DG-GAN: A High Quality Defect Image Generation Method for Defect Detection.

The surface defect detection of industrial products has become a crucial link in industrial manufacturing. It has a series of chain effects on the control of product quality, the safety of the subsequent use of products, the reputation of products, and production efficiency. However, in actual production, it is often difficult to collect defect image samples. Without a sufficient number of defect image samples, training defect detection models is difficult to achieve. In this paper, a defect image generation method DG-GAN is proposed for defect detection. Based on the idea of the progressive generative adversarial, D2 adversarial loss function, cyclic consistency loss function, a data augmentation module, and a self-attention mechanism are introduced to improve the training stability and generative ability of the network. The DG-GAN method can generate high-quality and high-diversity surface defect images. The surface defect image generated by the model can be used to train the defect detection model and improve the convergence stability and detection accuracy of the defect detection model. Validation was performed on two data sets. Compared to the previous methods, the FID score of the generated defect images was significantly reduced (mean reductions of 16.17 and 20.06, respectively). The YOLOX detection accuracy was significantly improved with the increase in generated defect images (the highest increases were 6.1% and 20.4%, respectively). Experimental results showed that the DG-GAN model is effective in surface defect detection tasks.

[Liquid-based cytology diagnosis of endoscopic ultrasound-guided fine needle aspiration of pancreatic lesions].

OBJECTIVE: To investigate the diagnostic value of liquid-based cytology test (LCT) in pancreatic lesions sampled by ultrasound-guided fine needle aspiration (EUS-FNA). METHODS: A retrospective analysis of 556 cases of LCT smears sampled by EUS-FNA of pancreatic lesions was performed, and 164 cases had histologic diagnosis with subsequent surgical resection or biopsy and immunohistochemistry. The accuracy of the cytologic diagnosis was assessed using the histologic diagnosis as the gold standard. The discrepant cases were reviewed to identify sources of errors. RESULTS: The satisfactory rate for EUS-FNA was 96.0%(534/556). The sensitivity, specificity, positive predictive value, negative predictive value and diagnostic accuracy were 87.7%(128/146), 13/16, 97.7%(128/131), 41.9%(13/31) and 87.0%(141/162) respectively. The diagnostic accuracy was lower in cystic lesions than that in solid lesions. The LCT sensitivities of adenocarcinoma, lymphoma and neuroendocrine tumors were higher than those of cystic tumors and mesenchymal tumors. False positive diagnosis was mainly due to epithelial abnormalities in inflammatory reaction. False negative diagnosis was mainly due to scanty or lack of tumor cells in the smears, or mild atypia that was insufficient for diagnosis. CONCLUSIONS: EUS-FNA is a valuable tool for the diagnosis of pancreatic lesions. Standardized terminology and nomenclature are helpful to improve the diagnostic accuracy.