基于SE注意力机制与残差网络的胃溃疡良恶性分类研究

Objective: This paper addresses the problems of low efficiency of endoscopic examination of gastric ulcer images and the misdiagnosis and missed diagnosis of malignant gastric ulcers, and proposes a benign and malignant classification algorithm for gastric ulcers based on convolutional neural networks. Methods: This paper has conducted in-depth research on the problem of small total amount of medical image data and uneven distribution of the number of sample categories. The bottleneck structure, SE attention mechanism, and residual structure structure of convolutional neural network models are discussed in general digestive endoscopy. Image classification of benign and malignant gastric ulcer lesions. Results: The model\'s F1-score score was 92.58%, the sensitivity was 90.95%, the specificity was 85.44%, and the accuracy was 94.28%. Conclusion: The convolutional neural network-based benign and malignant classification algorithm proposed in this study can effectively classify benign and malignant gastric ulcers.