patient
The present single-center retrospective study was conducted by the First Affiliated Hospital of Bengbu Medical University (Bengbu, China). An electronic ERCP database was used to obtain demographic information and ERCP procedure details of MOJ patients who underwent ERCP from January 2017 to October 2022. Participants in this study met the following requirements, including age requirement ≥18 years and diagnosis of MOJ by clinical or histopathological examination. Patients were excluded from this study for the following reasons: (a) acute cholangitis before ERCP, (b) organ failure existed before ERCP, (c) surgery was completed within 1 month before ERCP, (d) death from causes other than acute cholangitis after ERCP, and (e) lack of clinical data.
To determine the initial diagnosis and indication for ERCP, all patients with complete information from hospital records on general data, clinical examinations, and imaging tests were included, and contraindications were excluded.The study was approved by the Ethics Committee of the First Affiliated Hospital of Bengbu Medical University (Bengbu, China; approval number 2019KY030).
surgery
All ERCP procedures were performed or supervised by licensed physicians with at least 10 years of ERCP experience. The duodenoscope (Olympus JF-260/TJF-260) was slowly inserted through the mouth, entered the descending duodenum, revealed the duodenal papilla, and the short speculum was straightened to place the contrast catheter into the hole of the clamp placement tube of the endoscope. The angle button and clamp lifter were appropriately adjusted to ensure that the catheter was perpendicular to the open end of the papilla and placed in the bile duct or pancreatic duct within the papilla. After the guidewire was placed in the lesion, 10–20 ml of iohexol injection solution (GE Healthcare (Shanghai) Co., Ltd., specification: 50 ml) was injected for imaging, and the lesion was observed under X-ray. Depending on the patient's lesion, different endoscopic treatment methods were selected, and biliary plastic stents or metal stents were used for internal drainage.
Primary Study Endpoint
The end date of the study is December 2022, and the follow-up endpoint was 30 days after surgery with or without cholangitis. For the definition of cholangitis, please refer to the Tokyo Guidelines 2018.twenty two and the European Society for Gastrointestinal Endoscopy (ESGE) guidelines.twenty three: 24 hours after the end of surgery, body temperature is 38°C or higher and/or white blood cell count (WBC) is 10.0×109/L, elevated biochemical indices such as serum bilirubin and biliary enzymes reflecting cholestasis, and imaging changes related to the etiology of cholestasis.
Clinical Data Sets
The following information was collected and recorded: patients’ personal information (gender, age, BMI, medical history, surgical history), tumor-related indicators (tumor type, length of obstruction, location of obstruction, type of stent (single plastic stent or single metal stent), duration of obstruction), pre-ERCP blood routine indicators (white blood cell count, total bilirubin, direct bilirubin, albumin, blood glucose level), antibiotics used before ERCP, and operation time. A detailed description of the heterogeneous dataset, including 16 quantitative features and 11 qualitative features (n = 16 + 11 = 27), is shown in Table 1. Ultimately, 218 patients were enrolled in this study. The dataset was divided into a training set (70%) and a test set (30%) to evaluate the accuracy of the model.
Univariate statistical analysis
Univariate statistical analyses were performed using Python software (version 3.8.8). For normally distributed data, continuous variables were expressed as mean ± standard deviation (SD) and comparisons were performed using the t These categorical variables were examined with Pearson's chi-square test. Nonparametric tests were used to compare other continuous variables, which were expressed as median and interquartile range (IQR). The performance of each variable was compared by the area under the receiver operating characteristic curve (AUROC).
Model Development and Optimization
We applied six different ML classification methods: Random Forest (RF), Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), Naive Bayes (NB) with Gaussian class conditional density function, and Adaboost (AB). We used 10-fold cross-validation throughout the experiments. We selected AUROC as the primary evaluation metric. Parameter tuning of the RF model was completed by the BAS algorithm. Details of the hyperparameter settings and the BAS algorithm are provided in the supplementary material.
To make the ML model work more effectively, we evaluated the clinically significant variables and calculated their influence weights on the clinical outcomes. To do so, we applied the Shapley additive explanation (SHAP) and RF models to explore the clinical correlations of different variables.
The following experiment is an extension of the above experiment. Features were selected according to the evaluation of the relative importance of features by the RF model, and then the hyperparameters were tuned. In this experiment, this selection is made by the model. Table S2 shows the details of the parameters of the experiment. These parameters include the RF model parameters and features. 0 indicates that the feature was removed, and 1 indicates that the feature was reserved. Feature removal started from less relevant features. The AUROC of the test set was taken as the fitness value. Figure S1 shows a schematic diagram of the experiment.
Metrics
To evaluate the model performance, standard metrics were selected: accuracy (ACC), F1 score, mean squared error (MSE), and area under the receiver operating characteristic curve (AUROC) values.twenty fourThese metrics were used to display the final results of the classification. The model was implemented using Python 3.8.8. In this work, the Python language, Pandas, NumPy, and Sklearn packages were used.
Ethics approval and consent to participate
The study protocol conformed to the standards of the Declaration of Helsinki and was approved by the Ethics Committee of the First Affiliated Hospital of Bengbu Medical University (Bengbu, China; approval number 2019KY030).
The Clinical Medical Research Ethics Committee of the First Affiliated Hospital of Bengbu Medical University waived the need to obtain informed consent from study participants.
