Gastric polyp detection method and device based on deep learning

Abstract

The invention provides a gastric polyp detection method and device based on deep learning, and belongs to the technical field of combination of medical image processing and deep learning. The method comprises the following steps: acquiring a marked gastric polyp image, and dividing the marked gastric polyp image into a training set and a test set; improving an original Faster R-CNN model, and establishing a gastric polyp detection model; wherein a residual network ResNet-101 is adopted to replace a backbone network VGG16 in the original Faster R-CNN model, and a RoI Align operation is adopted to replace a RoIPooling operation in the original Faster R-CNN model, and the ResNet-101 is adopted to replace a backbone network VGG16 in the original Faster R-CNN model; performing iterative training on the established gastric polyp detection model by using the training set; and detecting the gastric polyp in the test set image by using the trained gastric polyp detection model. By adopting the method, the omission ratio of polyps in the traditional gastroscopy process can be reduced, and the polyp detection accuracy is improved.

Description

technical field [0001] The invention relates to the technical field of combining medical image processing and deep learning, in particular to a method and device for detecting gastric polyps based on deep learning. Background technique [0002] Gastric polyps are a relatively common disease in gastroenteropathy, which can be divided into adenomatous gastric polyps and non-adenomatous gastric polyps. Adenomatous gastric polyps have a high chance of developing stomach cancer. According to statistics, there are approximately 951,000 newly diagnosed gastric cancer patients worldwide every year. It is very important to detect and remove gastric polyps that may be cancerous in time. Gastroscopy is the main method to find polyps. The doctor observes the gastroscopy image with the naked eye to determine whether there are polyps and find out the specific location of the polyps. Although this method has many advantages, factors such as heavy workload and inexperience may cause doc...

AI Technical Summary

Problems solved by technology

The detection of gastric polyps is different from that of intestinal polyps in that gastric polyps are smaller in size than intestinal polyps due to the many folds in the stomach wall, making detection of gastric polyps more difficult

The existing gastric polyp detection algorithm has low accuracy and high missed detection rate, which cannot be really applied to practical problems

Images