CT classification model of pancreatic serous cystic neoplasms and mucinous cystic neoplasms based on a deep neural network

Background At present, numerous challenges exist in the diagnosis of pancreatic SCNs and MCNs. After the emergence of artificial intelligence (AI), many radiomics research methods have been applied to the identification of pancreatic SCNs and MCNs. Purpose A deep neural network (DNN) model termed Multi-channel-Multiclassifier-Random Forest-ResNet (MMRF-ResNet) was constructed to provide an objective CT imaging basis for differential diagnosis between pancreatic serous cystic neoplasms (SCNs) and mucinous cystic neoplasms (MCNs). Materials and methods This study is a retrospective analysis of pancreatic unenhanced and enhanced CT images in 63 patients with pancreatic SCNs and 47 patients with MCNs (3 of which were mucinous cystadenocarcinoma) confirmed by pathology from December 2010 to August 2016. Different image segmented methods (single-channel manual outline ROI image and multi-channel image), feature extraction methods (wavelet, LBP, HOG, GLCM, Gabor, ResNet, and AlexNet) and classifiers (KNN, Softmax, Bayes, random forest classifier, and Majority Voting rule method) are used to classify the nature of the lesion in each CT image (SCNs/MCNs). Then, the comparisons of classification results were made based on sensitivity, specificity, precision, accuracy, F1 score, and area under the receiver operating characteristic curve (AUC), with pathological results serving as the gold standard. Results Multi-channel-ResNet (AUC 0.98) was superior to Manual-ResNet (AUC 0.91).CT image characteristics of lesions extracted by ResNet are more representative than wavelet, LBP, HOG, GLCM, Gabor, and AlexNet. Compared to the use of three classifiers alone and Majority Voting rule method, the use of the MMRF-ResNet model exhibits a better evaluation effect (AUC 0.96) for the classification of the pancreatic SCNs and MCNs. Conclusion The CT image classification model MMRF-ResNet is an effective method to distinguish between pancreatic SCNs and MCNs. Graphic abstract

Mucinous Cystic Lesions of the Pancreas an EUS-FNA-based Study with ClinicoPathological Follow Up

Background: Mucinous cystic lesions of pancreas harbor a pre-malignant potential thus necessitating their distinction from the non-mucinous ones. To make this distinction, EUS-FNA cytology along with cyst fluid CEA and amylase levels are utilized in addition to endoscopic and radiological findings. Evaluation of K-ras mutations has emerged as a useful adjunct for the evaluation of mucinous cystic neoplasms of the pancreas. Aim: We aimed to study mucinous cystic lesions of the pancreas diagnosed on EUS-FNA cytology, in conjunction with cyst fluid CEA and amylase levels and the frequency of K-ras mutation in a cohort of patients seen at the largest cancer hospital in our country. Materials and Methods: After approval from the institutional review board, all the cases of mucinous cystic lesions of pancreas evaluated between July 2005 and August 2019 were reviewed. Patient data, including age, gender, endoscopic and radiological findings, cytological and/or histological diagnosis, cyst fluid CEA, and amylase levels were collected. Results: Twenty-three patients enrolled in the study demonstrated an equal gender distribution with a mean age of 67.4 years. The sensitivity of EUS-FNA for mucinous cystic lesions of the pancreas was 84.6%. Cyst fluid CEA levels were elevated in some MCNs but not IPMNs resulting in a sensitivity of 37.5%. The specificity of cyst fluid amylase was 90%. K-ras mutation was found to have a sensitivity and specificity of 50% and 100% respectively, for mucinous lesions of the pancreas. Conclusion: EUS-FNA is a useful technique for evaluation of pancreatic cystic lesions, especially since cytological diagnosis can be augmented by cyst fluid CEA and amylase levels. K-ras analysis can add further to the diagnostic utility of EUS-FNA

The Yield of String Sign in Differentiating Mucinous from Non-Mucinous Pancreatic Cysts: A Retrospective Cross-Sectional Study

Background and Objectives: The diagnosis of pancreatic cysts is mostly based on a combination of morphological appearance and fluid analysis of amylase and carcinoembryonic antigen (CEA). We aimed to assess the capability of the string sign in differentiating mucinous from non-mucinous pancreatic cysts. Materials and Methods: All patients who were referred for endoscopic ultrasound (EUS) for pancreatic cysts assessment from 2015 to 2020 were retrospectively analyzed. Results: Our cohort consisted of 112 patients. Of them, 92 patients (82.1%) had mucinous cystic neoplasms (group A) and 20 patients (17.9%) had non-mucinous cystic neoplasms (group B). The average age in groups A and B was 71.3 and 60.4 years, respectively. String sign was positive in 47 patients (51.1%) and negative in 21 patients (22.8%) in group A, while in group B, string sign was negative in 19 patients (95%). String sign showed significant correlation with the diagnosis of mucinous cystic neoplasms (OR 64.2, 95% CI 8.1–508.6, p = 0.0001). Cytology confirmed mucinous cystic neoplasms that included 32 patients; the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of string sign for mucinous cystic neoplasms were high, reaching 93.8%, 85.7%, 96.8%, and 75%, respectively, with an excellent accuracy rate of 92.3%. Conclusions: The string sign is highly accurate for predicting pancreatic mucinous cystic neoplasms, and should be used as an important aid for improving diagnostic accuracy.

Intralobular Distribution of Ovarian-like Stroma in Pancreatic Mucinous Cystic Neoplasms: Hypothesis on Its Tumorigenesis

Pancreatic mucinous cystic neoplasm (MCN) harbors two histological components, tumor epithelia and ovarian-like stroma (OLS). To examine the tumorigenesis of pancreatic MCNs, this study analyzed the distribution, amount, immunohistochemical phenotype, presence of theca cells of the OLS, and the alteration of tumor epithelium of 29 surgically resected MCN cases and compared them with tumor sizes. Non-mucinous type epithelium was present in all low-grade MCNs but its ratio decreased with tumor size (p < 0.05), suggesting that epithelial mucinous changes are a progression phenomenon. The intralobular distribution of OLS was observed in 27.6 % of MCN cases and its existence related to a smaller size (p< 0.05), suggesting intralobular generation of MCNs. Nuclear expression of β-catenin was observed for OLS of everywhere, suggesting consistent activation of the Wnt pathway for OLS. Three MCN cases (10.3%) contained a-smooth muscle actin (SMA)-negative OLS, where OLS surrounding dilated pancreatic ducts or MCN cysts were a-SMA-positive and otherwise negative, suggesting that a-SMA-positivity is an acquired phenomenon of OLS. With this study, we could hypothesize that pancreatic MCNs may generate intralobularly. Epithelial mucinous change and a-SMA-positivity of OLS may be progression phenomena. This is the first study to show the intralobular distribution of OLS.

Differentiation and management of hepatobiliary mucinous cystic neoplasms: a single centre experience for 8 years

Background Hepatobiliary mucinous cystic neoplasms (H-MCNs) are relatively rare cystic neoplasms in the liver. The differential diagnosis of H-MCNs remains big challenging, and the management and prognosis between the hepatic simple cyst (HSC) and H-MCNs are quite different. This study aimed to present our experience in the management of H-MCNs and provide a preoperative H-MCNs risk prediction nomogram to differentiating H-MCNs from liver cystic lesions. Methods 29 patients diagnosed with H-MCNs and 75 patients diagnosed with HSC between June 2011 and June 2019 at Zhejiang University School of medicine, Sir Run-Run Shaw Hospital were reviewed in this study. We analyzed the demographic and clinicopathological variables. Results US, CT, and MRI could accurately diagnose only 3.4%, 46.1%, and 57.1% of H-MCNs, respectively. After univariate analysis and multivariate logistic regression analysis, the variables significantly associated with H-MCNs were enhancement after contrast (p = 0.009), tumour located in the left lobe (p = 0.02) and biliary ductal dilation (p = 0.027). An H-MCNs risk predictive nomogram was constructed, which showed excellent discrimination (areas under the receiver operating characteristic curve were 0.940) and consistent calibration between the predicted probability and actual probability. Conclusion Among patients with H-MCNs, the location of the tumour, enhancement in CT scan, and biliary duct dilation are significantly independent risk factors. The appropriate treatment of H-MCNs is radical resection. Using our Nomogram could facilitate screening and identification of patients with liver cystic lesions.

Mesenteric and Retroperitoneal Mucinous Cystic Neoplasms: A Case Series

Aims. Mucinous cystic neoplasms (MCNs) are cystic neoplasms with mucinous epithelium surrounded by ovarian-like stroma. Extraovarian MCN occurring in the liver and pancreas have been well characterized. However, only rare case reports of MCN arising outside of these locations have been reported. MCNs arising in unusual locations should enter the differential diagnosis of mucinous intra-abdominal tumors and must be distinguished from more common mimics. Therefore, we aimed to examine a series of MCNs of the retroperitoneum and mesentery to characterize the clinicopathologic features of this entity. Methods and results. Seven MCNs arising in the abdominal mesentery or retroperitoneum were retrospectively identified. A clinicopathologic, histologic, and immunohistochemical (keratin 7, keratin 19, keratin 20, calretinin, inhibin-α, steroidogenic factor-1 (SF-1), estrogen receptor (ER), progesterone receptor (PR), PAX8, CDX2, and CD10) analysis was performed. All 7 MCNs were from females with a median age of 41 years old and a median size of 8 cm. All cases demonstrated mucinous with or without concomitant non-mucinous epithelium overlying spindle cell ovarian-like stroma. Luteinized cells were noted. The epithelium was positive for keratin 7 and keratin 19 in all 7 cases, while the stroma expressed ER, PR, and SF-1 in all cases stained. Calretinin was focally positive in the stroma of 3 of 7 cases, while inhibin-α was focally expressed in 5 of 6 cases. Conclusions. These results highlight the clinicopathologic, histologic, and immunophenotypic similarities between MCNs of the mesentery, retroperitoneum, pancreas, and liver. Overlapping features suggest a common histogenesis for all MCNs, which could include periductal fetal mesenchyme, aberrant migration of primordial germ cells, or abnormal differentiation or metaplasia of the embryonic coelomic epithelium.