Objective: To explore the efficacy of a novel injectable hydrogel (GelMA/P11/IL4@LIP) loaded with P11 bacteriophages and interleukin-4 (IL-4) liposomes (LIP) in preventing relapse after maxillary expansion in mice, providing experimental evidence for its clinical application. Methods: This study was approved by the experimental animal ethics committee of our hospital. First, 15 7-week-old C57BL/6 mice were used to establish a maxillary expansion model and divided into 5 groups (3 mice in each group): a control group, post expansion day 3 group (PED3 group), post expansion day 7 group (PED7 group), retention for 14 days group (RET group), and relapse for 7 days group (REL group). The mice in each group were sacrificed at their designated time points (day 0, 3, 7, 21, 28), and their maxilla and anterior cranial regions were collected. Bone parameters and the inter-crestal distance (ICD) of maxillary incisor mesial alveolar ridge were measured using micro-computed tomography (micro-CT). Histological staining was performed to evaluate bone formation and resorption, while immunohistochemistry (IHC) was performed for macrophage markers (CD86 and CD206), mesenchymal stem cell markers (glioma-associated oncogene homolog 1 [Gli1]), and osteogenic markers (Runt-related transcription factor 2 [Runx2] and Osterix [OSX]). Next, GelMA/P11/IL4@LIP was synthesized and administered to mouse models of maxillary expansion. A total of 24 7-week-old C57BL/6 mice were divided into 4 groups (6 mice in each group): a blank control group, GelMA group, GelMA/P11 group, and GelMA/P11/IL4@LIP group. All mice underwent palatal expansion. On PED7, the expanders of all 24 mice were cemented with resin to initiate the 14-day retention period. On day 1 of the retention phase, the mice in each group received injections of saline, GelMA, GelMA/P11, or GelMA/P11/IL4@LIP at the midpalatal suture. After the 14-day retention period, three mice in each group were randomly selected and sacrificed, while the other three had their expanders removed and underwent a 7-day relapse before being sacrificed on day 28 (REL). Micro-CT, histological staining, and IHC were performed to evaluate the preventive effect of GelMA/P11/IL4@LIP on post-expansion relapse. Results: The mice maxillary expansion model exhibited a decreased ICD at REL compared to RET in micro-CT analysis (P = 0.008). IHC analysis demonstrated prolonged M1 macrophage infiltration, scarce Gli1+ mesenchymal stem cells, and insufficient expression of osteogenic markers (RUNX2 and OSX) (P < 0.001). Compared to the blank control and GelMA groups, GelMA/P11/IL4@LIP hydrogel injection in the midpalatal suture led to increased ICD at REL, promoted the timely M2 polarization of macrophages, recruited Gli1+ mesenchymal stem cells, and upregulated the expression of RUNX2 and OSX (P < 0.05). Conclusion The mechanism of relapse after maxillary expansion involves the persistent infiltration of M1 macrophages, as well as the inadequate recruitment and insufficient osteogenic differentiation of MSCs in the midpalatal suture. The GelMA/P11/IL4@LIP composite enhanced orofacial mesenchymal stem cell recruitment and promoted the M2 polarization of macrophages, thereby enhancing osteogenesis in the midpalatal suture and preventing post-expansion relapse.

Objective:To develop and validate machine learning-based radiomics models using preoperative CT images for individualized prediction of mitotic index (MI) in patients with gastrointestinal stromal tumors (GIST).Methods:The study was a case-control study. The data of 348 GIST patients confirmed by pathology were retrospectively collected from two independent medical centers: the Affiliated Hospital of Qingdao University (center 1) and Shandong Provincial Hospital Affiliated to Shandong First Medical University (center 2), covering the period from January 2013 to June 2018. Patients from center 1 were divided into a training cohort (176 cases) and an internal validation cohort (75 cases) at a ratio of 7∶3 using random sampling. Patients from center 2 served as an independent external validation cohort (97 cases). The primary endpoint was MI, categorized into high MI (145 cases) and low MI (203 cases) groups. Radiomic features were extracted from the portal venous phase images of preoperative contrast-enhanced CT scans. Five machine learning algorithms, including logistic regression, support vector machine, random forest, decision tree, and extreme gradient boosting (XGBoost),were employed to construct MI prediction models. The optimal model was identified using receiver operating characteristic curves. An individualized prediction model was developed by integrating the the optimal machine learning model combined with selected independent clinical factors, and the importance of features was visualized using Shapley Additive Explanation (SHAP) analysis. Patients were followed up, and Kaplan-Meier curves along with log-rank tests were used to evaluate recurrence-free survival (RFS) differences between the predicted high MI and low MI groups.Results:Among the five constructed machine learning models, the XGBoost model demonstrated the best predictive performance, with area under the curve (AUC) of 0.809 (95% CI 0.738-0.872), 0.693 (95% CI 0.571-0.809), and 0.718 (95% CI 0.605-0.822) in the training cohort, internal validation cohort, and external validation cohort, respectively. An individualized prediction model combining the XGBoost model with independent clinical factors (tumor location and tumor size) was developed. The model achieved AUC of 0.843 (95% CI 0.785-0.899), 0.791 (95% CI 0.680-0.894), and 0.777 (95% CI 0.678-0.861) in the training cohort, internal validation cohort, and external validation cohort, respectively. SHAP analysis indicated that radiomic features had the highest predictive impact. In both the training cohort and internal validation cohort, the RFS of patients predicted to be in the high MI group was lower than that of the low MI group, with statistically significant differences ( χ2=14.58, 9.52, both P<0.001). However, there was no statistically significant difference in RFS in the external validation set ( χ2=6.18, P=0.080). Conclusions:The optimal XGBoost model based on radiomic features extracted from preoperative portal venous phase CT images, when combined with clinical factors, can effectively predict the MI of GIST patients.

Objective:To investigate the value of habitat radiomic features based on gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA)-enhanced MRI in establishing a predictive model for cytokeratin 19 (CK19) expression in hepatocellular carcinoma (HCC) and to evaluate its role in prognostic risk stratification.Methods:This multicenter case-control study retrospectively enrolled 489 patients with pathologically confirmed HCC who underwent Gd-EOB-DTPA-enhanced MRI between June 2016 and June 2024. Among them, 346 patients from the First Affiliated Hospital of Soochow University were divided into a training cohort ( n=245) and an internal test cohort ( n=101) via stratified sampling at a 7∶3 ratio. And 143 patients from Nantong Third Hospital Affiliated to Nantong University served as an external validation cohort. The training cohort included 53 CK19-positive and 192 CK19-negative patients. The internal test cohort included 21 CK19-positive and 80 CK19-negative patients. The external validation cohort included 30 CK19-positive and 113 CK19-negative patients. Univariate logistic regression analysis was performed to identify potential factors associated with CK19 expression, and a clinical-radiologic model was constructed. The k-means clustering algorithm was applied to segment target HCC lesions into 3 subregions. Radiomic features were extracted and selected from these habitat subregions. Habitat radiomics models were constructed for the arterial phase (AP), portal venous phase, hepatobiliary phase (HBP), and combined phases (CP). Multivariate logistic regression analysis identified independent clinical and radiologic predictors of CK19 expression, and the optimal habitat model score was integrated to build a clinical-radiologic-habitat combined model. The area under the receiver operating characteristic curve (AUC) was used to evaluate model predictive performance. Recurrence-free survival (RFS) was analyzed using the Kaplan-Meier method and the differences in survival curves were compared with the log-rank test. Results:Univariate logistic regression analysis revealed that alpha-fetoprotein (AFP) ( OR=2.629, 95% CI 1.412-4.896, P=0.002), AP enhancement ( OR=3.636, 95% CI 1.642-8.052, P=0.001), AP peritumoral enhancement ( OR=2.219, 95% CI 1.084-4.542, P=0.029), and HBP peritumoral hypointensity ( OR=2.010, 95% CI 1.004-4.021, P=0.049) were potential factors associated with CK19 expression, which were incorporated into the clinical-radiologic model. In the internal and external validation cohorts, the AUC of the clinical-radiologic model was 0.690 (95% CI 0.590-0.778) and 0.650 (95% CI 0.565-0.727), respectively. The habitat radiomics model based on CP images demonstrated the highest performance. It achieved AUC of 0.729 (95% CI 0.622-0.836) and 0.725 (95% CI 0.607-0.842) in the internal and external validation cohorts, respectively. Multivariate analysis identified AFP ( OR=2.494, 95% CI 1.163-5.348, P=0.019), AP enhancement ( OR=5.230, 95% CI 1.868-14.643, P=0.002) and habitat radiomics model score ( OR=4.105, 95% CI 2.643-6.368, P<0.001) as independent predictors of CK19 positivity. Based on these factors, a combined clinical-radiologic-habitat combined model was established. The clinical-radiologic-habitat combined model achieved AUCs of 0.767 (95% CI 0.671-0.846) and 0.730 (95% CI 0.649-0.801) in the internal and external validation cohorts, respectively. Significant differences in RFS were observed between the CK19-positive group (25.1 month) and CK19-negative group (51.0 month) as predicted by the clinical-radiologic-habitat model ( χ2=4.17, P=0.041). Conclusion:The clinical-radiologic-habitat combined model based on Gd-EOB-DTPA-enhanced MRI habitat radiomics demonstrates good predictive performance for CK19 expression in HCC and offers valuable prognostic stratification for clinical practice.

Objective:To explore the value of the deep learning model based on gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA) enhanced MRI in preoperatively predicting vessels completely encapsulating tumor clusters (VETC) in hepatocellular carcinoma (HCC).Methods:This study adopted a case-control design to retrospectively analyze 420 patients with HCC confirmed by postoperative pathology who underwent Gd-EOB-DTPA enhanced MRI between June 2016 and March 2023. A total of 420 patients were divided into a training set ( n=305) from the First Affiliated Hospital of Soochow University and an external validation set ( n=115) from Affiliated Nantong Hospital 3 of Nantong University. Based on postoperative pathological findings, patients were stratified into VETC-positive and VETC-negative groups. The training set comprised 161 VETC-positive cases and 144 VETC-negative cases, while the external validation set included 55 VETC-positive cases and 60 VETC-negative cases. Tumor regions of interest in arterial, portal venous, and hepatobiliary phases were manually delineated using ITK-SNAP software. Pre-trained Vgg19, Densenet121, and Vision Transformer (ViT) models were employed for transfer learning, extracting deep learning features from each image. Feature data were processed using FAE software, and 12 logistic regression models (arterial phase, portal venous phase, hepatobiliary phase, and combined three-phase models) were constructed to select the optimal deep learning model. Independent predictors in clinical characteristics were identified through univariate and multivariate logistic analyses to establish a clinical model for predicting VETC pattern. Subsequently, a clinical-deep learning fusion model was developed by integrating these clinical predictors with the optimal deep learning features. Model performance in predicting VETC-positive HCC was evaluated using receiver operating characteristic curves, calibration curves, and decision curve analysis (DCA). Results:In the external validation set, the area under the curve (AUC) of the Vgg19 model in the arterial phase, portal venous phase, hepatobiliary phase, and combined three-phase, respectively were 0.799,0.756,0.789,0.821, which were higher than those of Densenet121 (AUC: 0.544,0.581,0.544,0.583) and ViT (AUC: 0.740,0.752,0.785,0.767) model. The three-phase combined Vgg19 model achieved the highest AUC of 0.821 (95% CI 0.746-0.897). Multivariate logistic regression identified alpha-fetoprotein level ( OR=1.826,95% CI 1.069-3.120, P=0.028) and tumor diameter ( OR=1.329,95% CI 1.206-1.466, P<0.001) as independent predictors of VETC-positive HCC, forming the clinical model with an AUC of 0.789 (95% CI 0.703-0.859). The clinical-deep learning fusion model further achieved the AUC of 0.825 (95% CI 0.749-0.900). Calibration curves confirmed high concordance between predicted and actual probabilities for the three-phase Vgg19 model, while DCA revealed greater net clinical benefit for the combined Vgg19 and fusion models compared with the clinical model alone. Conclusions:The deep learning model based on Gd-EOB-DTPA enhanced MRI can be used to predict VETC of HCC preoperatively, among which the three-phase combined Vgg19 model and the clinical-deep learning model provide high predictive value.

Objective:To investigate the value of coronary CTA (CCTA)-based traditional features and radiomics of plaque in the identification of culprit lesions that caused acute myocardial infarction (AMI).Methods:This was a retrospective multicenter study. From July 2016 to November 2023, a total of 344 patients from the Affiliated Hospital of Qingdao University (training cohort, n=184), Shandong Provincial Hospital Affiliated to Shandong First Medical University (validation cohort, n=88) and Qilu Hospital of Shandong University (test cohort, n=72) who received percutaneous coronary intervention (PCI) due to AMI and underwent CCTA within 48 hours of AMI were enrolled. The culprit plaques and non-culprit plaques were identified using a combination of electrocardiogram, CCTA, and angiographic findings. The vessel, plaque location, plaque type, Coronary Artery Disease-Reporting and Data System (CAD-RADS) score, high-risk plaque characteristics, plaque length, plaque volume, and burden were analyzed, and 1 904 radiomics features were extracted for each plaque. The traditional imaging model, the radiomics model, and the combined model were established by using multivariate Logistic regression analysis. The area under the receiver operating characteristic curve (AUC) was used to evaluate the performance of each model in identifying culprit lesions. The DeLong test was used for the comparison of AUC between every two models. The net reclassification index (NRI) was used to evaluate the incremental value of the combined model to the traditional imaging model and the radiomics model. The decision curve analysis (DCA) was used to assess the clinical net benefit of these models. A correlation heatmap was used to evaluate the correlation between the radiomics score and traditional CCTA factors. The interpretable analysis of the decision process of the combined model was performed by the Shapley Additive exPlanations (SHAP). Results:In the validation cohort and the test cohort, the AUC of the traditional imaging model developed by the vessel, plaque type, positive remodeling and CAD-RADS score was 0.898 (95% CI 0.869-0.922) and 0.881 (95% CI 0.848-0.910), respectively. The radiomics model developed by six radiomics features was 0.863 (95% CI 0.831-0.891) and 0.863 (95% CI 0.827-0.864), respectively. The AUC of the combined model was 0.930 (95% CI 0.905-0.950)and 0.919 (95% CI 0.889-0.942), respectively. In the validation cohort and the test cohort, the AUC of the combined model was higher than that of the traditional imaging model ( Z=4.013, 4.272, P<0.001) and that of the radiomics model ( Z=4.819, 3.784, P<0.001), respectively. In the validation cohort, the combined model yielded an NRI of 20.43% (95% CI 10.43%-30.44%, P<0.001) and 20.21% (95% CI 9.62%-30.80%, P<0.001) for identifying culprit lesions compared with the traditional imaging model and the radiomics model, respectively. In the test cohort, the combined model yielded an NRI of 28.05% (95% CI 16.72%-39.38%, P<0.001) and 23.57% (95% CI 13.58%-33.56%, P<0.001) for identifying culprit lesions compared with the traditional imaging model and the radiomics model, respectively. DCA showed the combined model had the highest clinical net benefit. The correlation heatmap showed the radiomics score was not correlated or only weakly correlated with traditional CCTA factors. SHAP indicated the radiomics and CAD-RADS score contributed significantly to the model. Conclusion:The CCTA-based traditional features and radiomics of plaque have favorable performance for the identification of culprit plaques in patients with AMI.

Objective:To construct a combined prediction model based on dual-layer detector spectral CT radiomics features and quantitative parameters, and to evaluate its value in preoperative prediction of serosal invasion in gastric cancer.Methods:This case-control study retrospectively analyzed data from 253 gastric cancer patients confirmed by postoperative pathology at the First Affiliated Hospital of Soochow University (Center 1) and Changshu No.2 People′s Hospital (Center 2) from January 2022 to December 2023. Patients from Center 1 ( n=157) were randomly divided into training set ( n=110) and test set ( n=47) in a 7∶3 ratio, while patients from Center 2 ( n=96) served as an external validation set. Based on postoperative pathological serosal invasion status, patients were classified into serosal invasion group ( n=164) and non-serosal invasion group ( n=89), with distributions of 70/40, 30/17, and 64/32 in the training, test, and external validation sets, respectively. Spectral CT quantitative parameters, including arterial and venous phase iodine concentration (IC), normalized iodine concentration (NIC), arterial-venous IC differences, arterial-venous NIC differences (NIC pa), arterial enhancement fraction (AEF), and effective atomic number (Z eff), were measured. Radiomics features were extracted from venous-phase 40 keV monochromatic images. The least absolute shrinkage and selection operator (LASSO) algorithm was used for feature selection. The logistic regression classifier (LR-LASSO) was applied to construct the radiomics model. Univariate and multivariate logistic regression analyses identified independent risk factors for serosal invasion, including the radiomics signature (RadScore) and quantitative parameters. A clinical model was built using significant quantitative parameters, and a combined model integrated RadScore. An artificial model was based on cT4 staging assessed by two radiologists using venous-phase CT. The diagnostic performance of each model was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA). Results:A total of six radiomics features were selected to establish the radiomics model. RadScore ( OR=7.598, 95% CI 2.259-25.562, P=0.001) and NIC pa ( OR=4.598, 95% CI 1.404-15.050, P=0.012) served as independent risk factors. The NIC pa served as the clinical model. The AUCs (95% CI) of the combined model in the training, test, and external validation sets were 0.984 (0.969-1.000), 0.855 (0.728-0.982), and 0.773 (0.665-0.882), respectively. The AUCs of the artificial model were 0.741, 0.670, 0.644; of the clinical model were 0.709, 0.633, 0.626. The AUCs of the radiomics model were 0.963, 0.824, 0.741, respectively. Calibration curves showed good agreement between predicted probability and observed probability. The DCA confirmed higher clinical net benefits for the combined model. Conclusion:The combined model integrating dual-layer detector spectral CT radiomics features and quantitative parameters exhibits high efficacy for preoperative prediction of gastric cancer serosal invasion.

Objective To investigate the diagnostic value of multimodality-enhanced CT-based radiomics nomogram for muscle-inva-sive bladder urothelial carcinoma.Methods A retrospective analysis was performed on the preoperative data of 644 patients with pathologically confirmed bladder urothelial carcinoma from three medical centers.Region of interest(ROI)were drawn on preopera-tive contrast-enhanced CT images,and radiomics features were extracted.Patients from medical center 1 were randomly divided into training set and internal validation set in a 7∶3 ratio,while patients from medical centers 2 and 3 were combined as an external val-idation set.The diagnostic performance of the models was evaluated using receiver operating characteristic(ROC)curve.Results In the external validation set,the area under the curve(AUC)for diagnosing muscle-invasive bladder urothelial carcinoma using the multi-phase fusion radiomics model was 0.861[95％ confidence interval(CI)0.811-0.911].The nomogram constructed by combi-ning the multi-phase fusion radiomics model with clinical factors achieved an AUC of 0.901(95％ CI 0.862-0.939).Conclusion The nomogram combining multimodality-enhanced CT-based radiomics with clinical factors can effectively diagnose muscle-invasive bladder urothelial carcinoma.

Objective To investigate the diagnostic value of multimodality-enhanced CT-based radiomics nomogram for muscle-inva-sive bladder urothelial carcinoma.Methods A retrospective analysis was performed on the preoperative data of 644 patients with pathologically confirmed bladder urothelial carcinoma from three medical centers.Region of interest(ROI)were drawn on preopera-tive contrast-enhanced CT images,and radiomics features were extracted.Patients from medical center 1 were randomly divided into training set and internal validation set in a 7∶3 ratio,while patients from medical centers 2 and 3 were combined as an external val-idation set.The diagnostic performance of the models was evaluated using receiver operating characteristic(ROC)curve.Results In the external validation set,the area under the curve(AUC)for diagnosing muscle-invasive bladder urothelial carcinoma using the multi-phase fusion radiomics model was 0.861[95％ confidence interval(CI)0.811-0.911].The nomogram constructed by combi-ning the multi-phase fusion radiomics model with clinical factors achieved an AUC of 0.901(95％ CI 0.862-0.939).Conclusion The nomogram combining multimodality-enhanced CT-based radiomics with clinical factors can effectively diagnose muscle-invasive bladder urothelial carcinoma.

Objective To evaluate the value of biparametric MRI(bpMRI)radiomics for predicting postoperation Gleason score(GS)upgrade of prostate cancer(PCa).Methods Totally 344 PCa patients who underwent radical prostatectomy(RP)were retrospectively enrolled and divided into training set(n=241)and test set(n=103)at a ratio of 7∶3.T2WI,diffusion weighted imaging(DWI)and apparent diffusion coefficient(ADC)map radiomics signatures were constructed based on preoperative bpMRI,respectively,then logistic regression(LR)algorithm was used to establish bpMRI radiomics model.Univariate and multivariate logistic regression analyses were performed to screen independent risk factors for postoperation GS upgrade of PCa,and a clinical model was constructed.Then a clinical-radiomics combined model was established based on clinical model and bpMRI radiomics model.Receiver operating characteristic curves were drawn,the area under the curves(AUC)were calculated to evaluate the efficacy of each model for predicting postoperation GS upgrade of PCa.Results Elevated preoperative prostate imaging reporting and data system(PI-RADS)score and reduced biopsy Gleason grade group(GG)were both independent risk factors of postoperation GS upgrade of PCa(both P＜0.05).The AUC of bpMRI radiomics model and clinical-radiomics combined model for predicting postoperation GS upgrade of PCa were higher than that of single-sequence radiomics signatures and clinical model(all P＜0.05),while no significant difference was found between the former two(P＞0.05).The clinical-radiomics combined model demonstrated good efficacy for predicting postoperation GS upgrade of PCa with different biopsy GG before operation,with AUC ranging from 0.835 to 0.949 in training set and 0.803 to 0.948 in test set.Conclusion bpMRI radiomics model could effectively predict postoperation GS upgrade of PCa.

Objective To evaluate the diagnostic value of dual-layer spectral CT(DLCT)virtual non-contrast(VNC)imaging combined with iodine maps in differentiating early post-endovascular therapy(EVT)intracranial hemorrhage from contrast extravasation.Methods Retrospective analysis of 97 patients who underwent DLCT immediately after EVT was conducted.Taking 24-hour follow-up CT/MRI as the gold standard,patients were divided into hemorrhage and non-hemorrhage groups,and their clinical data were compared.VNC CT values and iodine concentration(IC)were measured.Spearman's rank correlation was used to analyze the relationship between VNC CT and IC values,and ROC curve analysis using R software to evaluate the diagnostic performance of VNC,iodine maps,and their combination.Results Among 97 patients,51(52.6%)showed no intracranial hyperdense lesions,while 46(47.4%)with abnormal densities were analyzed.Using 24-hour postoperative CT/MRI as reference stan-dard,among the 46 patients ultimately included in the analysis,38 cases(82.6%)were non-hemorrhagic and 8 cases(17.4%)hemorrhagic.No significant differences existed in age,sex,or treatment methods(all P>0.05).VNC CT values and IC showed significantly negative correlation(r=-0.537,P<0.01).ROC analysis revealed AUCs of 0.917(95%CI:0.786～0.999)for VNC,0.878(95%CI:0.719～0.999)for IC,and 0.919(95%CI:0.812～0.999)for the combination of the two(P<0.05 for combined vs.individual methods).Optimal thresholds were 53.6 HU for VNC and 0.605 mg/ml for IC.Based on the final analysis of 46 enrolled patients,the sensitivity of VNC,iodine map,and their combination in differentiating early cerebral hemorrhage from contrast extravasation was 88.9%,94.3%,and 91.4%,respectively;the specificity 94.3%,77.8%,and 88.9%,respectively;and the accuracy 90.9%,90.9%,and 93.2%,respectively.Conclusion The DLCT VNC-iodine map combination significantly im-proves differentiation between post-EVT hemorrhage and contrast extravasation,and it is recommended for routine clinical application.