Background: Contrast-enhanced endoscopic ultrasound (CE-EUS) is useful for the differentiation of pancreatic tumors. Using deep learning for the segmentation and classification of pancreatic tumors might further improve the diagnostic capability of CE-EUS. Aims: The aim of this study was to evaluate the capability of deep learning for the automatic segmentation of pancreatic tumors on CE-EUS video images and possible factors affecting the automatic segmentation. Methods: This retrospective study included 100 patients who underwent CE-EUS for pancreatic tumors. The CE-EUS video images were converted from the originals to 90-s segments with six frames per second. Manual segmentation of pancreatic tumors from B-mode images was performed as ground truth. Automatic segmentation was performed using U-Net with 100 epochs and was evaluated with 4-fold cross-validation. The degree of respiratory movement (RM) and tumor boundary (TB) were divided into 3-degree intervals in each patient and evaluated as possible factors affecting the segmentation. The concordance rate was calculated using the intersection over union (IoU). Results: The median IoU of all cases was 0.77. The median IoUs in TB-1 (clear around), TB-2, and TB-3 (unclear more than half) were 0.80, 0.76, and 0.69, respectively. The IoU for TB-1 was significantly higher than that of TB-3 (p < 0.01). However, there was no significant difference between the degrees of RM. Conclusions: Automatic segmentation of pancreatic tumors using U-Net on CE-EUS video images showed a decent concordance rate. The concordance rate was lowered by an unclear TB but was not affected by RM.

Gastric neuroendocrine tumors (GNETs) are a heterogeneous group of neoplasm with varying biological characteristics. This study aimed to investigate the clinical features and outcomes of GNET patients after endoscopic diagnosis and treatment in a multicenter registry. Patients with GNETs confirmed histologically were recruited from 17 hospitals between January 2010 and April 2016 in Taiwan. Clinical, laboratory, radiological, endoscopic, pathological data, treatment strategies, follow-up periods, and survivals were collected retrospectively. Totally 187 (107 female, 80 male) patients were recruited. Mean ( ± standard deviation [SD]) age and size of tumors were 63.2-year-old ( ± 14.6) and 2.3-cm ( ± 3.0). World Health Organization (WHO) grading were 93 (49.7%) G1, 26 (13.9%) G2, 40 (21.4%) G3, and 28 (15.0%) unknown. G3 patients were older (mean ± SD, 71.6 ± 12.4 vs. 60.9 ± 14.3/56.7 ± 15.4 years), larger (6.1 ± 4.0 vs.1.2 ± 1.3/2.4 ± 2.5 cm), more distally located (35.0% vs. 7.6%/15.4%), lower proportion of superficial lesions (17.5% vs. 61.9%/53.8%) and higher rates of lymphovascular invasion (32.5% vs. 3.2%/7.7%) than G1/G2. There was no nodal or distant organ metastases despite different grading of lesions≦10 mm and those 20 mm (log-rank test P = .02). Male gender (P = .01), deeper invasion (P = .0001), nodal (P < .0001), and distant organ metastases (P = .0001) were associated with worse outcome. In conclusion, treatment strategies for GNET should be decided by grading, size, invasiveness, and LN metastasis risk. Curative endoscopic resection is feasible for G1/2 lesions less than 20 mm and limited to mucosa/submucosa layers without lymphovascular invasion.

Autoimmune pancreatitis (AIP) shares overlapping clinical features with pancreatic cancer (PC). Importantly, treatment of the two conditions is different. We investigated the clinical usefulness of ¹⁸F-fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) in patients with suspected AIP before treatment. From September 2008 to July 2016, 53 patients with suspected AIP at National Taiwan University Hospital had PET/CT prior to therapy to exclude malignancy and evaluate the extent of inflammation. Their scans were compared with those from 61 PC patients. PET imaging features were analyzed using logistic regression. Significant differences in pancreatic tumor uptake morphology, maximum standardized uptake value, high-order primary tumor texture feature (i.e. high-gray level zone emphasis value), and numbers and location of extrapancreatic foci were found between AIP and PC. Using the prediction model, the area under curve of receiver-operator curve was 0.95 (P < 0.0001) with sensitivity, specificity, positive predictive, and negative predictive values of 90.6%, 84.0%, 87.9%, and 87.5% respectively, in differentiating AIP from PC. FDG PET/CT offers high sensitivity, albeit slightly lower specificity in differentiating AIP from PC. Nonetheless, additional systemic inflammatory foci detected by the whole body PET/CT help confirm diagnosis of AIP in these patients before initiating steroid therapy, especially when biopsy is inconclusive.