Objective:To fine-tune the mT5 (massively multilingual pre-trained text-to-text transformer) large language model, automatically generate report conclusions for teaching purposes from chest CT image descriptions, and assess the quality of automatically generated conclusions.Methods:The training set included 3 000 high-quality physical examination chest CT reports from one hospital, and the external validation set consisted of 600 physical examination chest CT reports from two other hospitals. Experienced radiology teaching physicians assessed the consistency between the generated conclusions and the original physician-written conclusions in the external validation set using a 5-point Likert scale across five linguistic indicators (correctness of examination information, correctness of lesion detection, standardization of terminology, applicability of the conclusions, and simplicity of conclusions). Using the original report conclusions as the reference, the accuracy of the conclusions generated based on the external validation set in describing four major thoracic conditions (pulmonary nodules, pneumonia, emphysema, pleural effusion) was evaluated. Perform chi square test using SPSS 25.0.Results:In the external validation set, the mean consistency score between the generated conclusions and the original conclusions given by the radiology teaching physicians was >4 points, indicating agreement with the original conclusions. In the generated conclusions, the description of the four major thoracic conditions demonstrated 0.95-1.00 (95% CI=0.91-1.00) accuracy, 0.76-1.00 (95% CI=0.59-1.00) sensitivity, and 0.97-1.00 (95% CI=0.91-1.00) specificity. Conclusions:The chest CT report conclusion generation system based on the mT5 large language model demonstrated high accuracy and is expected to provide immediate and efficient automated guidance for standardized residency training.

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.

VR technology can generate virtual,immersive,and interactive environments,allowing users to immerse themselves in these environments and interact with objects within them.AR technology,on the other hand,can accurately overlay virtual information onto real-world scenes,achieving a seamless integration of the virtual and the real.These two emerging technologies each possess unique advantages and exhibit broad development prospects.They have already begun to be applied in various aspects of medical education,such as basic theoretical teaching and skills training,with promising results.They can compensate for the shortcomings of traditional medical education,enhance students'learning enthusiasm and safety,and improve teaching effectiveness.However,limitations remain,such as the need for improved hardware infrastructure and a scarcity of teaching resources.Based on this,this paper systematically introduces the concepts of AR and VR technologies,reviews their application prospects,current status,advantages,and limitations in medical education,aiming to provide evidence-based support and feasible approaches for medical schools to develop digital teaching plans,promote educational reform,and drive research innovation.

Objective:To fine-tune the mT5 (massively multilingual pre-trained text-to-text transformer) large language model, automatically generate report conclusions for teaching purposes from chest CT image descriptions, and assess the quality of automatically generated conclusions.Methods:The training set included 3 000 high-quality physical examination chest CT reports from one hospital, and the external validation set consisted of 600 physical examination chest CT reports from two other hospitals. Experienced radiology teaching physicians assessed the consistency between the generated conclusions and the original physician-written conclusions in the external validation set using a 5-point Likert scale across five linguistic indicators (correctness of examination information, correctness of lesion detection, standardization of terminology, applicability of the conclusions, and simplicity of conclusions). Using the original report conclusions as the reference, the accuracy of the conclusions generated based on the external validation set in describing four major thoracic conditions (pulmonary nodules, pneumonia, emphysema, pleural effusion) was evaluated. Perform chi square test using SPSS 25.0.Results:In the external validation set, the mean consistency score between the generated conclusions and the original conclusions given by the radiology teaching physicians was >4 points, indicating agreement with the original conclusions. In the generated conclusions, the description of the four major thoracic conditions demonstrated 0.95-1.00 (95% CI=0.91-1.00) accuracy, 0.76-1.00 (95% CI=0.59-1.00) sensitivity, and 0.97-1.00 (95% CI=0.91-1.00) specificity. Conclusions:The chest CT report conclusion generation system based on the mT5 large language model demonstrated high accuracy and is expected to provide immediate and efficient automated guidance for standardized residency training.

Objective To observe the value of Swin2SR network based on Transformer architecture for reconstructing chest super-resolution CT images.Methods Chest CT data of 218 patients were retrospectively collected.Swin2SR model based on Transformer architecture was adopted to enhance standard 512 matrix(512 × 512)CT images(standard-512 group)into 1 024(SR-1 024 group)and 2 048(SR-2 048 group)matrix SR CT images,respectively.Subjective and objective evaluation of image quality were performed,and the results were compared among groups.Results The subjective scores of overall imaging quality and lesion clarity in SR-1 024 and SR-2 048 groups were both higher than those in standard-512 group(all P＜0.05),while no significant difference was found between the former two(P＞0.05).Meanwhile,no significant difference of objective indexes of imaging quality was observed among 3 groups(all P＞0.05).Conclusion Swin2SR model could reconstruct chest SR CT images without increasing noise and improve imaging quality.

Objective To observe the value of Swin2SR network based on Transformer architecture for reconstructing chest super-resolution CT images.Methods Chest CT data of 218 patients were retrospectively collected.Swin2SR model based on Transformer architecture was adopted to enhance standard 512 matrix(512 × 512)CT images(standard-512 group)into 1 024(SR-1 024 group)and 2 048(SR-2 048 group)matrix SR CT images,respectively.Subjective and objective evaluation of image quality were performed,and the results were compared among groups.Results The subjective scores of overall imaging quality and lesion clarity in SR-1 024 and SR-2 048 groups were both higher than those in standard-512 group(all P＜0.05),while no significant difference was found between the former two(P＞0.05).Meanwhile,no significant difference of objective indexes of imaging quality was observed among 3 groups(all P＞0.05).Conclusion Swin2SR model could reconstruct chest SR CT images without increasing noise and improve imaging quality.

VR technology can generate virtual,immersive,and interactive environments,allowing users to immerse themselves in these environments and interact with objects within them.AR technology,on the other hand,can accurately overlay virtual information onto real-world scenes,achieving a seamless integration of the virtual and the real.These two emerging technologies each possess unique advantages and exhibit broad development prospects.They have already begun to be applied in various aspects of medical education,such as basic theoretical teaching and skills training,with promising results.They can compensate for the shortcomings of traditional medical education,enhance students'learning enthusiasm and safety,and improve teaching effectiveness.However,limitations remain,such as the need for improved hardware infrastructure and a scarcity of teaching resources.Based on this,this paper systematically introduces the concepts of AR and VR technologies,reviews their application prospects,current status,advantages,and limitations in medical education,aiming to provide evidence-based support and feasible approaches for medical schools to develop digital teaching plans,promote educational reform,and drive research innovation.

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 evaluate the relationship between body mass index(BMI)and chronic metabolic diseases.Methods·The elderly(≥60 years old)who were underwent physical examination in the Physical Examination Center of Renji Hospital,Shanghai Jiao Tong University School of Medicine from 2014 to 2021 were studied.Their results of biochemical indicators were collected.Their height,body weight,and blood pressure were measured by trained nurses.The history of chronic metabolic diseases was collected by self-reported questionnaire.Systolic blood pressure≥140 mmHg(1 mmHg=0.133 kPa),diastolic blood pressure≥90 mmHg,or self-reported hypertension history was defined as hypertension.Fasting blood glucose≥7.0 mmol/L or self-reported history of diabetes was defined as diabetes.Total cholesterol≥6.2 mmol/L,triglyceride≥2.3 mmol/L,or self-reported history of dyslipidemia was defined as dyslipidemia.The relationship between BMI and hypertension,diabetes,and dyslipidemia was evaluated by using receiver operator characteristic(ROC)curve analysis and binary logistic regression.Results·Data of 59 083 subjects were collected[30 807 men and 28 276 women,average age:(67.9±6.3)years old].The prevalence of hypertension,diabetes and dyslipidemia was 76.5%(45 219/59 083),24.1%(14 225/59 083)and 50.0%(29 544/59 083),respectively.Compared to the elderly people aged 60?74 years,those aged 75 years and above had a higher proportion of hypertension and diabetes,and a lower proportion of dyslipidemia and no metabolic abnormalities.With ROC analysis,the BMI cut-off values for hypertension,diabetes,and dyslipidemia were 24.3,23.9,and 23.9 kg/m2.The BMI cut-off values for hypertension and diabetes in elderly men were similar to those in elderly women(for hypertension:24.3 kg/m2 in elderly men vs 24.2 kg/m2 in elderly women;for diabetes:24.0 kg/m2 in elderly men vs 23.7 kg/m2 in elderly women);however,BMI cut-off value for dyslipidemia was obviously higher in elderly men than that in elderly women(24.0 kg/m2 in elderly men vs 22.5 kg/m2 in elderly women).The BMI cut-off value for chronic metabolic diseases was higher in the elderly people aged 60?74 years than that in the elderly people aged 75 years and above(24.2?24.7 kg/m2 vs 22.9?23.8 kg/m2).Conclusion·Elderly people aged 60?74 years should maintain the BMI below 24.0 kg/m2,while those aged 75 years and above should aim for the BMI below 23.0 kg/m2,so as to reduce the risk of chronic metabolic diseases.

Primary hypertrophic osteoarthropathy(PHO)is a rare disease also known as pachydermo-periostosis.We reported a painless case whose diagnosis was confirmed by genetic test.A 24-year-old male presented a series of symptoms that first began at 14.He suffered from progressive clubbed-fingers accompa-nied by swelling of the wrist and ankle joints.Facial skin concentric thickening and alar nose broadening ap-peared simultaneously and increased progressively.He was also prone to acne and hyperhidrosis.X-rays showed thickening of the metacarpal and phalangeal bones,as well as symmetrical periosteal ossification of both the tibia and fibula.Clinical diagnosis of PHO is difficult because of the variable features.With acromeg-aly excluded,the diagnosis was confirmed by a genetic test.Whole exome sequencing revealed a heterozygous SLCO2A1 c.611C＞T(p.Ser204Lue)and SLCO2A1 c.1602C＞A(p.Asn534Lys)mutation from each par-ent.It suggests that primary hypertrophic osteoarthropathy should be considered for young limb hypertrophic patients especially when periosteal thickening signs were showed in X-ray.A confirmatory diagnosis can be made through the genetic test.