A Breast Cancer CT Image Detection Algorithm Based on Improved Feature Pyramid Network
Yongmei Zhang, Tong Cheng, Ruiqi Li · 2021
In medical diagnosis, medical imaging detection is often affected by training data and image backgrounds, which easily leads to lower detection accuracy. Aiming at the problem, a region convolution neural network breast cancer image detection algorithm based on improved feature pyramid structure is proposed. On the basis of the feature pyramid network structure, the bottom-up feature fusion is added, and the ideas of up-down fusion and bypass connection are adopted. And the fusion feature maps generate after two convolution operation. The improved algorithm is compared with the algorithms of original models for Mask R-CNN and ResNet-50 on the datasets DDSM (Digital Database for Screening Mammography). The experiment results show the correct detection rate of the improved algorithm is greatly enhanced in both normal and fuzzy positive sample images under the same feature extraction network and data set.