BACKGROUND: Over 20 million colonoscopies are performed in China annually. An automatic clinical decision support system (CDSS) with accurate semantic recognition of colonoscopy reports and guideline-based is helpful to relieve the increasing medical burden and standardize the healthcare. In this study, the CDSS was built under a hierarchical-label interpretable classification framework, trained by a state-of-the-art transformer-based model, and validated in a multi-center style. METHODS: We conducted stratified sampling on a previously established dataset containing 302,965 electronic colonoscopy reports with pathology, identified 2041 patients' records representative of overall features, and randomly divided into the training and testing sets (7:3). A total of five main labels and 22 sublabels were applied to annotate each record on a network platform, and the data were trained respectively by three pre-training models on Chinese corpus website, including bidirectional encoder representations from transformers (BERT)-base-Chinese (BC), the BERT-wwm-ext-Chinese (BWEC), and ernie-3.0-base-zh (E3BZ). The performance of trained models was subsequently compared with a randomly initialized model, and the preferred model was selected. Model fine-tuning was applied to further enhance the capacity. The system was validated in five other hospitals with 3177 consecutive colonoscopy cases. RESULTS: The E3BZ pre-trained model exhibited the best performance, with a 90.18% accuracy and a 69.14% Macro-F1 score overall. The model achieved 100% accuracy in identifying cancer cases and 99.16% for normal cases. In external validation, the model exhibited favorable consistency and good performance among five hospitals. CONCLUSIONS The novel CDSS possesses high-level semantic recognition of colonoscopy reports, provides appropriate recommendations, and holds the potential to be a powerful tool for physicians and patients. The hierarchical multi-label strategy and pre-training method should be amendable to manage more medical text in the future.
Humans ; Colonoscopy/methods* ; Deep Learning ; Decision Support Systems, Clinical ; Female ; Male

BACKGROUND: Lung cancer is the leading malignancy in China in terms of both incidence and mortality. With increased health awareness and the widespread use of low-dose computed tomography (CT), early diagnosis rates have been steadily improving. Surgical intervention remains the primary treatment option for early-stage lung cancer, and video-assisted thoracoscopic surgery (VATS) has become a common approach due to its minimal invasiveness and rapid recovery. However, post-discharge recovery remains incomplete, underscoring the importance of postoperative care. Traditional follow-up methods, lack standardization, consume significant medical resources, and increase the burden of the patients. Artificial intelligence (AI)-driven disease management platforms offer a novel solution to optimize postoperative follow-up. This study followed 463 lung cancer surgery patients using an AI-based platform, aiming to identify common postoperative issues, propose solutions, improve quality of life, reduce recurrence-related costs, and promote AI integration in healthcare. METHODS: Using the AI disease management platform, this study integrated educational videos, collaboration between healthcare teams and AI assistants, daily health logs, health assessment forms, and personalized interventions to monitor postoperative recovery. The postoperative rehabilitation status of the patients was assessed by the Leicester Cough Questionnaire (LCQ-MC). Two independent t-test and one-way ANOVA were used to analyze the causes of postoperative cough in lung cancer. RESULTS: Most issues occurred within 7 d post-discharge, significantly declined on 14 d post-discharge. Factors such as gender, smoking history, and surgical approaches were found to influence cough recovery. The incidence of cough on 7 d post-discharge in females was higher than that in males (P<0.01), while the incidence of cough on 14 d post-discharge in elderly patients was lower than that in young patients (P=0.03). The AI-based platform effectively addressed cough, pain, and sleep disturbances through phased interventions. CONCLUSIONS The AI-based platform significantly enhanced postoperative management efficiency and the self-care capabilities of the patients, particularly in phased cough management. Future integration with wearable devices could enable more precise and personalized postoperative care, further advancing the application of AI technology across multidisciplinary healthcare domains.
Humans ; Lung Neoplasms/rehabilitation* ; Male ; Female ; Middle Aged ; Aged ; Patient Discharge ; Artificial Intelligence ; Adult ; Postoperative Care ; Postoperative Period ; Disease Management ; Quality of Life

Objective:To preliminarily explore the correlation between the quantitative parameters of spectral CT before proton radiotherapy and the changes in apparent diffusion coefficient (ADC) of MRI before and after radiotherapy in chordoma patients.Methods:A retrospective analysis was conducted on imaging data from 28 patients with chordoma who underwent proton radiotherapy at Shandong Cancer Hospital and Institute from August 2022 to December 2024. Spectral CT images obtained prior to treatment were used to extract four quantitative parameters of the lesion area: relative iodine concentration (RIC), electron density (ED), effective atomic number (Z eff), and the slope of spectral attenuation curve (λHU). In parallel, MRI-DWI scans before and after treatment were collected to compute the ADC difference (ΔADC). The correlation between spectral CT parameters and ΔADC was assessed using scatter plots, simple linear regression, LOWESS curve fitting, correlation matrix, and bootstrap resampling methods. Results:The ADC of 28 patients was 1 137.05 (921.07, 1 643.91) before treatment, and 1 197.10 (994.75, 1 785.57) after treatment, and the ΔADC was 133.18 (-36.46, 253.04). The RIC was 0.45 (0.12, 0.67), the ED was 38.01±12.72, the Z eff was 8.40±0.64, and the λHU was -2.20±1.05. The scatter plots results indicated a positive distribution trend between RIC and ΔADC, and a negative trend between λHu and ΔADC. Simple linear regression showed that RIC had the highest goodness of fit with ΔADC ( R2=0.75) and the largest regression coefficient ( β=518.34), followed by λHU ( R2=0.64, β=-121.94). The goodness of fit between Z eff and ΔADC was low ( R2=0.25). No correlation was found between ED and ΔADC. LOWESS curve fitting showed a consistent trend with simple linear regression results, without significant deviation. The correlation matrix indicated that RIC was positively correlated with ΔADC ( r=0.88, P<0.001), Z eff was moderately positively correlated with ΔADC ( r=0.51, P=0.006), and λHU was negatively correlated with ΔADC ( r=-0.84, P<0.001). Bootstrap resampling analysis showed that the r values of RIC, Z eff, λHU and ΔADC were 0.87, 0.50, -0.80, respectively. Conclusions:The spectral CT parameter RIC of chordoma patients before proton radiotherapy is positively correlated with ΔADC, while λHU exhibits a negative correlation with ΔADC.

Objective To analyze the risk factors and re-intervention strategies for mid- and long-term residual after arterial switch operation (ASO). Methods The clinical data of children with complex congenital heart disease who underwent ASO surgery in Shanghai Children’s Medical Center from January 2006 to June 2022 were retrospectively collected, and the risk factors for mid- and long-term residual after ASO were analyzed. Results A total of 952 children undergoing ASO were enrolled in this study, including 654 males and 298 females with an average age of (102.9±90.1) d and weight of (4.6±1.6) kg. There were 421 patients with D-transposition of the great arteries with intact ventricular septum (D-TGA/IVS), 357 patients with D-transposition of the great arteries with ventricular septal defect (D-TGA/VSD), and 174 patients with right ventricle double outlet combined with subpulmonary ventricular septal defect (Taussig-Bing malformation). Eighty-nine patients died early after the surgery, the mortality rate was 9.3%. The 746 surviving children were regularly followed up after the surgery (follow-up rate 86.4%), with a median follow-up time of 79.4 (12.0-188.0) months. During the follow-up, 53 children underwent surgical re-intervention due to residual, including 33 males and 20 females, with a median age of 62.5 (17.0-214.0) months. The median surgical weight was 19.0 (8.2-86.0) kg, and the mean time of re-intervention was 28.0-170.0 (77.5±45.4) months after the ASO. Residual problems included common trunk and branch stenosis of the pulmonary artery in 23 patients, right ventricular outflow tract (RVOT) obstruction in 11 patients, left ventricular outflow tract obstruction in 6 patients, aortic arch restenosis in 5 patients, aortic insufficiency in 5 patients, residual shunt of ventricular septal defect in 2 patients, and tricuspid valve insufficiency in 1 patient. The early postoperative mortality rate was 3.8% (2/53), with the causes of death being acute myocardial infarction due to coronary artery injury and acute left heart failure, respectively. The mean follow-up time of the surviving children was (52.4±28.6) months, and no mid- and long-term death occurred. Two patients underwent the third operations due to pulmonary restenosis. The multivariate analysis result showed that combined aortic arch surgery and early postoperative RVOT velocity>3 m/s were independent risk factors for mid- and long-term residual after ASO. Conclusion ASO is an ideal procedure for the treatment of D-TGA/IVS, D-TGA/VSD and Taussig-Bing malformations. Combined aortic arch surgery and early postoperative RVOT velocity>3 m/s are independent risk factors for mid- and long-term residual after ASO.

ObjectiveTo investigate a suspected outbreak of healthcare-associated infection with carbapenem-resistant Klebsiella pneumoniae (CRKP) in a geriatric emergency ward, and to provide references for the prevention and control of multidrug-resistant bacteria in a hospital in Shanghai. MethodsOn-site epidemiological investigation, combined with environmental hygiene monitoring and pulsed field gel electrophoresis (PFGE) molecular typing method, were adopted to investigate a suspected outbreak of CRKP infection in the geriatric emergency ward of a hospital from October to November 2022, aiming at finding out factors caused the outbreak before taking corresponding control measures. ResultsA total of 3 cases of healthcare-associated CRKP infection were identified, of which 2 cases were homologous to a previous case of community-associated CRKP infection. What’s more, the 2 cases lived in the same ward with the latter and with adjacent beds, but the third case was non-homologous to the community-associated infection case. A total of 46 samples were collected from the environmental surfaces and the hands of healthcare workers, of which 7 samples tested positive for CRKP and were identical to the strains from the 2 healthcare-associated infection cases and the 1 community-associated infection case, originating from the bedrails, bedside tables, surface of non-invasive ventilator, bed curtains and panels of monitoring equipment, with a detection rate of 15.22%. But none of the 11 samples from the hands of healthcare workers tested positive for CRKP. The outbreak was effectively controlled after taking specific prevention and control measures such as strengthening personnel management, intensifying environmental cleaning and disinfection and strictly enforcing hand hygiene among healthcare workers. Subsequently, no similar new cases were reported during the 14-day follow-up period. ConclusionIncomplete environmental cleaning and disinfection, as well as inadequate enforcement of hand hygiene among heatheare workers may have contributed to the suspected outbreak of CRKP in the geriatric emergency ward. Early warning and timely investigation of suspected outbreaks of multidrug-resistant bacteria are crucial for preventing and controlling such outbreaks in hospitals.

Objective:To explore the clinical effect of the whole-process management scheme of daytime minimally invasive liver resection surgery based on the enhanced recovery after surgery (ERAS) concept.Methods:This is a retrospective case series study. The data of 55 patients who underwent minimally invasive daytime liver resection surgery under the ERAS concept at the Department of General Surgery,Sir Run Run Shaw Hospital, Zhejiang University School of Medicine from January 2023 to August 2024. There were 22 males and 33 females;aged (48.2±15.1) years (range: 16 to 77 years). All patients were classified as Grade 2 according to the American Society of Anesthesiologists physical status classification. Among them, 7 cases were complicated with liver cirrhosis and 10 cases had fatty liver. A multidisciplinary team was formed, consisting of surgeons, anesthesiologists, rehabilitation physicians, psychologists, pharmacists, acute pain management team, operating room nurses, day surgery ward nurses, and ERAS specialized nurses. After strict evaluation by surgeons and anesthesiologists, patients suitable for daytime liver resection surgery were implemented with the ERAS whole-process management plan for liver resection on the basis of routine nursing care.Results:Among the 55 patients, 50 were discharged smoothly within 48 hours, while 5 were transferred to specialized departments for further treatment due to not meeting the discharge criteria, with a smooth daytime discharge rate of 90.9%. Among the 50 patients, 30 underwent laparoscopic surgery and 20 underwent robotic-assisted surgery. The surgery time was (91.6±28.2)minutes(range:45 to 165 minutes), with the intraoperative blood loss of only (30.5±25.5)ml(range:5 to 100 ml). Pathological examination results showed that among the 50 patients, 13 cases had hepatocellular carcinoma, 21 cases had hepatic hemangioma, 4 cases had hepatic cyst, 8 cases had focal nodular hyperplasia, 1 case had low-grade dysplastic nodule, 1 case had hepatolithiasis, 1 case had lymphoma, and 1 case had vascular, fibrous and lymphoid tissue proliferation. There were 44.0% patients who were able to get out of bed on the day of surgery. The hospital stay was (1.8±0.4)days(range:1 to 2 days), and the hospitalization cost was (34 499±20 330)yuan(range:11 724 to 73 488 yuan). No complications requiring special treatment outside the conventional pathway were observed during the hospital stay and follow-up period. At the 2-week outpatient follow-up, no significant abnormalities were found in all patients, and the wound healing was good.Conclusions:The daytime liver resection surgery based on the ERAS whole-process management plan has shown good feasibility in clinical practice. It helps to simplify medical process, shorten hospital stay, and reduce medical costs.

Metabolic-bariatric surgery(MBS)has become an important treatment for pathological obesity and metabolic diseases.However,common postoperative nutritional complications—such as protein-energy malnutrition,iron deficiency anemia,and vitamin B12 deficiency—significantly affect patients' long-term prognosis.Traditional nutritional management models rely on static monitoring and standardized supplementation,which are insufficient to address individual variability and dynamic postoperative changes.Artificial intelligence(AI),through integrating multimodal data(such as biochemical indicators,imaging information,and wearable device monitoring)and intelligent modeling,offers new approaches for dynamic monitoring,risk prediction,and personalized intervention.Based on literature from 2017 to 2025,this article systematically evaluates the application of AI in perioperative nutritional management for MBS,covering key technologies including machine learning,deep learning,and natural language processing.It also analyzes current challenges in clinical translation,such as data fragmentation,lack of model interpretability,and limited long-term validation.In the future,enhanced multi-center collaboration,the development of standardized databases,and explainable models will be essential to advancing nutritional management in MBS from empirical practice to precision medicine.

AIM:To explore the mechanism of doxorubicin/copper(DOX/Cu)complex induced copper death in hepatocellular carcinoma cells.METHODS:Human hepatocellular carcinoma cell line Huh7 was treated with DOX/Cu 0,2.5,4,7.5,10 and 12.5 μmol/L.The cell viability was detected by CCK-8 method.The cell proliferation level was observed by laser microscopy and proliferation kit.The cell invasion ability was determined by cell scratch assay.The flow cytometry was used to de-tect intracellular reactive oxygen species(ROS)and copper ion levels.And the western blot was used to detect intracellular iron-sulfur cluster proteins expression levels.RESULTS:With the increase of DOX/Cu concentration,cell viability,cell prolifera-tion and invasion ability decreased gradually.The copper ion chelating agent(TTM)can significantly restore the effects of DOX/Cu on cell viability.After DOX/Cu treatment,intracellular copper ion and ROS levels related to coproptosis were significantly increased,accompanied by the loss of iron-sulfur cluster proteins.CONCLUSION:DOX/Cu can inhibit hepatocellular carcinoma cells through cuproptosis.

Objective:To systematically retrieve, evaluate, extract, and integrate the best evidence on sleep disordered breathing (SDB) assessment and intervention in stroke patients to inform clinical practice.Methods:Based on the "6S" evidence model, the literature on SDB assessment and intervention in stroke patients was searched top-down on clinical decision-making websites, guideline websites, professional association websites, and databases. The search period was from the establishment of the database to September 1, 2024. The evidence-based team was formed to select the corresponding tool for quality assessment based on the type of literature. The evidence was also summarized using the Joanna Briggs Institute Center for Evidence-Based Health Care quality level of evidence and grade of recommendation.Results:A total of twenty papers were included, including six expert consensus, six guidelines, four clinical decision-making, two evidence summaries, and two systematic reviews. Eighteen pieces of evidence were summarized around four aspects of assessment, treatment, adherence, and health education.Conclusions:The assessment and intervention of SDB in stroke patients is multifaceted. It is recommended that healthcare professionals dynamically observe the actual condition of patients when applying the evidence to scientifically assess and manage SDB in stroke patients.

Metabolic-bariatric surgery(MBS)has become an important treatment for pathological obesity and metabolic diseases.However,common postoperative nutritional complications—such as protein-energy malnutrition,iron deficiency anemia,and vitamin B12 deficiency—significantly affect patients' long-term prognosis.Traditional nutritional management models rely on static monitoring and standardized supplementation,which are insufficient to address individual variability and dynamic postoperative changes.Artificial intelligence(AI),through integrating multimodal data(such as biochemical indicators,imaging information,and wearable device monitoring)and intelligent modeling,offers new approaches for dynamic monitoring,risk prediction,and personalized intervention.Based on literature from 2017 to 2025,this article systematically evaluates the application of AI in perioperative nutritional management for MBS,covering key technologies including machine learning,deep learning,and natural language processing.It also analyzes current challenges in clinical translation,such as data fragmentation,lack of model interpretability,and limited long-term validation.In the future,enhanced multi-center collaboration,the development of standardized databases,and explainable models will be essential to advancing nutritional management in MBS from empirical practice to precision medicine.