【摘要】 目的:将深度学习技术与胃镜图像相结合,准确稳定地检测胃癌病灶的区域和类别,提高检测的效率与准确性,辅助医生进行早期胃癌的筛查和诊断。方法:对具有不同类别胃癌病灶的胃镜图像进行标注,构建胃镜图像数据集,使用Faster RCNN目标检测算法进行训练,不断优化参数,最终形成一个最优的目标检测模型。结果与结论:目标检测模型可以较好地完成对胃癌病灶区域的检测和病灶类别量化分级的工作,可以辅助医生进行诊断,具有一定的临床价值和科研价值。更多还原

【Abstract】 Objective: Deep learning technology is combined with magnifying gastroscope images to accurately and stably detect the area and category of gastric cancer lesions, improve the efficiency and accuracy of detection and assist doctors in screening and diagnosis of early gastric cancer. Methods: Labeling the gastroscope images which have different categories of gastric cancer lesions, and constructing the gastroscope image data set. Using the Faster RCNN target detection algorithm for training, and optimizing the parameters constantly to form an optimal target detection model. Results and Conclusion: The target detection model can well complete the detection of gastric cancer lesion areas and the quantification and classification of lesions. It can assist doctors in the diagnosis and has certain clinical and scientific research value.更多还原