendon sheath and localized giant cell tumor of tendon sheath, and to improve the level of differential diagnosis. Methods MRI images of 10 patients with fibroma of tendon sheath and 28 patients with localized giant cell tumor of tendon sheath confirmed by pathology were retrospectively analyzed. The characteristics evaluated included tumor length, morphology, signal characteristics (signal homogeneity on T₁-weighted imaging, proportion of high signal, hypointense morphology and distribution, and hypointense ring on fat-suppressed T₂ or proton density weighted imaging), and surrounding tissue manifestations (bone compression and absorption, degree of wrapping tendon, and soft tissue edema). The mean signal intensity ratio of tumor to tendon was measured and compared. Inter-observer agreement was assessed using kappa statistics. Continuous variables were compared by independent samples t-test or Mann-Whitney U test based on normality (Shapiro-Wilk test). Categorical variables were analyzed with χ² or Fisher’s exact test. Statistical significance was set at P<0.05. Results Fibromas of tendon sheath were mostly round/quasi-round (n=7), mostly band-like with low signal intensity on fat-suppressed T₂ or proton density weighted imaging (n=6), and some tumors showed low signal ring at the edge (n=4). The mean signal intensity ratio of tumor to tendon was 3.81 (2.73-6.81). Most of the localized giant cell tumors of tendon sheath showed cast growth (n=17), and the low signal on fat-suppressed T₂ or proton density weighted imaging was granular/separated (n=20). The low signal shadow was mostly located at the tumor edge (n=18), and most of the tumors showed low signal ring at the tumor edge (n=23). The median signal intensity ratio of tumor to tendon was 2.78 (2.08-3.42). There were significant differences between the two groups in morphology, hypointensity on fat-suppressed T₂ or proton density weighted imaging, distribution, margin of tumor with hypointensity ring, and average signal intensity ratio of tumor to tendon. Conclusion MRI plays an important role in the differential diagnosis of fibroma of tendon sheath and localized giant cell tumor of tendon sheath. The key points to distinguish the two types of tumors include: tumor morphology, hypointensity on fat-suppressed T₂ or proton density weighted imaging, distribution and hypointensity of the ring, and the average signal intensity ratio of tumor to tendon.

Objective A deep learning model based on radiograph(s) of the shoulder joint was constructed to evaluate the degree of rotator cuff tear. Methods The plain radiograph(s) data of the shoulder joint (including 613 cases of anteroposterior images and 524 cases of outlet view radiographs) and corresponding clinical information (age, gender and trauma history) of 1137 patients with rotator cuff injury were collected retrospectively. Based on the MRI or arthroscopic examination results of the patients, the supraspinatus tendon tears were divided into two groups, namely no/mild tears and severe tears, according to whether they were > 50%, to construct a deep learning model. The classification performance of the model was evaluated by using the area under the receiver operating characteristic curve (AUC), accuracy rate, sensitivity and specificity. Results Among 1137 patients with rotator cuff injuries, 719 cases had rotator cuff tears less than 50% and 418 cases had tears more than 50%. There was no statistically significant difference in clinical information (age, gender and trauma history) between the two groups. The deep learning model performance constructed based on the anteroposterior shoulder joint radiographs: the accuracy rate was 93.9%, the AUC was 0.981, the sensitivity was 96.4%, and the specificity was 91.5%; The model performance constructed based on the acromion outlet view shoulder joint radiographs : the accuracy rate was 79.2%, the AUC was 0.808, the sensitivity was 74.4%, and the specificity was 82.8%. The heat map shows that the subacromial area and the acromial area are the most sensitive areas. Conclusion The constructed deep learning model can evaluate the degree of supraspinatus tendon tear based on the changes in the subacromial area and acromion on radiograph(s).

Objective To explore the feasibility of constructing a prediction model for the severity of mycoplasma pneumoniae pneumonia in children based on CT radiomics. Methods A retrospective analysis was conducted on cases of mycoplasma pneumoniae pneumonia diagnosed in children who were hospitalized from January 2023 to July 2024 at the maternal and child health hospital of Guangxi Zhuang Autonomous Region (Center 1,n=424),the First affiliated hospital of Guangxi medical university (Center 2, n=84),and Nanning maternal and child health care hospital (Center 3, n=80).Clinical information, laboratory data, and CT morphological features of all children were collected.Clinical data with P<0.05 were included in the multivariate logistic regression analysis to determine the independent risk factors for severe mycoplasma pneumoniae pneumonia and construct a clinical prediction model. CT images were segmented using a semi-automatic method, and those identified as incomplete or inaccurate by AI were manually adjusted. Radiomics features were extracted and screened to determine the radiomics markers most relevant to severe mycoplasma pneumoniae pneumonia. Prediction models based on CT radiomics and the combination of clinical and radiomics features were constructed, and their performance was evaluated using the area under the curve (AUC) and accuracy. Results The AUC of the prediction model based on CT radiomics features was 0.8382.Multivariate Logistic regression analysis showed that shortness of breath, D-dimer, and CT scores for lung consolidation, ground-glass opacity, atelectasis, and enlarged mediastinal and hilar lymph nodes were independent risk factors for severe mycoplasma pneumoniae pneumonia. The combined model of CT radiomics and clinical features demonstrated the best predictive performance, with AUCs of 0.9005, 0.7603, and 0.8051 in the training set, external validation set 1, and external validation set 2, respectively. Conclusion A logistic regression model constructed by integrating clinical, laboratory characteristics, CT morphological features and omics features demonstrated good performance in predicting the severity of mycoplasma pneumoniae pneumonia in children,which is expected to provide an Objective risk stratification tool for front-line doctors and manage patients more effectively.

Objective Exploring the value of a U-Net deep learning model based on chest CT images for diagnosing mycoplasma pneumoniae pneumonia(MPP)in children. Methods A retrospective analysis was conducted on the pre-treatment CT images and clinical data of 434 pediatric pneumonia patients(220 MPP cases and 214 non-MPP cases).The data were stratified and sampled in a 7∶3 ratio to form the training set and validation set.The U-Net deep learning algorithm was used to establish separate clinical model,imaging model,and combined clinical and imaging model.The diagnostic performance,reliability,and patient benefits of these models were evaluated using the area under the receiver operating characteristic curve(AUC),calibration curves, and decision curves(DCA). Results Clinical model,imaging model,and combined clinical-imaging model based on the U-Net deep learning algorithm can all be used for diagnosing MPP.Compared with clinical models and imaging models,the combined clinical-imaging model demonstrates the highest diagnostic performance,with an AUC of 0.873(95%CI:0.831-0.909)in the training set,with sensitivity and specificity of 88.96% and 90.67%,respectively.In the validation set,the AUC was 0.839(95%CI:0.765-0.898),with sensitivity and specificity of 87.88% and 85.94%,respectively. Conclusion A U-Net deep learning model based on chest CT is helpful for the diagnosis of MPP in children.

Objective To investigate the efficacy and safety of transarterial chemoembolization（TACE）sequentially combined with bevacizumab and sintilimab in the treatment of intermediate and advanced hepatocellular carcinoma（HCC）,and to analyze the risk factors associated with adverse events. Methods The clinical data of 55 patients with intermediate and advanced HCC who underwent TACE followed by sequential therapy with bevacizumab and sintilimab were collected,and the efficacy and safety of this treatment were analyzed.The perioperative adverse reactions were analyzed by measurement of tumor-feeding artery diameter and other information with digital subtraction angiography（DSA）during TACE. Results Our results demonstrated that the TACE sequential therapy with bevacizumab and sintilimab achieved an Objective response rate（ORR）of 65.3%,and the disease control rate（DCR）was 86.9%.The median progression-free survival（mPFS）was 8.2 months,and a median overall survival（mOS）was 22.5 months.Among the 55 patients,most adverse events were grade 1-2,with 2 cases of grade 3 immune-related adverse events and 1 case of major gastrointestinal bleeding.No fatal adverse events occurred.The diameter of the tumor-feeding hepatic artery is a potential predictive factor for perioperative adverse events. Conclusion sThis study provides evidence that TACE sequential therapy with bevacizumab and sintilimab is a safe and effective treatment option for intermediate and advanced-stage HCC.The diameter of the tumor-supplying hepatic artery may predict treatment-related adverse events associated with this combination therapy.

Objective To explore the technical experience of C-type ipsilateral oblique 40°+cephalothoracic 20° radiography in uterine artery embolization. Methods 1683 cases of uterine artery catheterization in our hospital from June 2012 to June 2025 were collected,including 251 cases of cervical cancer,483 cases of scar pregnancy,448 cases of postpartum hemorrhage,211 cases of adenomyosis and 290 cases of uterine fibroids.According to the choice of catheterization methods,they were divided into traditional group,rotation group and experimental group.The traditional group was cannulated by DSA in positive oblique position to show the opening of uterine artery.The rotation group was cannulated by rotating DSA to find the opening of uterine artery.The experimental group was cannulated by C arm ipsilateral oblique 40°+cephalothoracic 20° DSA to find the opening of uterine artery. Results Uterine arterial embolization/chemoembolization was successfully performed in all 1683 cases.The operation time,fluoroscopy time,contrast medium dosage,radiation dose and uterine artery catheterization time in the experimental group were shorter than those in the rotation group and the traditional group. Conclusion C-type 40° ipsilateral+20° cephalothoracic angiography technique in uterine artery embolization has shorter operation time,lower radiation dose and better protection for patients and doctors.

Objective To investigate the diagnostic value of non-enhanced magnetic resonance lymphography（MRL）and ^99mTc-DX lymphography in primary Sjögren's syndrome（pSS）combined with lymphatic reflux disorder. Methods The clinical and imaging data of 14 patients with pSS were retrospectively collected,and the characteristics of MRL and ^99mTc-DX lymphography were analyzed in all patients. Results There were 9 cases of thoracic duct outlet（left venous angle area）obstruction,and 4 cases of bilateral venous angle drainage of thoracic duct with outlet obstruction.There was no statistically significant difference between MRL and ^99mTc-DX lymphatic imaging in the demonstration of thoracic duct outlet obstruction（left venous angle area）,bilateral venous angle drainage of thoracic duct with outlet obstruction,tortuous dilatation of thoracic segment of thoracic duct,abnormal demonstration of broncho-mediastinal trunks,and bilateral dilatation of iliac lymphatic ducts and lumbar trunks.The Kappa values were 1.000,1.000,0.632,0.632,0.440,and the diagnostic consistency was almost perfect,almost perfect,good,good,good,and moderate,respectively.MRL was superior to ^99mTc-DX lymphatic imaging in the display of anomalies of the infraclavicular trunk,and the difference was statistically significant（P<0.05）.^99mTc-DX lymphatic imaging showed 3 cases of bilateral lower extremity lymphedema,1 case of subcutaneous lymphedema in the buttocks and perineum;8 cases of celiac pleural effusion,3 cases of celiac peritoneal effusion,1 case of celiac pleural effusion with peritoneal effusion,and 1 case of celiac pericardial effusion. Conclusion The combined application of MRL with ^99mTc-DX lymphography is valuable in the diagnosis of pSS combined with lymphatic reflux disorders.