MRI characteristics and malignancy risk prediction model for intraductal papillary mucinous neoplasm of the pancreas
10.3760/cma.j.cn115667-20211014-00176
- VernacularTitle:胰腺导管内乳头状黏液性肿瘤MRI影像特征及恶变风险预测模型建立
- Author:
Xu FANG
1
;
Jing LI
;
Tiegong WANG
;
Na LI
;
Yinghao MENG
;
Xiaochen FENG
;
Yun BIAN
;
Chengwei SHAO
;
Jianping LU
;
Li WANG
Author Information
1. 海军军医大学第一附属医院放射诊断科，上海 200433
- Keywords:
Pancreas;
Intraductal papillary mucinous neoplasm;
Magnetic resonance imaging;
Nomograms
- From:
Chinese Journal of Pancreatology
2021;21(6):426-432
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To investigate the MRI features of intraductal papillary mucinous tumor (IPMN) of the pancreas and establish a prediction model for predicting the malignancy risk.Methods:The clinical data of 260 IPMN patients who underwent MRI and pathological confirmed in the First Affiliated Hospital of Naval Medical University from October 2012 to April 2020 were retrospectively analyzed. According to the pathological results, all patients were divided into benign group (including IPMN with low-grade dysplasia) and malignant group (including IPMN with high grade dysplasia and invasive carcinoma). According to international consensus of prediction model modeling, patients were divided into training set and validation set in chronological order. A prediction model was developed based on a training set consisting of 193 patients (including 117 patients with benign IPMN and 76 patients with malignant IPMN) between October 2012 and April 2019, and the model was validated in 67 patients (including 40 patients with benign IPMN and 27 patients with malignant IPMN) between May 2019 and April 2020. The multivariable logistic regression model was adopted to identify the independent predictive factors for IPMN malignancy and establish and visualized a nomogram. The ROC was drawn and AUC was calculated. The decision curve analysis was used to evaluate its clinical usefulness.Results:The IPMN type, cyst size, thickened cyst wall, mural nodule size, diameter of main pancreatic duct (MPD) and the abrupt change in the caliber of the MPD with distal pancreatic atrophy in the training set and validation set, and jaundice and lymphadenopathy in the training set were significantly different between benign group and malignant group ( P<0.05). The multivariable logistic regression model of characteristics included the jaundice, cyst size, mural nodule size ≥5 mm, the abrupt change in caliber of the MPD with distal pancreatic atrophy were independent risk factors for IPMN maligancy. The model for predicting IPMN malignancy was -0.35+ 2.28×(jaundice)+ 1.57×(mural nodule size ≥5 mm)+ 2.92×(the abrupt change in caliber of the MPD with distal pancreatic atrophy)-1.95×(cyst <3 cm)-1.05×(cyst≥3 cm). The individualized prediction nomogram using these predictors of the malignant IPMN achieved an AUC of 0.85 (95% CI 0.79-0.91) in the training set and 0.84 (95% CI 0.74-0.94) in the validation set. The sensitivity, specificity and accuracy of the training set were 72.37%, 85.47% and 80.31%, respectively. The sensitivity, specificity and accuracy of the validation set were 81.48%, 75.00% and 77.61%, respectively. The decision curve analysis demonstrated that when the IPMN malignancy rate was >0.16, the nomogram diagnosing IPMN could benefit patients more than the strategy of considering all the patients as malignancy or non-malignancy. Conclusions:The nomogram based on MRI features can accurately predict the risk of malignant IPMN, and can be used as an effective predictive tool to provide more accurate information for personalized diagnosis and treatment of patients.
