A method, apparatus, storage medium, and device for polyp segmentation

Abstract

The application discloses a polyp segmentation method and device, a storage medium and equipment, and belongs to the polyp segmentation technical field in image processing. The method comprises the following steps: extracting high-frequency edge features based on a medical image; performing feature extraction on the image to obtain four groups of feature maps; gradually inputting the last two-stage feature maps into a global semantic aggregation module and a third multi-scale channel perception module to obtain third deep-layer semantic features; inputting the second-stage feature maps, the high-frequency edge features and the third deep-layer semantic features into a second semantic guided boundary exploration module and a second multi-scale channel perception module to obtain second shallow-layer semantic features; inputting the first-stage feature maps, the high-frequency edge features and the second shallow-layer semantic features into a first semantic guided boundary exploration module and a first multi-scale channel perception module to obtain first shallow-layer semantic features; and obtaining a polyp segmentation map based on the third deep-layer semantic features and the first shallow-layer semantic features. The application can enhance the robustness of the model and effectively improve the polyp segmentation precision.

Images