One-year recurrence rate of acute pancreatitis at about 20%. 36% of the patients with recurrent acute pancreatitis will develop into chronic pancreatitis. In addition to negative impact on patient's quality of life, chronic pancreatitis is also associated with the occurrence of pancreatic cancer. The etiology of recurrent acute pancreatitis (RAP) can be divided into mechanical obstructive factors (e.g. cholelithiasis, cholestasis), metabolic abnormality and toxic substance factors (e.g. hyperlipidemia and alcoholism), and other or idiopathic factors. At present, the diagnosis and treatment of RAP remains highly challenging. Early identification and intervention on risk factors of recurrence will be effective in reducing incidence and improving prognosis. Contrast-enhanced Computed Tomography (CT) can not only provide more imaging information and further assess the severity of acute pancreatitis, but also aid in the differentiation of other diseases associated with acute abdominal pain. In addition, radiomics based on raw radiographic data has become a research hotspot in recent years. The purpose of this study is to establish and validate a deep learning model based on high concentration iopromide-enhanced abdominal CT images which is designed to predict the recurrence of pancreatitis in patients with first episode of pancreatitis within the 1-year follow-up period.
Primary objective(s) To evaluate the sensitivity and specificity of the deep-learning integrated model established with relevant clinical factors and radiomic features based on the high concentration (370 mgI/ml) Iopromide-enhanced pancreatic CT obtained within 14 days after the first onset of symptoms for quantitative prediction of (the first) recurrence of acute pancreatitis in 12 months follow-up period. Sample Size: According to previously published data, the average time of occurrence of RAP is 12.5 ± 3.6 months and one-year recurrence rate of acute pancreatitis is about 20%. In addition, the recurrence rate is estimated to be about 17% within the 12-month follow-up window in this study, when taking into account the clinical experience of our hospital. The calculation parameters for sample size of the training set in the study are as follows: 1. Z1-α/2 is1.96 at α=0.05 2. L, the width of the acceptable 95% confidential interval of sensitivity or specificity, 0.03-0.1 3. The sensitivity is 0.85, the specificity is 0.98, and the disease prevalence is 0.17 Calculated based on sensitivity, N1= 1.962X0.85 X (1-0.85)/0.12 X 0.17= 0.490/0.0017=288 Calculated based on specificity, N2=1.962X0.98 X (1-0.98)/0.12 X (1-0.17) = 0.075/0.008=61 The sample size of the training set is 288 x 1.2 = 346 considering a dropout rate of 20% in the study. The training set, test set and validation set are estimated in a ratio of 5:2:3. The total sample size for this prospective study is 694. According to the order of patient enrollment, the last 200 patients recruited will form the validation set. Statistical Analyses: At baseline and follow-up, descriptive statistics will be used to describe the entire population and subgroups of interest. Summary statistics such as mean, median, standard deviation and range will be used to describe continuous variables. Categorical variables will be presented in a frequency table. * Primary endpoint analysis For patients with acute pancreatitis undergoing enhanced CT scan, the model that used the combination of radiomics and clinical features is used to predict the recurrence probability of acute pancreatitis within 12 months. The sensitivity and specificity of prediction and corresponding 95% CI are calculated. * Secondary endpoint analysis Chi-square testing for all potential clinical risk factors included (as described in the Chapter on Variables and Criteria Used in Determining Primary Endpoints). The variables with p\< 0.05 are analyzed for multivariate logistic regression and clinical modeling. Also based on the logistic regression model, a combination model of radioomic features and clinical factors is established. The sensitivity, specificity and corresponding 95% CI for prediction of recurrence within 1, 3, 6 and 12 months are calculated based on the model that used both clinical features and/or radiomics features. Only the first recurrence is calculated. Brief statistics of the quality of CT images will be provided. • Baseline demographic characteristics Demographic and baseline characteristics will be summarized descriptively. Sensitivity= TP/(TP+FN) Specificity=TN/(TN+FP) Accuracy = (TP+TN)/(TP+FN+TN+FP) TP=True positive TP=True negative FN=False negative FP=False positive TP+FN+TN+FP=Total number of patients Statistical analyses are performed using R software (R Core Team, Vienna, Austria) version 3.4.3 All tests are two-sided. A P value \< 0.05 is considered statistically significant. All therapies will be coded using the World Health Organization - Drug Dictionary (WHO-DD). Medical history and any disease will be coded using the most current version of ICH Medical Dictionary for Regulatory Activities (MedDRA).
Study Type
OBSERVATIONAL
Enrollment
694
A total of 694 consecutive eligible patients with first episode of "acute pancreatitis" that recommended for iopromide 370 mgI/ml enhanced pancreatic CT scan within 14 days after onset of symptoms (the sample size ratio of training set: test set: validation set is 5:2:3) will be sequentially enrolled at 10 sites (This study is observational in nature. The routine enhanced pancreatic CT protocol at all selected sites is similar or equivalent to that used in the PI site in the study). Relevant clinical information of the enrolled subjects is collected. Radiomics features are extracted from the selected regions of interest on pancreatic CT images and classified. The subjects are followed up for 12 months and divided into recurrent group and non-recurrent group according to the first recurrence status.
The sensitivity and specificity of the model established with relevant clinical factors and radiomic features
Image acquisition at each site is performed by an independent radiologist with ≥ 5 years of work experience. Radiomics features will be extracted from 370 mgI/ml Iopromide-enhanced pancreatic CT obtained within 14 days after the first onset of symptoms to quantitative predict the first recurrence of acute pancreatitis in 12 months follow-up period.Sensitivity= TP/(TP+FN) Specificity=TN/(TN+FP) Accuracy = (TP+TN)/(TP+FN+TN+FP) TP=True positive TP=True negative FN=False negative FP=False positive TP+FN+TN+FP=Total number of patients.
Time frame: 12 months
The total number of subjects who developed the first recurrence of acute pancreatitis (subjects with multiple recurrences, calculated and analyzed according to the time of the earliest recurrence) within 3, 6 and 12 months
Recurrent acute pancreatitis: relapse occurs more than three months after disappearance of symptoms of the first episode, with the exclusion of re-hospitalization due to local or systemic complications of the initial episode and chronic pancreatitis.
Time frame: in 12 months follow up period
The sensitivity and specificity of the deep-learning integrated model established with relevant clinical factors and radiomic features to quantitative predict overall recurrence of acute pancreatitis in 3 and 6 months follow-up period;
Sensitivity= TP/(TP+FN) Specificity=TN/(TN+FP) Accuracy = (TP+TN)/(TP+FN+TN+FP) TP=True positive TP=True negative FN=False negative FP=False positive
Time frame: 3 months, 6 months
The image quality of Iopromide-enhanced pancreatic CT images obtained within 14 days after onset of symptoms
The Imaging Department, Peking University People's Hospital served as quality control center. Objective quantitative evaluation and subjective evaluation of the quality of the images from all sites are performed by two radiologists with ≥ 10 years of abdominal imaging experience. A 4-point scale is used for subjective evaluation of the overall image quality in terms of noise, sharpness and contrast. Where, a score of 0 denotes poor image quality; 1 denotes fair image quality; 2 denotes good image quality; 3 denotes excellent image quality. The images with a score of 2 to 3 are classified into the high-quality image set.
Time frame: 14 days
The sensitivity and specificity of the model determined by radiomics features extracted from images with high quality scores (2-3 points)
Sensitivity= TP/(TP+FN) Specificity=TN/(TN+FP) Accuracy = (TP+TN)/(TP+FN+TN+FP) TP=True positive TP=True negative FN=False negative FP=False positive
Time frame: 3 months, 6 months and 12 months
The sensitivity and specificity of the radiomics features in predicting (the first) recurrence of different types and severities of acute pancreatitis at 3, 6 and 12 months
Sensitivity= TP/(TP+FN) Specificity=TN/(TN+FP) Accuracy = (TP+TN)/(TP+FN+TN+FP) TP=True positive TP=True negative FN=False negative FP=False positive
Time frame: 3 months, 6 months and 12 months
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.