Objective:To develop and evaluate a medical visual question answering (VQA) model for neonatal lung ultrasound (LUS) images to enhance intelligent auxiliary diagnosis of neonatal pulmonary diseases.Methods:Using data from neonates admitted to Beijing Obstetrics and Gynecology Hospital, Capital Medical University (January 2023 to December 2024), an image-question-answer dataset comprising 251 LUS images was constructed [43 pneumonia (17.1%), 42 neonatal respiratory distress syndrome (16.7%), 83 transient tachypnea (33.1%), and 83 normal (33.1%) images] with a four-tier medical question-answer framework. Building upon the Qwen2.5-VL-7B base model and integrating LoRA fine-tuning with chain-of-thought prompting, we developed the NLUS-VQA model to enhance visual-language semantic alignment and enable stepwise clinical reasoning, achieving efficient small-sample adaptation. Model performance was comprehensively assessed through natural language generation metrics (BLEU-4, ROUGE-1/2/L), qualitative evaluation of characteristic recognition, and clinical consistency analysis.Results:(1) Quantitative evaluation demonstrated that NLUS-VQA achieved scores of 22.38 (BLEU-4), 48.26 (ROUGE-1), 22.40 (ROUGE-2), and 37.20 (ROUGE-L), representing significant improvements over baseline models. (2) Qualitatively, the model exhibited strong performance in identifying lung consolidation, coalescent B-lines, and snowflake signs, with its chain-of-thought strategy enhancing clinical interpretability and answer accuracy. (3) Clinically, NLUS-VQA achieved a Cohen's Kappa coefficient of 0.78 and diagnostic accuracy of 80.8% (21/26), indicating substantial agreement with clinical experts.Conclusion:The NLUS-VQA model demonstrates robust interpretability in recognizing key sonographic patterns (e.g. lung consolidation, confluent B-lines, and snowflake signs), providing a scalable framework for small-sample medical image analysis, though diagnostic performance on complex conditions remains limited by dataset scale and minority class representation.

Objective To observe the clinical efficacy of thumb-tack needling for subcutaneous embedding combined with repetitive transcranial magnetic stimulation(rTMS)in the treatment of post-stroke dysphagia.Methods A total of 120 cases of patients with post-stroke dysphagia admitted to the 908th Hospital of the Joint Logistics Support Force of the People's Liberation Army of China from February 2022 to February 2024 were randomly divided into the observation group and the control group according to random number table method,with 60 cases in each group.Both groups of patients were given conventional symptomatic supportive treatment,such as trophic nerve,improving circulation,etc.,and the control group was given high-frequency rTMS treatment,while the observation group was treated with thumb-tack needling for subcutaneous embedding therapy on the basis of the treatment in the control group.The course of treatment covered four weeks continuously.After one month of treatment,the clinical efficacy of the two groups was evaluated,and the changes in scores of traditional Chinese medicine(TCM)syndrome,Functional Oral Intake Scale(FOIS)and Standardized Swallowing Assessment(SSA),as well as levels of serum total protein(TP),serum albumin(ALB),and hemoglobin(Hb)before and after treatment were observed in the patients of the two groups.The changes in swallowing wave amplitude and swallowing duration before and after treatment were compared between the two groups of patients.Results(1)During the study,there were none in the observation group and eight cases in the control group were failed to follow-up.Finally,64 cases in the observation group and 56 cases in the control group were included in the efficacy statistics.(2)The total effective rate was 92.19%(59/64)in the observation group and 76.79%(43/56)in the control group,and the efficacy of the observation group was superior to that of the control group,the difference being statistically significant(P<0.05).(3)After treatment,the TCM syndrome scores of the patients in the two groups were significantly improved(P<0.05),and the improvement in the observation group was significantly superior to that in the control group,with statistically significant differences(P<0.05).(4)After treatment,the FOIS score and SSA score of the patients in the two groups improved significantly(P<0.05),and the improvement in the observation group was significantly superior to that in the control group,the difference being statistically significant(P<0.05).(5)After treatment,the swallowing wave amplitude and swallowing duration of patients in the two groups improved significantly(P<0.05),and the improvement in the observation group was significantly superior to that in the control group,the difference being statistically significant(P<0.05).(6)After treatment,the TP,ALB and Hb levels of patients in the two groups improved significantly(P<0.05),and the improvement in the observation group was significantly superior to that in the control group,with statistically significant differences(P<0.05).Conclusion Thumb-tack needling for subcutaneous embedding combined with rTMS in the treatment of post-stroke dysphagia can significantly shorten the patients'swallowing duration and enlarge the swallowing wave amplitude,thus improving their swallowing function and improving their quality of life.

Objective To explore conventional MRI and synthetic MRI(syMRI)quantitative parameter histogram features contributing to predicting tumor infiltrating lymphocyte(TIL)level of breast cancer.Methods A total of 114 female patients with solitary breast cancer were retrospectively included and divided into the low-TIL group(L group,0-10%,n=70)and high-TIL group(H group,＞10%,n=44).Conventional MRI features of breast cancer were analyzed based on dynamic contrast-enhanced MRI(DCE-MRI),and histogram features of syMRI quantitative parameters(T1,T2 and proton density[PD])images were extracted,including the 10th and 90th percentile(abbreviated to 10th and 90th,respectively),mean,entropy,kurtosis and skewness,which were then compared between groups.Multivariate logistic regression was used to screen the independent predictors for TIL level among histogram features of syMRI quantitative parameters.Results The proportion of Luminal A breast cancer was lower,while the proportion of HER2-positive breast cancer was higher in H group than those in L group(both adjusted P＜0.05).H group had a higher proportion of lesions with regular margin than L group(P＜0.05).H group exhibited higher value of T1-10th,T1-90th,T1-mean,T2-10th,T2-mean and lower value of PD-skewness compared to those in L group(all P＜0.05).T2-mean was identified as an independent predictor of TIL level of breast cancer(P＜0.05),and its area under the curve(AUC)of receiver operating characteristic(ROC)was 0.780.Conclusion Tumor margin characteristics on conventional MRI and histogram feature T2-mean based on syMRI contributed to predicting TIL level of breast cancer.

Objective To explore the value of artificial intelligence-assisted compressed sensing(ACS)-MR cine imaging(Cine)for assessing biventricular function and left ventricular myocardial strain compared with conventional cardiac Cine and united compressed sensing(uCS)-Cine.Methods A total of 30 subjects who underwent conventional Cine,uCS-Cine and ACS-Cine were prospectively enrolled.Based on a 5-point scale,subjective scoring of image quality was performed and compared among 3 sequences.Biventricular function parameters and myocardial strain parameters of the left ventricle,including left ventricular end-diastolic volume(LVEDV),left ventricular end-systolic volume(LVESV),left ventricular ejection fraction(LVEF),left ventricular end-diastolic mass(LVEDM),right ventricular end-diastolic volume(RVEDV),right ventricular end-systolic volume(RVESV),right ventricular ejection fraction(RVEF),left ventricular global radial strain(LVGRS),left ventricular global circumferential strain(LVGCS)and left ventricular global longitudinal strain(LVGLS)were measured and compared among sequences.Results The subjective image quality scores of all 3 sequences were≥3,of ACS-Cine and conventional Cine were not significantly different(adjusted P=0.306)but both slightly higher than that of uCS-Cine(both adjusted P＜0.05).LVEDV,LVEF and LVEDM based on ACS-Cine and uCS-Cine were not significantly different(all adjusted P＞0.05),but all lower than those of conventional Cine(all adjusted P＜0.05).No significant difference of LVESV was found between each two sequences(all adjusted P＞0.05).RVEDV based on ACS-Cine was lower than that based on conventional Cine,RVESV based on ACS-Cine was higher than that based on uCS-Cine,and RVEF based on ACS-Cine was lower than that based on uCS-Cine and conventional Cine(all adjusted P＜0.05),while no significant difference of right ventricular functional parameters was observed between uCS-Cine and conventional Cine(all adjusted P＞0.05).Both ACS-Cine and uCS-Cine had lower LVGRS and LVGCS than conventional Cine(all adjusted P＜0.05),with no significant difference was found between uCS-Cine and ACS-Cine(both adjusted P＞0.05).No significant difference of LVGLS was observed among 3 sequences(P＞0.05).Conclusion ACS-Cine could be used as appropriate supplement to conventional Cine for clinical assessment of biventricular function and left ventricular myocardial strain.

Objective To explore conventional MRI and synthetic MRI(syMRI)quantitative parameter histogram features contributing to predicting tumor infiltrating lymphocyte(TIL)level of breast cancer.Methods A total of 114 female patients with solitary breast cancer were retrospectively included and divided into the low-TIL group(L group,0-10%,n=70)and high-TIL group(H group,＞10%,n=44).Conventional MRI features of breast cancer were analyzed based on dynamic contrast-enhanced MRI(DCE-MRI),and histogram features of syMRI quantitative parameters(T1,T2 and proton density[PD])images were extracted,including the 10th and 90th percentile(abbreviated to 10th and 90th,respectively),mean,entropy,kurtosis and skewness,which were then compared between groups.Multivariate logistic regression was used to screen the independent predictors for TIL level among histogram features of syMRI quantitative parameters.Results The proportion of Luminal A breast cancer was lower,while the proportion of HER2-positive breast cancer was higher in H group than those in L group(both adjusted P＜0.05).H group had a higher proportion of lesions with regular margin than L group(P＜0.05).H group exhibited higher value of T1-10th,T1-90th,T1-mean,T2-10th,T2-mean and lower value of PD-skewness compared to those in L group(all P＜0.05).T2-mean was identified as an independent predictor of TIL level of breast cancer(P＜0.05),and its area under the curve(AUC)of receiver operating characteristic(ROC)was 0.780.Conclusion Tumor margin characteristics on conventional MRI and histogram feature T2-mean based on syMRI contributed to predicting TIL level of breast cancer.

Objective To explore the value of artificial intelligence-assisted compressed sensing(ACS)-MR cine imaging(Cine)for assessing biventricular function and left ventricular myocardial strain compared with conventional cardiac Cine and united compressed sensing(uCS)-Cine.Methods A total of 30 subjects who underwent conventional Cine,uCS-Cine and ACS-Cine were prospectively enrolled.Based on a 5-point scale,subjective scoring of image quality was performed and compared among 3 sequences.Biventricular function parameters and myocardial strain parameters of the left ventricle,including left ventricular end-diastolic volume(LVEDV),left ventricular end-systolic volume(LVESV),left ventricular ejection fraction(LVEF),left ventricular end-diastolic mass(LVEDM),right ventricular end-diastolic volume(RVEDV),right ventricular end-systolic volume(RVESV),right ventricular ejection fraction(RVEF),left ventricular global radial strain(LVGRS),left ventricular global circumferential strain(LVGCS)and left ventricular global longitudinal strain(LVGLS)were measured and compared among sequences.Results The subjective image quality scores of all 3 sequences were≥3,of ACS-Cine and conventional Cine were not significantly different(adjusted P=0.306)but both slightly higher than that of uCS-Cine(both adjusted P＜0.05).LVEDV,LVEF and LVEDM based on ACS-Cine and uCS-Cine were not significantly different(all adjusted P＞0.05),but all lower than those of conventional Cine(all adjusted P＜0.05).No significant difference of LVESV was found between each two sequences(all adjusted P＞0.05).RVEDV based on ACS-Cine was lower than that based on conventional Cine,RVESV based on ACS-Cine was higher than that based on uCS-Cine,and RVEF based on ACS-Cine was lower than that based on uCS-Cine and conventional Cine(all adjusted P＜0.05),while no significant difference of right ventricular functional parameters was observed between uCS-Cine and conventional Cine(all adjusted P＞0.05).Both ACS-Cine and uCS-Cine had lower LVGRS and LVGCS than conventional Cine(all adjusted P＜0.05),with no significant difference was found between uCS-Cine and ACS-Cine(both adjusted P＞0.05).No significant difference of LVGLS was observed among 3 sequences(P＞0.05).Conclusion ACS-Cine could be used as appropriate supplement to conventional Cine for clinical assessment of biventricular function and left ventricular myocardial strain.

Objective:To develop and evaluate a medical visual question answering (VQA) model for neonatal lung ultrasound (LUS) images to enhance intelligent auxiliary diagnosis of neonatal pulmonary diseases.Methods:Using data from neonates admitted to Beijing Obstetrics and Gynecology Hospital, Capital Medical University (January 2023 to December 2024), an image-question-answer dataset comprising 251 LUS images was constructed [43 pneumonia (17.1%), 42 neonatal respiratory distress syndrome (16.7%), 83 transient tachypnea (33.1%), and 83 normal (33.1%) images] with a four-tier medical question-answer framework. Building upon the Qwen2.5-VL-7B base model and integrating LoRA fine-tuning with chain-of-thought prompting, we developed the NLUS-VQA model to enhance visual-language semantic alignment and enable stepwise clinical reasoning, achieving efficient small-sample adaptation. Model performance was comprehensively assessed through natural language generation metrics (BLEU-4, ROUGE-1/2/L), qualitative evaluation of characteristic recognition, and clinical consistency analysis.Results:(1) Quantitative evaluation demonstrated that NLUS-VQA achieved scores of 22.38 (BLEU-4), 48.26 (ROUGE-1), 22.40 (ROUGE-2), and 37.20 (ROUGE-L), representing significant improvements over baseline models. (2) Qualitatively, the model exhibited strong performance in identifying lung consolidation, coalescent B-lines, and snowflake signs, with its chain-of-thought strategy enhancing clinical interpretability and answer accuracy. (3) Clinically, NLUS-VQA achieved a Cohen's Kappa coefficient of 0.78 and diagnostic accuracy of 80.8% (21/26), indicating substantial agreement with clinical experts.Conclusion:The NLUS-VQA model demonstrates robust interpretability in recognizing key sonographic patterns (e.g. lung consolidation, confluent B-lines, and snowflake signs), providing a scalable framework for small-sample medical image analysis, though diagnostic performance on complex conditions remains limited by dataset scale and minority class representation.

Microgravity in space will lead to the decline of astronauts'motor sensation,which will affect the ability of balance maintenance and posture control.Which limits the ability of astronauts to quickly get out of danger in case of extravehicular activities or landing emergencies,and seriously affects the completion of space missions and endangers the safety of astronauts.The current weightlessness countermeasure based on exercise can effectively alleviate the problems of skeletal muscle and cardiovascular system caused by weightlessness to a certain extent,but it cannot effectively prevent the motor sensory dysfunction after space flight.Therefore,new weightlessness countermeasures of special physical conditioning are needed to attenuate the decline of human balance control ability after long-term space flight.By reviewing the previous research results of space flight and simulated weightlessness experiments,this paper systematically analyzes the influencing factors of human body balance control,compares the balance control promotion schemes of special physical conditioning in the field of sports,and puts forward the potential protection strategies for weightlessness environment based on the integration of sports and medicine.The feedback regulation of muscle tension,tactile and proprioceptive play an important role in human posture balance and motion control.The unloading of human gravity load in space weightlessness environment will greatly reduce the input-output feedback regulation activity of the above senses which may lead to balance dysfunction.The post-flight balance control dysfunction may be improved by special physical conditioning,such as loading of z-axial load,enhancement of tactile feedback and proprioceptive response based on unstable mechanical stress stimulation during spaceflight.

Amino-containing compounds,including amino acids,aliphatic amines,aromatic amines,small peptides and catecholamines,are involved in various biological processes and play vital roles in multiple meta-bolic pathways.Previous studies indicated that some amino-containing metabolites are significant diagnostic and prognostic biomarkers of gastric cancer.However,the discovery of precise biomarkers for the preoperative diagnosis of gastric cancer is still in an urgent need.Herein,we established a polarity-regulated derivatization method coupled with liquid chromatography-mass spectrometry(LC-MS)for amino-containing metabolites profiling in the serum samples of patients with gastric cancer and healthy controls,based on our newly designed and synthesized derivatization reagent(S)-3-(1-(diisopropox-yphosphoryl)pyrrolidine-2-carboxamido)-N-hydroxysuccinimidyl ester(3-DP-NHS).Enhanced separa-tion efficiency and detection sensitivity for amino-containing metabolites were achieved after derivatization.This method exhibited good linearity,recovery,intra-and inter-day precision and accu-racy.Only 5 μL serum is needed for untargeted analysis,enabling 202 amino-containing metabolites to be detected.Statistical analysis revealed altered amino acid metabolisms in patients with gastric cancer.Furthermore,ultra high performance liquid chromatography coupled with mass spectrometry(UHPLC-MS/MS)analysis quantification revealed increased serum levels of tryptamine and decreased concen-trations of arginine and tryptophan in patients with gastric cancer.Receiver operating characteristic(ROC)curves indicated that an increased tryptamine/tryptophan ratio could serve as a potential biomarker for gastric cancer diagnosis.This study demostrated the possibility of using serum amino acid biomarkers for gastric cancer diagnosis,providing new avenues for the treatment of gastric cancer.