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.

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.

Amid the wave of medical digitalalization,pathology departments urgently require digital and intelligent transformation to address challenges posed by surging clinical samples and resource shortages.Based on an interpreta-tion of the White Paper on the Development of Digital Intelligent Pathology Departments,this article proposes a"Four-Comprehensive"framework(encompassing all modules,full slide coverage,full workflows,and full ecosystem)as the core of intelligent pathology department development.It emphasizes a tiered implementation strategy,analyzes critical challenges and solutions,and provides tailored recommendations for hospitals at different levels.The paper further ad-vocates for industry-wide collaboration to establish unified data standards,accelerate the advancement of AI-driven pathological models and multimodal applications,and foster equitable resource distribution and precision medicine,thereby advancing China's digital-intelligent pathology ecosystem.

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.

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.

Amid the wave of medical digitalalization,pathology departments urgently require digital and intelligent transformation to address challenges posed by surging clinical samples and resource shortages.Based on an interpreta-tion of the White Paper on the Development of Digital Intelligent Pathology Departments,this article proposes a"Four-Comprehensive"framework(encompassing all modules,full slide coverage,full workflows,and full ecosystem)as the core of intelligent pathology department development.It emphasizes a tiered implementation strategy,analyzes critical challenges and solutions,and provides tailored recommendations for hospitals at different levels.The paper further ad-vocates for industry-wide collaboration to establish unified data standards,accelerate the advancement of AI-driven pathological models and multimodal applications,and foster equitable resource distribution and precision medicine,thereby advancing China's digital-intelligent pathology ecosystem.

Objective:To investigate the distribution characteristics and analyze the clinical significance of dermatophagoides pteronyssinus allergen components in children with allergic rhinitis and asthma in Shenzhen.Methods:This study was a cross-sectional study. The clinical data of children with allergic rhinitis and asthma induced by dust mites admitted to the allergy clinic of Shenzhen Children′s Hospital from 2021 to 2024 were collected and the serum sIgE levels of dermatophagoides pteronyssinus, dermatophagoides farinae (Der p, Der f) and dermatophagoides pteronyssinus components (Der p 1, Der p 2, Der p 10, Der p 23) were detected by magnetic bead chemiluminescence method. The correlation between dermatophagoides pteronyssinus allergen components and clinical data of children was analyzed. According to the diagnosis, the children were divided into allergic rhinitis (AR) group and AR with asthma (ARAS) group. According to the age, the children were divided into preschool age (5 years ≤age<7 years), school age (7 years ≤age<10 years) and adolescence (10 years ≤age≤15 years). The expression differences of dermatophagoides pteronyssinus components among AR group and ARAS group and different age groups were compared.Results:A total of 314 children with allergic rhinitis and asthma caused by dust mites were included in the study, of whom 112 were male and 202 were female. There were 188 cases of AR and 126 cases of ARAS, aged 5-15 years, with a median age of 7.54 years and an average age of (8.02±2.24) years. BMI was 13.89-31.76 kg/m 2,the median BMI was 15.87 kg/m2 and average BMI was (16.55±3.05) kg/m2. There was not statistically significant difference in gender, age, BMI, blood eosinophils, blood basophils, FeNO, FVC and FEV1 between the AR group and the ARAS group ( P>0.05). There was significant difference in FEV1/FVC and small airway function indexes MMEF, MEF75%, MEF50% and MEF25% between the AR group and the ARAS group ( P<0.05). In the 314 children, the dermatophagoides pteronyssinus allergen components sensitization rates were in the order of Der p 1 (97.1%), Der p 2 (89.8%), Der p 23 (55.1%), Der p 10 (8.6%), and the difference in the positive rate was statistically significant (χ 2=658.31, P<0.001). There was not significant difference in Der p 1, Der p 2 and Der p 10 among children of different ages ( P>0.05). There was significant difference in Der p 23 among children of different ages (χ 2=7.29, P=0.03). A correlation analysis showed that Der p, Der f, Der p 1 and Der p 2 had a high positive correlation ( P<0.001). Eosinophils are positively correlated with Der p, Der f, Der p 1, Der p 2, Der p 10 and Der p 23 ( P<0.001). FeNO is positively correlated with Der p, Der f, and Der p 23 ( P<0.05). Small airway function indicators MMEF, MEF50% and MEF25% are negatively correlated with Der p, Der f and Der p 1 ( P<0.05). The sIgE levels of Der p, Der f, Der p 1, Der p 2 and Der p 10 in the AR group were significantly lower than those in the ARAS group ( P<0.05). In the ARAS group, 120 cases (95.24%) showed positive results for at least 2 dermatophagoides pteronyssinus components, while 71 cases (56.35%) showed positive results for at least 3 dermatophagoides pteronyssinus components. In the AR group, 171 cases (90.96%) showed positive results for at least 2 dermatophagoides pteronyssinus components, while 94 cases (50.00%) showed positive results for at least 3 dermatophagoides pteronyssinus components. Conclusion:Der p 1, Der p 2 and Der p 23 may be the main dermatophagoides pteronyssinus allergen components that induce allergic rhinitis and asthma in Shenzhen City. The elevation of sIgE levels in the dermatophagoides pteronyssinus components can aggravate the severity of lower airway eosinophilic inflammation and airway obstruction. Attention should be paid to the detection of dermatophagoides pteronyssinus components in children with poor response to dust mite-allergen specific immunotherapy.

Objective To investigate the injury mechanisms of three-year-old child occupants by reconstructing a real traffic accident.Methods A traffic accident case from the CIREN database was reconstructed using a vehicle finite element model and a three-year-old child occupant injury bionic model(TUST IBMs 3YO-O).The Δv,mass of the vehicle,and deformation energy were comprehensively analyzed to calculate the collision velocity of the vehicle.This accident was simulated to present injuries to a child occupant,and the injury mechanisms were analyzed in depth.Results The TUST IBMs 3YO-O fully reconstructed the injuries of the child occupant in this case.The kinematic and biomechanical responses of the children's heads differed.The biomechanical response of the internal tissues and organs in the chest cavity showed no injury,however,the result ant chest acceleration at 3 ms reached 54 g,which exceeded the threshold.Conclusions In the future,it will be necessary to adopt biomechanical parameters for occupant safety evaluations.The application of human biomechanical models with high biofidelity to reconstruct occupant injuries in traffic accidents can not only be used to observe the kinematic responses of the occupant in the accident and analyze the injury mechanisms in depth,but also to provide references for virtual testing,as well as for the research and development of child occupant protection devices and the formulation of safety regulations.

The study presents a method for the personalized biomechanical modeling of the human head and validates the generated model.Based on the TUST 50th percentile head biomechanical model,the method utilizes head CT data of the target model,and employs three-dimensional point cloud registration and free-form deformation techniques to rapidly develop a personalized head finite element model with detailed brain tissue structures.By reconstructing classic cadaver tests,it is found that the personalized head biomechanical model created by the proposed method shows a good consistency with the results of cadaver tests in kinematic and biomechanical responses.Furthermore,no significant differences are observed when compared with the head biomechanical model developed using reverse engineering method,thus verifying the effectiveness of the developed model.Consequently,the proposed method can be used to quickly construct personalized head biomechanical models with detailed anatomical structures,providing a fundamental computational analysis tool for researches in injury biomechanics,clinical medicine,and forensic identification.

Objective To provide basic data for developing automobile crash safety standards with Chinese human body characteristics,the influence of the muscle active force on the kinematic response of an occupant's head and neck under load impact was investigated.Methods Based on computed tomography(CT)images of the 50th percentile male volunteers with Chinese physical characteristics,a finite element model of the neck containing the cervical vertebrae,muscles,and fat was constructed.The validity of frontal and side impact simulation was verified,and a beam unit was added to the model to simulate the active force of neck muscles.Results The developed neck model consisted of 143 793 units and 165 077 nodes.The simulation experimental data were consistent with the trend of volunteer experimental data,which had a good consistency and verified the effectiveness of the model.A comparison of the simulation results of the activated and passive models showed that the peak motion of the activated model was lower than that of the passive model.Under the side impact,the horizontal displacement of the head of the activated model in the y-direction on the coronal plane did not fully match the experimental channel of the volunteer.Conclusions The muscle active force can maintain the posture and stability of the body.The activation curves,as well as the muscle active force produced by different individuals,vary owing to the different physiological cross-sectional areas of the muscles and other factors.The finite element model of the male neck developed in this study is based on the most recent statistical data of male physiques in China.It has a detailed anatomical structure and high biological fidelity.The model can be used to study the neck injury mechanisms of medium-sized Chinese male physiques.