Image Quality Enhancement of Digital Breast Tomosynthesis Images by Deblurring with Deep Residual Convolutional Neural Network

Abstract

Digital Breast Tomosynthesis (DBT) is widely used for early diagnosis of breast cancer due to its higher detectability compared to mammography. However, since the DBT acquires projection data within a limited angular range, blurring artifact occurs in reconstructed images. In this work, we proposed a method to reduce the blurring artifact in DBT images by applying Deep Residual Convolutional Neural Network (DRCNN), which was shown to be very effective for deblurring caused by camera motion. To evaluate the performance of the proposed method, we calculated Mean Squared Error (MSE), Peak Signal to Noise Ratio (PSNR) and structural similarity index (SSIM) between reference and output image from the proposed network. The results show that the DRCNN reduces the blurring artifact effectively, producing the improved image quality.