Ultrasonic image breast tumor classification method based on feature fusion and attention mechanism

The invention discloses an ultrasonic image breast tumor classification method based on feature fusion and an attention mechanism, and belongs to the field of computer-aided diagnosis of breast tumors, and the method comprises the steps: building a breast ultrasonic image data set, and segmenting a breast tumor lesion region based on a MobileNet trunk and feature fusion coding-decoding lesion segmentation network; the prior information of tumor segmentation is used to carry out segmentation enhancement preprocessing on the breast ultrasound image. Performing noise removal, detail enhancement and edge enhancement on the original image by using methods of morphological operation, bilateral filtering, CLAHE and the like, and generating an RGB image for tumor classification; a lightweight visual attention network VAN is improved and designed for tumor classification, a hybrid feedforward neural network module is introduced, local and detail information of an image is extracted by using a big kernel attention mech