Abstract

Info

Title:: Space-frequency domain difference with super-resolution feedback network for breast microcalcification detection

Author(s):: XING Suxia1; SHEN Nan1; LIU Zijiao1; JU Zihan1; HE Xiangping2; PAN Ziyan1; WANG Yu1; 1. School of Artificial Intelligence, Beijing Technology and Business University, Beijing 100048, China 2. Department of Breast,Haidian District Maternal and Child Health Care Hospital, Beijing 100080, China

Keywords:: Keywords: breast cancer microcalcification difference image technique super-resolution reconstruction GAB-DS

Abstract:: Abstract: The early symptoms of breast cancer are mainly characterized by microcalcifications in mammogram. The true and false positive detections of microcalcifications are of great significance for the early screening of breast cancer. After preprocessing the images from DDSM breast data for removing noise and irrelevant tissue interference, the suspected microcalcifications are segmented based on the space-frequency domain difference image technique, obtaining a sensitivity of 91.00% and a false positive rate of 34.00%. According to the centroid position of the suspected point, the region of interest is automatically intercepted. Then, super-resolution feedback network algorithm is used to realize the super-resolution reconstruction of the microcalcifications. Finally, the texture features of the region of interest are extracted, and the Gentle AdaBoost algorithm is combined with the single layer decision tree algorithm to construct a strong classifier GAB-DS for classifying the regions, separating the microcalcifications from the normal tissues. The accuracy, sensitivity and specificity achieved by GAB-DS model were 96.25%, 94.38% and 98.13%, respectively. The experimental results show that the model has superior performance in detecting microcalcifications, and can be used to assist in clinical detection and diagnosis of breast cancer, with high application value in clinic.