ObjectiveThis paper aims to conduct the feature analysis and correlation analysis on the ocular collateral features and differential metabolites in patients with chronic hepatitis B （CHB） complicated by glucose metabolism disorder （GMD），particularly those with the damp-heat syndrome type，by integrating shadowless scleral imaging and metabolomics technologies. MethodsA total of 313 patients were recruited from the Hepatology and Endocrinology Outpatient Departments of Public Health Clinical Center of Chengdu according to the inclusion/exclusion criteria，and they were divided into a CHB group and a CHB complicated by GMD groups （damp-heat syndrome group and non-damp-heat syndrome group）. All patients underwent high-definition ocular image acquisition and feature extraction using an intelligent analysis system for shadowless scleral imaging to analyze the differences in the counting of morphological feature scores of ocular collaterals among groups. By using a digital sampling method，24 patients from each group were randomly selected，along with 20 healthy volunteers，for untargeted metabolomic analysis of peripheral serum. Differential metabolites were identified，statistically analyzed，and subjected to potential biomarker analysis and pathway enrichment. Spearman method was performed to conduct the correlation analysis on the differential ocular collateral features and differential metabolites，followed by correlation network construction. ResultsCompared with those in the CHB group，patients with CHB complicated by GMD showed significant changes in ocular collateral feature scores such as "hillock"，"blood vessels"，and "pale dusky coloration" （P<0.05）. In comparison with those in the healthy group，metabolites including N-acetylglucosamine，acetylhomoserine，and myo-inositol （AUC>0.7） were identified as potential biomarkers for the disease. Compared with those in the CHB complicated by GMD group with non-damp-heat syndrome，patients with damp-heat syndrome exhibited significant changes in feature scores of "plaques"，"yellow coloration"，"spleen"，and "gallbladder" （P<0.05）. In comparison with those in the healthy group，metabolites such as O²′-4a-cyclic tetrahydrobiopterin，theobromine，xanthurenic acid，and L-glutamic acid 5-phosphate （AUC>0.7） were identified as potential biomarkers for the damp-heat syndrome type. The Spearman correlation analysis reveals weak to moderate linear correlations between the differential scleral collateral features and metabolites. By constructing a "disease-syndrome" network of ocular diagnosis and metabolites，"xanthurenic acid-gallbladder" and "theobromine-plaque/yellow coloration" were identified as specific molecular-phenotypic correlated biomarker clusters for CHB complicated by GMD with dampness-heat syndrome. ConclusionPatients with CHB complicated by GMD demonstrate differential ocular diagnostic features and serum metabolites corresponding to disease states and dampness-heat syndrome. These objective biomarkers can guide both clinical syndrome differentiation and medication. The macro-micro integration based on ocular feature clusters and potential metabolic biomarkers offers an innovative approach to a combined traditional Chinese and Western medicine diagnosis and treatment model for this disease.

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 denoising performance of different deep learning (DL) strategies on low-dose brain 18F-FDG PET images. Methods:This retrospective methodological study was conducted on brain PET/CT images of 50 patients (35 males, 15 females, age 20-87 years) who received 3.7MBq/kg 18F-FDG at the Fifth Affiliated Hospital of Sun Yat-sen University between May 2023 and January 2024. Full-dose PET data were acquired with 2min scan. CT scans were acquired before PET scanning. Low-dose PET sinograms were generated by down-sampling the full-dose list mode data to 1/2, 1/4, and 1/20 of full-dose count level. Both full-dose and low-dose sinograms were reconstructed with random, CT-based attenuation and scatter corrections using the three-dimensional (3D) ordered-subsets expectation maximization (OSEM) algorithm (2 iterations, 20 subsets). A total of 4 DL denoising methods were established: (1) 3D conditional generative adversarial networks (GAN) using only low-dose PET as input (GAN-1); (2) 3D attention-based GAN (AttGAN) with low-dose PET input (AttGAN-1); (3) 3D AttGAN with low-dose PET and CT inputs (AttGAN-2); (4) 3D AttGAN with frequency-separation using low-dose PET and CT inputs (AttGAN-FS-2). For AttGAN-FS-2, during the frequency division process, high- and low-frequency components were extracted from the PET reconstructed images via Fourier transform, then inversed Fourier transform, denoised separately, and finally combined to produce the final denoised images. The dataset was separated into training (70%), validation (10%) and testing (20%) sets using simple random sampling without replacement with a fixed random seed. A 5-fold cross-validation scheme was then applied to test all 50 patients. Performance was evaluated against full-dose PET using normalized mean square error (NMSE), structural similarity (SSIM), peak signal-to-noise ratio (PSNR), contrast-to-noise ratio (CNR), SUV mean and SUV max bias of selected brain ROIs. Wilcoxon signed rank test was used to analyze the differences between the denoising methods. Results:AttGAN-FS-2 showed the best performance among all dose levels, with statistical difference as compared by low-dose PET and GAN-1 denoised images for NMSE, SSIM, PSNR, and CNR ( Z values: 2.92-6.15, all P<0.005). NMSE, SSIM quantitative evaluation results (median) of each model at 1/20 dose were: GAN-1: 0.08, 0.87, AttGAN-1: 0.08, 0.88, AttGAN-2: 0.07, 0.89, AttGAN-FS-2: 0.06, 0.91, respectively ( Z values: 3.24-5.77, all P<0.005). Conclusion:The DL-based method combined with multiple strategies AttGAN-FS-2 shows improved denoising performance for low-dose brain PET images.

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 denoising performance of different deep learning (DL) strategies on low-dose brain 18F-FDG PET images. Methods:This retrospective methodological study was conducted on brain PET/CT images of 50 patients (35 males, 15 females, age 20-87 years) who received 3.7MBq/kg 18F-FDG at the Fifth Affiliated Hospital of Sun Yat-sen University between May 2023 and January 2024. Full-dose PET data were acquired with 2min scan. CT scans were acquired before PET scanning. Low-dose PET sinograms were generated by down-sampling the full-dose list mode data to 1/2, 1/4, and 1/20 of full-dose count level. Both full-dose and low-dose sinograms were reconstructed with random, CT-based attenuation and scatter corrections using the three-dimensional (3D) ordered-subsets expectation maximization (OSEM) algorithm (2 iterations, 20 subsets). A total of 4 DL denoising methods were established: (1) 3D conditional generative adversarial networks (GAN) using only low-dose PET as input (GAN-1); (2) 3D attention-based GAN (AttGAN) with low-dose PET input (AttGAN-1); (3) 3D AttGAN with low-dose PET and CT inputs (AttGAN-2); (4) 3D AttGAN with frequency-separation using low-dose PET and CT inputs (AttGAN-FS-2). For AttGAN-FS-2, during the frequency division process, high- and low-frequency components were extracted from the PET reconstructed images via Fourier transform, then inversed Fourier transform, denoised separately, and finally combined to produce the final denoised images. The dataset was separated into training (70%), validation (10%) and testing (20%) sets using simple random sampling without replacement with a fixed random seed. A 5-fold cross-validation scheme was then applied to test all 50 patients. Performance was evaluated against full-dose PET using normalized mean square error (NMSE), structural similarity (SSIM), peak signal-to-noise ratio (PSNR), contrast-to-noise ratio (CNR), SUV mean and SUV max bias of selected brain ROIs. Wilcoxon signed rank test was used to analyze the differences between the denoising methods. Results:AttGAN-FS-2 showed the best performance among all dose levels, with statistical difference as compared by low-dose PET and GAN-1 denoised images for NMSE, SSIM, PSNR, and CNR ( Z values: 2.92-6.15, all P<0.005). NMSE, SSIM quantitative evaluation results (median) of each model at 1/20 dose were: GAN-1: 0.08, 0.87, AttGAN-1: 0.08, 0.88, AttGAN-2: 0.07, 0.89, AttGAN-FS-2: 0.06, 0.91, respectively ( Z values: 3.24-5.77, all P<0.005). Conclusion:The DL-based method combined with multiple strategies AttGAN-FS-2 shows improved denoising performance for low-dose brain PET images.

Background::Artificial intelligence (AI) technology represented by deep learning has made remarkable achievements in digital pathology, enhancing the accuracy and reliability of diagnosis and prognosis evaluation. The spatial distribution of CD3 + and CD8 + T cells within the tumor microenvironment has been demonstrated to have a significant impact on the prognosis of colorectal cancer (CRC). This study aimed to investigate CD3 CT (CD3 + T cells density in the core of the tumor [CT]) prognostic ability in patients with CRC by using AI technology. Methods::The study involved the enrollment of 492 patients from two distinct medical centers, with 358 patients assigned to the training cohort and an additional 134 patients allocated to the validation cohort. To facilitate tissue segmentation and T-cells quantification in whole-slide images (WSIs), a fully automated workflow based on deep learning was devised. Upon the completion of tissue segmentation and subsequent cell segmentation, a comprehensive analysis was conducted.Results::The evaluation of various positive T cell densities revealed comparable discriminatory ability between CD3 CT and CD3-CD8 (the combination of CD3 + and CD8 + T cells density within the CT and invasive margin) in predicting mortality (C-index in training cohort: 0.65 vs. 0.64; validation cohort: 0.69 vs. 0.69). The CD3 CT was confirmed as an independent prognostic factor, with high CD3 CT density associated with increased overall survival (OS) in the training cohort (hazard ratio [HR] = 0.22, 95% confidence interval [CI]: 0.12–0.38, P <0.001) and validation cohort (HR = 0.21, 95% CI: 0.05–0.92, P = 0.037). Conclusions::We quantify the spatial distribution of CD3 + and CD8 + T cells within tissue regions in WSIs using AI technology. The CD3 CT confirmed as a stage-independent predictor for OS in CRC patients. Moreover, CD3 CT shows promise in simplifying the CD3-CD8 system and facilitating its practical application in clinical settings.

Objective: To explore the association between greenness surrounding school and aggression among primary school students, and to explore the potential mediating roles of social support, loneliness, particulate matter (PM2.5) and Nitrogen Dioxide (NO2) in this association, in order to provide a scientific reference for preventing and ameliorating aggressive behaviors of primary school students. Methods: The data was used from a survey of children and adolescents conducted in 2015. The Chinese version of the Buss-Warren Aggression Questionnaire was used to assess total and subtypes of aggression, and the mean values of normalized difference vegetation index (NDVI) of 100 m, 500 m, 1 000 m circular buffers surrounding school were used to indicate the participants greenness exposure. PM2.5 and NO2 datas were obtained from the China High Air Pollutants Dataset.Generalized Linear Mixed Models were used to assess the associations of greenness surrounding school with total and subtypes of aggression. Results: Per IQR increment of NDVI-500 m [OR(95%CI)=1.09(1.03-1.15)] and NDVI-1 000 m[OR(95%CI)=1.07(1.02-1.13)] were positively correlated with physical aggression among primary school children, and per IQR increment of NDVI-100 m [OR(95%CI)=0.94(0.90-0.99)], NDVI-500 m [OR(95%CI)=0.93(0.89-0.98)] and NDVI-1 000 m [OR(95%CI)=0.95(0.91-1.00)] were negatively associated with verbal aggression (P<0.05). Mediation analyses revealed that social support partially mediated the association between the NDVI-500 m and physical aggression (mediation ratio:18.0%) and verbal aggression (mediation ratio:-8.3%) among primary school students, and loneliness partially mediated the association between the NDVI-500 m and physical aggression and verbal aggression among elementary school students effects, with proportion mediated ratios of -10.0% and 21.0%, respectively (P<0.05). Conclusions Exposure to school surrounding greenness is likely to associated with physical aggression and verbal aggression in primary school students, and social support and loneliness may partially mediate these associations.

Objective To investigate the clinical manifestations of familial adenomatous polyposis(FAP).Methods The clinical data of 100 patients with FAP diagnosed in the First Affiliated Hospital of Zhengzhou University from 2011 to 2021 were analyzed retrospectively.Results The main clinical manifestations were bloody stool(44.0%),abdominal pain(40.0%),changes of stool characteristics(25.0%),abdominal distension(18.0%)and diarrhea(17.0%);The most common types of adenomas were villous tubular adenoma(44.9%)and tubular adenoma(32.1%);Extraintestinal manifestations:3 cases of desmoid fibroma,2 cases of osteoma and 1 case of lipoma.47 cases of gastric polyps and the mostly pathological type was fundic gland polyp(34.0%).18 cases of duodenal polyps including 6 cases of adenomatous polyps and most were located at descending duodenum(61.1%).Adenoma canceration occurred in 21 patients.The average age was 38.7 and the canceration mainly occurs in the rectum.The gender,family history,age of onset,number of adenomas,diameter and pathological type of adenomas were statistically compared between patients with and without canceration.It was found that there were significant differences in gender,average age of onset,number of adenoma polyps,diameter of adenoma and pathological type between the two groups(P<0.05).Conclusions FAP is mainly characterized by bloody stool and abdominal pain,with high canceration rate.The risk factors include the age of onset,the number and size of adenomas and the pathological type.The main pathological types are villous tubular adenoma and tubular adenoma.At present,preventive colorectal resection is still the most effective way to treat FAP.No matter what type of methods patients choose,they should be followed up strictly under endoscopy in order to find the lesions in time and treat if necessary.

Objective To investigate the clinical application value of MRI Radiomics score(Radscore)combined with clinicopatho-logical features in predicting postoperative progression-free survival(PFS)of patients with endometrial cancer(EC).Methods A total of 127 patients with EC were selected.The radiomic features of the lesions were extracted from T2 WI,diffusion weighted imaging(DWI)and apparent diffusion coefficient(ADC)images.The features were screened by random forest model and Radscore was calcu-lated.Simultaneously,clinical and pathological characteristics of patients were collected and incorporated,and multivariate Cox regression analysis was used to screen the risk factors related to PFS.The MRI Radscore and clinicopathological features were mapped to the nomogram,and the performance of nomogram was evaluated by receiver operating characteristic(ROC)curve and calibration curve.Results Multivariate Cox regression analysis showed that progesterone receptor(PR),human epididymis protein 4(HE4)and MRI Radscore were independent risk factors for predicting PFS in patients with EC(P＜0.05).The area under the curve(AUC)of the predicted PFS at 1,3 and 5 years after surgery were 0.91,0.804 and 0.776,respectively.Calibration curves showed that nomogram had a good fit in predicting PFS in patients with EC 1,3 and 5 years after surgery.Conclusion The nomogram con-structed based on multi-sequence MRI Radscore and clinicopathological features has favorable accuracy and stability in predicting postoperative PFS in individuals diagnosed with EC.

Objective:To predict the mechanism of Danggui Buxue Decoction for anti-myocardial ischemia-reperfusion injury and "treating different diseases with the same method" in ischemic stroke based on network pharmacology and molecular docking.Methods:The active components and targets of Danggui Buxue Decoction were screened by retrieving the database of TCMSP and literature; the corresponding targets of myocardial ischemia-reperfusion injury and ischemic stroke were found by OMIM and GeneCards database; the intersection targets of Danggui Buxue Decoction and disease were obtained by using Venny diagram, and the common target network and protein-protein interaction network were constructed by Cytoscape 3.7.1 software and STRING database. The GO and KEGG pathways were enriched by David Database, and the Bio GPS database was used to obtain the tissue distribution information of the key targets. The molecular docking technology was used to verify the results.Results:There were 21 active components in Danggui Buxue Decoction, 181 effective targets and 93 cross targets with diseases. The key components were quercetin, Kaempferol, β-sitosterol, formononetin and isorhamnetin. The key targets were AKT1, TNF, IL6, IL-1β and VEGFA. The enrichment results showed that the main action pathways were fluid shear force and arteriosclerosis, lipid and arteriosclerosis, AGE-RAGE signal pathway in diabetic complications, and the core targets were mainly located in the medullary cells, dendritic cell, smooth muscle, prostate, thyroid and other tissues. The results of molecular docking showed that quercetin had the best binding effect to IL-1β, while isorhamnetin had the best binding effect to IL-1β.Conclusion:Danggui Buxue Decoction is against myocardial ischemia-reperfusion injury and ischemic stroke through hemodynamics, lipid metabolism, inflammatory reaction, oxidative stress, immune reaction and cell apoptosis, plays the role of "treating different diseases with the same method".