A synthesis of 2D mammographic image using super-resolution technique: A phantom study

Abstract

The gold standard for early detection of breast cancer has been the mammogram. However, this technique still has limitation for women with dense breast. Combining mammogram with digital breast tomosynthesis overcomes the limitation but increases exposure dose approximately twice. This study focuses on reducing radiation dose by synthesizing the 2D mammographic image from multiple tomosynthesis projection images using an image Super-Resolution technique based on sparse representation. We evaluated the result images using peak signal to noise ratio (PSNR), mean structure similarity (MSSIM) and phantom passing score. We compared the synthesized 2D mammographic image from multiple projection images to the one from a single central projection image. The one from multiple images yields better result with. 27.2426 PSNR and 0.4436 MSSIM. For the phantom passing score, we obtained 5, 2, 4 for fibers, group of micocalcifications, and masses, respectively.