1.Explainable Machine Learning Model for Predicting Prognosis in Patients with Malignant Tumors Complicated by Acute Respiratory Failure: Based on the eICU Collaborative Research Database in the United States
Zihan NAN ; Linan HAN ; Suwei LI ; Ziyi ZHU ; Qinqin ZHU ; Yan DUAN ; Xiaoting WANG ; Lixia LIU
Medical Journal of Peking Union Medical College Hospital 2026;17(1):98-108
To develop and validate a model for predicting intensive care unit (ICU) mortality risk in patients with malignant tumors complicated by acute respiratory failure (ARF) based on an explainable machine learning framework. Clinical data of patients with malignant tumors and ARF were extracted from the eICU Collaborative Research Database in the United States, including demographic characteristics, comorbidities, vital signs, laboratory test indicators, and major interventions within the first 24 hours after ICU admission.The study outcome was ICU death.Enrolled patients were randomly divided into a training set and a validation set at a ratio of 7:3.Predictor variables were selected using least absolute shrinkage and selection operator (LASSO) regression.Five machine learning algorithms-extreme gradient boosting (XGBoost), support vector machine (SVM), Logistic regression, multilayer perceptron (MLP), and C5.0 Decision Tree-were employed to construct predictive models.Model performance was evaluated based on the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and other metrics.The optimal model was further interpreted using the Shapley additive explanations (SHAP) algorithm. A total of 3196 patients with malignant tumors complicated by ARF were included.The training set comprised 2, 261 patients and the validation set 935 patients; 683 patients died during ICU stay, while 2513 survived.LASSO regression ultimately selected 12 variables closely associated with patient ICU outcomes, including sepsis comorbidity, use of vasoactive drugs, and within the first 24 hours after ICU admission: minimum mean arterial pressure, maximum heart rate, maximum respiratory rate, minimum oxygen saturation, minimum serum bicarbonate, minimum blood urea nitrogen, maximum white blood cell count, maximum mean corpuscular volume, maximum serum potassium, and maximum blood glucose.After model evaluation, the XGBoost model demonstrated the best performance.The AUCs for predicting ICU mortality risk in the training and validation sets were 0.940 and 0.763, respectively; accuracy was 88.3% and 81.2%;sensitivity was 98.5% and 95.9%.Its predictive performance also remained optimal in sensitivity analyses.SHAP analysis indicated that the top five variables contributing to the model's predictions were minimum oxygen saturation, minimum serum bicarbonate, minimum mean arterial pressure, use of vasoactive drugs, and maximum white blood cell count. This study successfully developed a mortality risk prediction model for ICU patients with malignant tumors complicated by ARF based on a large-scale dataset and performed explainability analysis.The model aids clinicians in early identification of high-risk patients and implementing individualized interventions.
2.Construction of novel transmembrane fusion antioxidant enzymes and their protective effect against hydrogen peroxide-mediated cellular oxidative damage.
Jianru PAN ; Ziyi ZHANG ; Jinnan CHU ; Yanan HAN ; Xueying ZHENG ; Shirong CAI ; Huocong HE
Chinese Journal of Biotechnology 2025;41(4):1547-1558
Reactive oxygen species (ROS) are major contributors to radiation therapy-induced side effects in cancer patients. A fusion antioxidant enzyme comprising glutathione S-transferase (GST), superoxide dismutase 1 (SOD1), and a transmembrane peptide has been shown to effectively mitigate ROS-induced damage. To enhance its targeting capability, the fusion protein was further modified by incorporating a matrix metalloproteinase-2/9 substrate peptide (X) and the transmembrane peptide R9, yielding the antioxidant enzyme GST-SOD1-X-R9 (GS1XR). This modification reduced its transmembrane ability in tumor cells, thereby selectively protecting normal cells from oxidative stress. However, the use of non-human GST poses potential immunogenicity risks. In this study, we employed seamless cloning technology to construct an expression vector containing the human GST gene to replace the non-human GST gene, and then expressed and purified novel fusion antioxidant enzymes GS1R and GS1XR. The protective effects of newly constructed GS1R and GS1XR against hydrogen peroxide (H2O2)-induced oxidative damage in L-02 cells were then evaluated using GS1 as a control. Enzymatic activity assays revealed that the specific activity of GST in GS1XR remained unchanged compared to the unmodified protein, while SOD activity was enhanced. Exposure to 200 μmol/L H₂O₂ transiently activated the nuclear factor-erythroid 2-related factor 2 (Nrf2) pathway; however, this activation diminished after 24 h, reducing cell viability to 48.4%. Both GS1R and GS1XR effectively scavenged intracellular ROS, directly counteracting oxidative stress and promoting Nrf2 nuclear translocation, thereby activating antioxidant pathways and restoring cell viability to normal levels. The two enzymes showed comparable efficacy. In contrast, GS1, lacking transmembrane capability, was restricted to scavenging extracellular ROS and provided only limited protection. In conclusion, both novel fusion antioxidant enzymes demonstrated significant potential in safeguarding normal cells from ROS-mediated oxidative damage. The findings provide a foundation for further investigation in related field.
Humans
;
Oxidative Stress/drug effects*
;
Hydrogen Peroxide
;
Antioxidants/metabolism*
;
Glutathione Transferase/metabolism*
;
Recombinant Fusion Proteins/pharmacology*
;
Superoxide Dismutase-1
;
Reactive Oxygen Species/metabolism*
;
Superoxide Dismutase/biosynthesis*
3.Research on Five Proven Prescriptions for the Treatment of Epidemic Diseases in Wucheng Hanjian
Ziyi HAN ; Qi-Hang WANG ; Yiliang XIONG
Journal of Zhejiang Chinese Medical University 2024;48(2):224-227
[Objective]To explore the six medicinal ingredients,namely"(臧)Zangdu,Luru,Shashen,Baishashen,Shuchong and Fangji(ji)",in the Wucheng Hanjian prescription,in order to restore their original appearance.[Methods]Through literature search and double evidence method,a large number of relevant papers and books were consulted,and the research results were compared and summarized.[Results]This prescription consists of twenty-six medicinal ingredients,and the methods of preparation are diverse."(臧)Zangdu"should be"Daizhe"used by modern people;"Shashen"should be"Nanshashen",and"Baishashen"should be"Beishashen";"Shuchong"should be"Zhechong";there are two possible specific medicinal ingredients for"Luru",one is"Qucao",and the other is"lvru";"Fangji(ji)"should be"Fangkui".This prescription is used to treat"cold and dampness epidemic"with the etiology of cold and dampness,the main symptoms of constipation,coughing,and body pain,and the pathogenesis of external cold,internal drinking,and heat stagnation.[Conclusion]The research and restoration of the true appearance of the Wucheng Hanjian prescription is not only beneficial for further exploring the development process of traditional medicine,but also has practical reference significance for modern research on Chinese medicine,and also has certain value for the tracing research of Zhe school of traditional Chinese medicine.
4.Evaluation value of contrast-enhanced ultrasound and contrast-enhanced spiral CT in the TACE treatment of primary hepatocellular carcinoma
Zongbao HAN ; Xi LI ; Ziyi LIANG ; Rengui WANG
Chinese Journal of Radiological Health 2024;33(5):600-606
Objective Analyze and compare the therapeutic value of contrast-enhanced ultrasound (CEUS) and enhanced spiral CT in patients with primary hepatocellular carcinoma (PHC) undergoing transcatheter arterial chemoembolization (TACE). Methods A retrospective analysis was conducted on 108 PHC patients (122 lesions) treated at Beijing Shijitan Hospital affiliated with Capital Medical University from July 2021 to June 2023. All patients were treated with TACE. After treatment, CEUS and contrast-enhanced spiral CT examinations were performed to evaluate the inactivation status of tumor lesions, with pathological results as the gold standard. The receiver operating characteristic curve and decision tree model were used to comprehensively evaluate the diagnostic value of CEUS, contrast-enhanced spiral CT, and their combination in PHC. Results The pathological results showed that 79 lesions remained active (positive) and 43 lesions were completely inactivated (negative). The areas under the receiver operating characteristic curve for evaluating PHC treatment efficacy by CEUS, contrast-enhanced spiral CT, and their combination were 0.975, 0.918, and 0.994, respectively. The sensitivity, positive predictive value, and actual application accuracy of CEUS and the combined method were higher than those of contrast-enhanced spiral CT (P < 0.05). A decision tree diagnostic model for PHC treatment efficacy was constructed using CEUS, contrast-enhanced spiral CT, and their combination as independent variables and PHC treatment efficacy as the dependent variable. The misjudgment rates were 3.3%, 10.7%, and 0.8%, respectively, and the prediction accuracy was 96.7%, 89.3%, and 99.2%, respectively. Conclusions CEUS and the combination of CEUS and contrast-enhanced spiral CT have high evaluation value for PHC treatment efficiency.
5.Evaluation value of contrast-enhanced ultrasound and contrast-enhanced spiral CT in the TACE treatment of primary hepatocellular carcinoma
Zongbao HAN ; Xi LI ; Ziyi LIANG ; Rengui WANG
Chinese Journal of Radiological Health 2024;33(5):600-606
Objective Analyze and compare the therapeutic value of contrast-enhanced ultrasound (CEUS) and enhanced spiral CT in patients with primary hepatocellular carcinoma (PHC) undergoing transcatheter arterial chemoembolization (TACE). Methods A retrospective analysis was conducted on 108 PHC patients (122 lesions) treated at Beijing Shijitan Hospital affiliated with Capital Medical University from July 2021 to June 2023. All patients were treated with TACE. After treatment, CEUS and contrast-enhanced spiral CT examinations were performed to evaluate the inactivation status of tumor lesions, with pathological results as the gold standard. The receiver operating characteristic curve and decision tree model were used to comprehensively evaluate the diagnostic value of CEUS, contrast-enhanced spiral CT, and their combination in PHC. Results The pathological results showed that 79 lesions remained active (positive) and 43 lesions were completely inactivated (negative). The areas under the receiver operating characteristic curve for evaluating PHC treatment efficacy by CEUS, contrast-enhanced spiral CT, and their combination were 0.975, 0.918, and 0.994, respectively. The sensitivity, positive predictive value, and actual application accuracy of CEUS and the combined method were higher than those of contrast-enhanced spiral CT (P < 0.05). A decision tree diagnostic model for PHC treatment efficacy was constructed using CEUS, contrast-enhanced spiral CT, and their combination as independent variables and PHC treatment efficacy as the dependent variable. The misjudgment rates were 3.3%, 10.7%, and 0.8%, respectively, and the prediction accuracy was 96.7%, 89.3%, and 99.2%, respectively. Conclusions CEUS and the combination of CEUS and contrast-enhanced spiral CT have high evaluation value for PHC treatment efficiency.
6.Analysis of the diagnostic efficiency of wrist joint ulnar deviation supination 45° palmar oblique position for scaphoid waist fracture and displacement
Shifei CHEN ; Jun ZHANG ; Shuman HAN ; Wenjuan WU ; Zhiwei ZHONG ; Jing LI ; Zhe GUO ; Huizhao WU ; Ziyi QIAO
Journal of Practical Radiology 2024;40(5):805-808
Objective To investigate the clinical application value of wrist joint ulnar deviation supination 45° palmar oblique posi-tion in the diagnosis of scaphoid waist fracture and displacement.Methods The imaging and clinical data such as digital radiography(DR),CT of 93 wrist joint trauma patients were analyzed.The four position views including wrist joint anteroposterior+lateral view,scaphoid position,wrist joint ulnar deviation supination 45° palmar oblique position,scaphoid position+wrist joint ulnar devia-tion supination 45° palmar oblique position were analyzed by three readers.The consistency of the evaluation among different readers and the diagnostic efficiency of the diagnosis of scaphoid fracture and displacement were compared.Results The inter-observer agreement,sensitivity,specificity,accuracy,and other diagnostic efficiency of scaphoid waist fracture and displacement was evaluated,wrist joint ulnar deviation supination 45° palmar oblique position+scaphoid position and wrist joint ulnar deviation supination 45° palmar oblique position were better than those of scaphoid position and wrist joint anteroposterior+lateral view.The combination of wrist joint ulnar deviation supination 45° palmar oblique position+scaphoid position obtained the best diagnostic efficiency.Conclusion The wrist joint ulnar deviation supination 45° palmar oblique position shows the long axis of the scaphoid,which has a high diagnostic efficiency in the diagnosis of scaphoid waist fracture and displacement and would be used as a useful supplement to other scaphoid imaging.
7.A cortical thickness study of insomnia disorder patients based on MRI
Wei CHEN ; Hui WANG ; Ziyi LIU ; Yu SHANG ; Haining LI ; Wenxuan HAN ; Qiange ZHU ; Ming ZHANG
Journal of Practical Radiology 2024;40(11):1766-1769
Objective To investigate the changes of cortical thickness in patients with insomnia disorder(ID).Methods High-resolution MRI data were collected from 32 ID patients(ID group)and 30 healthy controls(HC)(HC group).The cortical thickness of both groups were analyzed using statistical parametric mapping 12(SPM12)software,while considering age,gender,and educational level as covariates.The cortical thickness in brain regions showed statistically significant differences was extracted for Pearson's correla-tion analyses with sleep and mood-related scales.Results Compared with the HC group,the ID group exhibited significantly decreased cortical thickness in brain regions such as the left insula,fusiform gyrus,orbitofrontal lobe,superior temporal gyrus,middle temporal gyrus,lateral occipital lobe and right caudal anterior cingulate gyrus[P<0.05,family-wise error(FWE)correction].Furthermore,reduced cortical thickness of the cingulate gyrus was negatively correlated with the Pittsburgh sleep quality index(PSQI)score(r=-0.437,P=0.012).Conclusion The cortical thickness of several brain regions associated with sleep and mood are significantly reduced in patients with ID,providing potential neuroimaging evidence for understanding the pathophysiological mechanism of ID.
8.Artificial intelligence federated learning system based on chest X-ray films for pathogen diagnosis of community-acquired pneumonia in children
Ziyi WEI ; Yi TANG ; Ze TENG ; Hongfeng LI ; Yun PENG ; Jiangfeng CAO ; Tianzi GAO ; Heng ZHANG ; Hongbin HAN
Chinese Journal of Interventional Imaging and Therapy 2024;21(6):368-373
Objective To explore the value of artificial intelligence federated learning system based on chest X-ray films for pathogen diagnosis of community-acquired pneumonia(CAP)in children.Methods Totally 900 cases of CAP children from 2 hospitals were retrospectively enrolled,including bacterial,viral and mycoplasma CAP(each n=300),and chest posteroanterior X-ray films were collected.Meanwhile,chest posteroanterior X-ray films of 5856 children from the publicly available dataset GWCMCx were collected,including 4273 CAP images and 1583 healthy chest images.All above 6756 images were divided into training set(n=5359)and validation set(n=1397)at the ratio of 8∶2.Then a pathogen diagnosis model of children CAP was established based on attention mechanism.Binary and ternary diagnostic algorithms were designed,and federated deployment training was performed.The efficacy of this system for pathogen diagnosis of children CAP was analyzed and compared with DenseNet model.Results Based on all data,the accuracy of the obtained artificial intelligence federated learning system model for diagnosing children CAP was 97.00%,with the area under the curve(AUC)of 0.990.Based on hospital data,the AUC of this system using single imaging data and clinical-imaging data for pathogen diagnosis of children CAP was 0.858 and 0.836,respectively,both better than that of DenseNet model(0.740,both P<0.05).Conclusion The artificial intelligence federated learning system based on chest X-ray films could be used for pathogen diagnosis of children CAP.
9.Cardiac CT angiography for left atrial appendage closure:Animal experiment
Ziyi ZHANG ; Binbin ZHANG ; Yingjian CUI ; He HAN ; Zhixiao XUE
Chinese Journal of Interventional Imaging and Therapy 2024;21(11):690-695
Objective To observe the value of cardiac CT angiography(CCTA)for left atrial appendage closure(LAAC).Methods Twelve Labradors were enrolled and underwent LAAC.CCTA examinations were performed before and 7,45 and 90 days after LAAC,and the left atrial appendage(LAA)parameters before and post LAAC occluder compression ratio were recorded.Ninety days after LAAC,the dogs were euthanized for gross anatomical observation,micro CT imaging and pathological examination,and the value of CCTA for LAAC was analyzed.Results Based on CCTA before LAAC,LAA opening depth was(18.8±2.6)mm,the minimum diameter was(16.3±2.0)mm and the maximum diameter was(22.6±3.2)mm,so 28 mm diameter occluders were selected for LAAC.Seven,45 and 90 days after LAAC,no displacement of occluder,thrombi nor device leakage was found on CCTA,while no significant difference of compression ratio was found among different time points(P>0.05),nor between those measured with CCTA and micro CT 90 days after LAAC(P>0.05).Ninety days after LAAC,gross dissection revealed complete endothelialization of occlude in 11 dogs and incomplete endothelialization in 1 dog.Pathological examination showed that collagen fibers arranged neatly,with endothelial cells covering the surface,and no inflammatory cell infiltration was observed.Conclusion CCTA could be used to observe LAA and surrounding structures before LAAC,hence guiding to select appropriate occluder,evaluate occlusion effect and monitor complications of LAAC in dogs.
10.Dermabrasion for the treatment of familial benign chronic pemphigus: a clinical observation of 6 cases
Yuqing SONG ; Nan YANG ; Linlin LIU ; Ziyi FENG ; Shixin HAN ; Meijuan ZHOU
Chinese Journal of Dermatology 2024;57(8):743-746
Objective:To analyze clinical efficacy and safety of dermabrasion for the treatment of familial benign chronic pemphigus.Methods:Clinical data were retrospectively collected from 6 patients with familial benign chronic pemphigus, who underwent dermabrasion in the Department of Dermatology, the First Affiliated Hospital of Dalian Medical University from February 2019 to July 2020. There were 3 males and 3 females, they were aged from 39 to 65 years, and their disease duration ranged from 10 to 40 years. All the patients were postoperatively followed up for 14 - 34 months. Response rates were calculated, and adverse reactions and recurrence were observed.Results:Dermabrasion was performed on the 6 patients in 55 body areas. After operation, complete recovery was observed in 36 areas (65.5%), marked improvement was noted in 13 areas (23.6%), and moderate improvement was observed in 6 areas (10.9%) ; there were no areas with no response, resulting in an overall response rate of 89.1%. Postoperatively, the patients experienced no obvious discomfort, and only slight hypopigmentation and mild scars remained. During the follow-up of 14 to 34 months, no recurrence was observed in the treated areas.Conclusion:Dermabrasion was safe and effective for the treatment of familial benign chronic pemphigus.

Result Analysis
Print
Save
E-mail