Deep Learning Based Ultrasound Image Segmentation Method for Breast Cancer
For the problem of breast cancer ultrasound image segmentation,an algorithm based on Mask Region Convolutional Neural Network(Mask R-CNN)is proposed.Firstly,the algorithm design of ultrasound image segmentation network was carried out;secondly,the organized dataset was made into the form of image segmentation of COCO dataset,and the training and test sets were extracted;finally,the network was trained by using the migration learning method,and the resultant graphs of breast cancer ultrasound image segmentation of all test sets were derived,and the image was evaluated using Dice coefficients,intersections and concatenation ratios and the average class accuracy rate as evaluation metrics.The segmentation results were evaluated using Dice coefficient,intersection ratio and average class accuracy as evaluation metrics.The results show that the Dice coefficient of the algorithm reaches 0.91,which significantly improves the segmentation accuracy of lesion tissue in breast cancer ultrasound images.
万方数据
维普
