BACKGROUND:Rapid and accurate segmentation of brain tissue in medical images is of great significance for three-dimensional biomechanical modeling and diagnosis of craniocerebral injuries.Currently,artificial intelligence(AI)-based baseline models exhibit excellent generalization capabilities on large-scale datasets.However,due to the specificity and complexity of craniocerebral tissues,these models have certain limitations in their application to craniocerebral tissue segmentation.Additionally,the scarcity of craniocerebral tissue samples makes it difficult for baseline models to achieve precise segmentation results through fine-tuning.OBJECTIVE:To construct a craniocerebral tissue segmentation model based on texture feature retrieval enhancement to improve segmentation accuracy under a small number of samples.METHODS:Segment Anything in Medical Images(MedSAM)model was selected as the basic framework,and texture features were combined with deep learning to build a brain tissue segmentation model based on texture feature retrieval enhancement(DP-MedSAM).Dice Coefficient and mean intersection over union(MIoU)were selected to evaluate the efficiency of image segmentation results.In comparison with the original MedSAM model,the ablation experiment systematically evaluated the influence of key components on the model performance.The sensitivities of MedSAM,the Segment Anything Model(SAM)for medical image segmentation(SAM-Med2D)and DP-MedSAM in the mandible,left optic nerve,and left parotid gland were compared.RESULTS AND CONCLUSION:(1)By verifying the impact of the number of point prompts on segmentation results on the HaN-Seg dataset,the experimental results indicated that the optimal Dice score was achieved with the addition of three points.(2)DP-MedSAM demonstrated performance improvements compared with MedSAM and SAM-Med2D on two datasets(HaN and Public Domain Database for Computational Anatomy).Especially on the Public Domain Database for Computational Anatomy dataset,in terms of the MIoU metric,DP-MedSAM outperformed MedSAM by 6.59％and SAM-Med2D by 37.35％;in terms of the Dice metric,DP-MedSAM outperformed MedSAM and SAM-Med2D by 4.34％and 25.32％,respectively.(3)The ablation experiment results showed that removing the texture feature extraction module in the DP-MedSAM model,relying solely on original image features,led to a significant decrease in results on the test set.Furthermore,removing the vector cache database and its retrieval enhancement function from the model,which deprived the ability of the model to perform similarity retrieval using an external knowledge base,further reduced model performance.(4)Under conditions of limited data resources,the DP-MedSAM model outperformed the other two models in all evaluation metrics.The DP-MedSAM model performed excellently when processing simple and moderately difficult samples,demonstrating a clear advantage over the other two models and indicating good generalization ability.Processing the fine structures of difficult samples placed higher demands on the model's segmentation capabilities.Although the performance of the DP-MedSAM model declined slightly,it still outperformed the other two models.(5)This study proposes an innovative craniocerebral tissue segmentation model,DP-MedSAM,which improves the baseline model's performance in capturing local details and global structural information in medical images by introducing target region texture feature extraction.Through vector similarity retrieval technology,DP-MedSAM can retrieve the feature vector most similar to the current target region from a pre-constructed vector database,providing more precise guiding information for the segmentation process.

This paper proposes a two-stage method integrating visual foundation models （VFM） and diffusion models. The segment anything model （SAM） as VFM is combined with the SegRefiner diffusion model to construct the SAM-SegRefiner framework for automated segmentation of edema， inflammation， and thrombus regions in histopathological images of hemorrhoidal tissue， providing a reproducible technical tool for the objective quantification of pathological morphology and its application in traditional Chinese medicine （TCM） syndrome research. Trained and validated on multi-center retrospective data， the SAM-SegRefiner model achieved an average pixel accuracy of 95.32% and a mean intersection over union （mIoU） of 66.81% on an independent test set， significantly outperfor-ming comparative models such as U-Net， MixU-Net， and SAM-Med2D， and also demonstrating robust cross-center generalization capability. Furthermore， by correlating the quantitatively segmented results from the model with the patients' TCM syndrome types， the potential associations between pathomorphological features and TCM syndrome differentiation have been explored. The analysis revealed no statistically significant differences in the degree of inflammatory infiltration and thrombus formation among different syndrome types， suggesting a complex relationship between local pathological changes and systemic syndrome manifestations.

BACKGROUND:Rapid and accurate segmentation of brain tissue in medical images is of great significance for three-dimensional biomechanical modeling and diagnosis of craniocerebral injuries.Currently,artificial intelligence(AI)-based baseline models exhibit excellent generalization capabilities on large-scale datasets.However,due to the specificity and complexity of craniocerebral tissues,these models have certain limitations in their application to craniocerebral tissue segmentation.Additionally,the scarcity of craniocerebral tissue samples makes it difficult for baseline models to achieve precise segmentation results through fine-tuning.OBJECTIVE:To construct a craniocerebral tissue segmentation model based on texture feature retrieval enhancement to improve segmentation accuracy under a small number of samples.METHODS:Segment Anything in Medical Images(MedSAM)model was selected as the basic framework,and texture features were combined with deep learning to build a brain tissue segmentation model based on texture feature retrieval enhancement(DP-MedSAM).Dice Coefficient and mean intersection over union(MIoU)were selected to evaluate the efficiency of image segmentation results.In comparison with the original MedSAM model,the ablation experiment systematically evaluated the influence of key components on the model performance.The sensitivities of MedSAM,the Segment Anything Model(SAM)for medical image segmentation(SAM-Med2D)and DP-MedSAM in the mandible,left optic nerve,and left parotid gland were compared.RESULTS AND CONCLUSION:(1)By verifying the impact of the number of point prompts on segmentation results on the HaN-Seg dataset,the experimental results indicated that the optimal Dice score was achieved with the addition of three points.(2)DP-MedSAM demonstrated performance improvements compared with MedSAM and SAM-Med2D on two datasets(HaN and Public Domain Database for Computational Anatomy).Especially on the Public Domain Database for Computational Anatomy dataset,in terms of the MIoU metric,DP-MedSAM outperformed MedSAM by 6.59％and SAM-Med2D by 37.35％;in terms of the Dice metric,DP-MedSAM outperformed MedSAM and SAM-Med2D by 4.34％and 25.32％,respectively.(3)The ablation experiment results showed that removing the texture feature extraction module in the DP-MedSAM model,relying solely on original image features,led to a significant decrease in results on the test set.Furthermore,removing the vector cache database and its retrieval enhancement function from the model,which deprived the ability of the model to perform similarity retrieval using an external knowledge base,further reduced model performance.(4)Under conditions of limited data resources,the DP-MedSAM model outperformed the other two models in all evaluation metrics.The DP-MedSAM model performed excellently when processing simple and moderately difficult samples,demonstrating a clear advantage over the other two models and indicating good generalization ability.Processing the fine structures of difficult samples placed higher demands on the model's segmentation capabilities.Although the performance of the DP-MedSAM model declined slightly,it still outperformed the other two models.(5)This study proposes an innovative craniocerebral tissue segmentation model,DP-MedSAM,which improves the baseline model's performance in capturing local details and global structural information in medical images by introducing target region texture feature extraction.Through vector similarity retrieval technology,DP-MedSAM can retrieve the feature vector most similar to the current target region from a pre-constructed vector database,providing more precise guiding information for the segmentation process.

Objective To explore the construction of heart preservation model of empty beating donor based on extracorporeal membrane oxygenation (ECMO). Methods From January 2022 to August 2023, 20 Guangxi Bama miniature pigs weighing 25-30 kg were selected, half male and half female. Under general anesthesia and heparinization, a midline thoracotomy was performed. The pericardium was cut after freeing the anterior and posterior vena cavae, and a perfusion needle was inserted near the brachiocephalic artery in the ascending aorta, connected to a blood collection bag to collect 500-600 mL of blood. The anterior and posterior vena cavae were ligated, the aorta was blocked and perfused with HTK solution to stop the heart beating. The superior and inferior vena cavae were cut off, the right pulmonary vein was decompressed, the aorta and left and right pulmonary arteries and veins were cut off, and the whole heart was removed. An ECMO device was used to continuously perfuse a cardioprotective solution mainly composed of oxygenated warm blood, maintaining the isolated pig heart beating for 8 hours, monitoring (once/hour) ECMO perfusion parameters, blood gas indicators, perfusate electrolytes, inflammatory factors, myocardial enzymes, myoglobin, and troponin levels. Myocardial tissue was taken for hematoxylin-eosin (HE) staining to observe myocardial cell damage and evaluate the quality of heart preservation. Results Among the 20 isolated beating pig hearts, 17 successfully resumed beating, 3 experienced ventricular fibrillation, resuscitated after intracardiac electrical defibrillation, and all 20 pig hearts successfully beat for 8 hours. There was no statistical difference in ECMO perfusion parameters, blood gas indicators, perfusate electrolytes, and inflammatory factors at each time point (P>0.05). There were statistical increases in myocardial enzymes, myoglobin, and troponin levels (P<0.05). HE staining results suggested that there was no severe myocardial damage. Conclusion ECMO technology can be used for pig heart preservation with good results, and this study provides experimental evidence for improving heart preservation research in clinical heart transplantation.

Purpose: Our previous study showed that circulating tumor cell (CTC) count combined with gene mutation detection might help differentiate benign and malignant pulmonary nodules (PNs). Herein, we aimed to expand the study cohort and conduct further sequencing analysis. Materials and Methods: Patients with PNs were included, and CTCs were identified before operation. Low-coverage whole-genome sequencing (LC-WGS) and lung cancer-related targeted gene sequencing were performed on CTCs. The diagnostic efficacy was evaluated by receiver operating characteristic (ROC) curve. The differences in CTC counts among subgroups classified by demographic–clinical characteristics were analyzed. LC-WGS–based copy number variation (CNV) analysis and targeted gene mutation analysis were conducted. Results: A total of 172 patients were included. CTC count of 2.5 was identified by the ROC curves as the optimal diagnostic cutoff. The sensitivity and specificity of CTC count for differentiating benign and malignant PNs were 54.2% and 78.6%, respectively. The diagnostic sensitivity and specificity of combined CTC count, radiological nodule type, and any malignant imaging features were 84.7% and 71.4%, respectively. The CTC counts were significantly greater in patients with aggressive tumors, later stage, and spread through air spaces. CTCs from malignant cases had more CNVs than those from benign cases. Conclusion CTC count can be used in identifying malignant PNs. The diagnostic efficacy can be improved if combined with computed tomography imaging characteristics. Further CNV analysis might help differential diagnosis. Greater CTC count might suggest more aggressive tumors. CTC detection can provide important information and guidance for subsequent management of PNs.

Objective:To investigate, for multi-sensitized allergic rhinitis (AR) patients allergic to dust mites combined with other allergens (pollen, mold, animal dander, etc.), whether the single dust mite subcutaneous immunotherapy (SCIT) can improve the specific symptoms caused by other allergens in the patients, and to analyze the relationship between the effectiveness of symptom improvement in these patients and the type, quantity and severity of the allergens.Methods:A multicenter retrospective study was conducted to collect mul-sensitized AR patients from allergy or respiratory departments of 5 hospitals who received house dust mite allergen preparation SCIT for 12 to 36 months and met other inclusion and exclusion criteria from February to July 2024. General clinical data were collected and the perennial or seasonal symptoms before and after treatment were evaluated with visual analogue scale (VAS) to assess whether there was an perennial or allergen-specific symptom improvement (VAS score decrease ≥30%), by which the patients were divided into effective group and ineffective. R software was used to analyze the differences between groups by using Fisher′s exact test and Mann-Whitney U test. Results:A total of 62 patients were enrolled, and the treatment were effective in 39 of them, with an effective rate of 62.9%. For allergen-specific symptoms, the median age of the effective group was higher than that of the ineffective group (12 years old vs. 8 years old, P=0.039), and the effective rate in dust mite specific immunoglobin E (sIgE) grade ≤5 group was higher than that in sIgE grade >5 group (81.6% vs. 45.5%, P=0.008), and the effective rate of mold sIgE grade ≤2 group was higher than that of sIgE grade >2 group (83.3% vs. 28.6%, P=0.045), and there was no statistically significant correlation between the other allergen grades and the effective rate ( P>0.05). For perennial symptoms, the effective rate in the mold grade ≤2 group was higher than that in the sIgE grade >2 group (91.3% vs. 28.6%, P=0.010), and there was no statistically significant correlation between the other allergen grades and the effective rate ( P>0.05). There was no significant correlation between the treatment effectiveness of perennial or allergen-specific symptoms and the number of combined allergens, the grade of skin test, and the difference between the grade of combined allergens and that of dust mites ( P>0.05). Conclusion:Among the patients with multi-sensitized AR allergic to dust mites included in this study, single dust mite SCIT is effective in some of them, and for allergen-specific symptoms, the effective group was elder, and dust mite sIgE grade 6 and mold sIgE grade ≥2 was related to the low effective rate of SCIT. The present results are insufficient for selecting single or multiple AIT in any type of multi-sensitized patients.

To develop vaccines for the prevention and control of porcine circovirus 2d genotype(PCV2d)and pseudorabies virus(PRV),the PCV2d ORF2 gene was amplified by PCR,and cloned into the BamH Ⅰ site of PRV transfer plasmid pG vector harboring the enhanced green fluorescent protein(EGFP)gene.The resulting recombinant transfer plasmid pG-PCV2d-EGFP was transfect-ed into ST cells infected with the three gene deleted PRV variant strain gE-/g-/TK-PRV NY to generate a recombinant virus rPRV-PCV2d-EGFP+,and then the EGFP gene was knocked out to harvest the rPRV-PCV2d using gene-editing technology termed CRISPR/Cas9 system.The recom-binant virus rPRV-PCV2d had similar genetic stability to the parental PRV as indicated by PCR and one-step growth curve test,and the expression of PCV2d capsid(Cap)protein was validated by Western blot.In animal experiment,higher PCV2-specific ELISA antibodies and detectable PCV2-specific neutralizing antibodies could be elicited in mice immunized with the recombinant vi-rus rPRV-PCV2d compared to commercial PCV2 inactivated vaccine.rPRV-PCV2d significantly re-duced the PCV2d loads in tissues such as the heart,liver and spleen of mice following virulent PCV2d challenge.Moreover,rPRV-PCV2d elicits PRV-specific immune responses in mice and can prevent PRV virulent infection in mice,indicating the recombinant virus rPRV-PCV2d has strong immunogenicity.

Based on CT scan data,a bionic model of lumbar spine injuries with high biofidelity is developed and validated through cadaver experiments.Decoupling the constraint system that affects occupants during collisions due to inertial forces and the subsequent pressure exerted by the seat upon returning to position,a simulated fall experiment is designed.The simulated outcomes are trained and predicted using deep learning algorithms,and the accuracy of the trained neural network prediction model is verified.Key parameters are analyzed for correlation using principal component analysis and cross-reverse methods.The results shows that the predicted lumbar spine injury model obtained from training has high reliability(R2>0.9).Comprehensive analysis reveals that after experiencing axial impact,the L4 vertebral body bears the highest impact load and can be used as a representative measure of lumbar spine injury.Among the environmental variables,the axial force on the L4 lumbar spine is mainly affected by torso mass and fall height,both of which have positive correlations.Torso mass,fall height,and posture angle all have positive effects on internal energy.Conversely,torso mass and fall height have negative correlations with stress.These research findings provide a scientific basis for further elucidating lumbar spine injury mechanisms in intelligent cockpit environments,devising corresponding safety protection measures,and evaluating occupant safety in automobiles.

Objective To construct mice hepatocellular carcinoma models with different tumor immune microenvironments(TIME)and explore the differences.Methods H22 and hepa1-6 were used to construct subcutaneous transplantation tumor model of C57 mice as homologous hepatocellular carcinoma cell lines(denoted as H22 group and hepal-6 group,each n=8),and the differences of TIME were evaluated.Immunohistochemistry was used to detect and quantify the infiltration of T cells,CD4+T cells,CD8+T cells,regulatory T cells and B cells in TIME.Flow cytometry was performed to detect the differences of composition of immune cell subpopulations in peripheral blood and tumor parenchyma.Gene expression profile characteristics of tumor tissue were analyzed based on high-throughput transcriptome sequencing technology,and enrichment analyses of immune-related signaling pathways were evaluated combined with gene ontology(GO)and Kyoto encyclopedia of genes and genomes(KEGG).Results H22 group showed cold and hepa1-6 group showed hot TIME characteristics.The number of T cells,CD4+T cells and CD8+T cells in tumor tissue of H22 group were all lower,while the proportion of T cells,CD4+T cells and CD8+T cells in peripheral blood were all higher than those of hepa1-6 group(all P＜0.05).Compared with H22 group,up-regulated genes of tumor tissue in hepa1-6 group expressed significantly enriched in tumor immune activation-related signaling pathways.Conclusion H22 and hepa1-6 hepatocellular carcinoma models showed distinct TIME characteristics of cold and hot tumors,respectively,and the amount of immune cells in tumor tissue of the former were significantly lower than those in the latter.

Objective:To evaluate the value of ultrasound radiomics combined with machine learning for early diagnosis of seronegative Hashimoto’s thyroiditis (SN-HT) .Methods:This retrospective study included 164 patients from Liaoning Provincial People’s Hospital , Lixin County People’s Hospital, Linghai Dalinghe Hospital, Fengcheng Phoenix Hospital, who underwent thyroidectomy for solitary nodules with normal thyroid function between Nov. 2016 and Jan. 2024. Postoperative pathology confirmed Hashimoto’s thyroiditis (HT) in some cases, who were further categorized into antibody-positive and antibody-negative groups based on serum antibody status. Patients without Hashimoto’s thyroiditis served as the control group. A total of 298 ultrasound images were analyzed. Radiomics features were extracted from hypoechoic non-nodular areas within 0.5 cm surrounding the tumor. Two senior pathologists and two senior ultrasound physicians independently assessed lymphocytic infiltration, eosinophilic changes of follicular epithelium, and the proportion of hypoechoic areas in pathology and ultrasound images, respectively. A machine learning model, CCH-NET, was developed using linear regression and t-distributed stochastic neighbor embedding (t-SNE) techniques. The dataset was divided into a training set (80%) and a validation set (20%) to compare the diagnostic accuracy of CCH-NET with that of senior ultrasound physicians. Results:In internal validation, CCH-NET achieved a diagnostic accuracy of 88.89% for both antibody-positive and antibody-negative groups, significantly higher than the 66.67% accuracy of senior ultrasound physicians ( P<0.01). In external validation, CCH-NET achieved 75.00% and 66.67% accuracy for the two groups, compared to 50.00% by senior ultrasound physicians. For the control group, both methods achieved 93.33% accuracy. The AUC of CCH-NET was 0.848, outperforming senior ultrasound physicians (0.681) ,demonstrating superior diagnostic performance. Conclusion:The radiomics-based CCH-NET model, using non-nodular hypoechoic areas as a specific indicator, can accurately identify early SN-HT in euthyroid patients. It significantly outperforms senior ultrasound physicians, improving diagnostic accuracy and reducing missed diagnoses.