Research And Implementation Of Object Detection Based Bilateral Fused Mammogram Mass Detection

Breast cancer is one of the most common malignant tumors of women,and mammogram is one of the most effective method for the early diagnosis of breast cancer.By diagnosing breast cancer early,doctors can make suggestion for further examination to avoid the risk of cancer.Recent clinical studies have proved that computer-aided diagnosis(CAD)systems can help radiologists to detect breast lesions early and improve the diagnosis efficiency.However,it is still a great challenge to detect masses correctly because of the masses' vari-ability in shape,size and boundary.While conventional methods always rely on decision threshold and feature design,it is poor in generalization ability.With the development of deep learning in the field of computer vision,mammogram mass detection algorithms based on deep learning only use unilateral images and the false positive rate is still high.Clinical experiences by radiologists indicate that comparing bilateral breasts improves the detection accuracy of abnormalities in the breast.The fusion of information from bilateral breasts is expected to improve the performance of CAD systems.In this paper,breast mass detection algorithm based on fusion of bilateral feature is presented.First,we preprocess the image and register the regions of interest.Then,we use registration results as input of object de-tection network.After that,we extract and combine feature map of bilateral breasts with fusion network to integrate bilateral information.We exploit Mask R-CN:N as base framework to do object detection and segmentation simulta-neously.Experimental results on two public datasets show that the proposed algorithm based on fusion of bilateral features can improve the recall of mass detection effectively by 6.95%on DDSM and 4.79%on INbreast at 1 false positive per image compared to the conventional unilateral algorithm.