2018/04/12 by Şahin Olut, Yusuf H. Şahin, Olut, Sahin +5 · 2 citations
Computer Science · #Advanced Image Processing Techniques #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Image and Signal Denoising Methods
paper · pdf · doi:10.48550/arxiv.1804.04366
openalex publication_date 2018/04/12 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Magnetic Resonance Angiography (MRA) has become an essential MR contrast for\nimaging and evaluation of vascular anatomy and related diseases. MRA\nacquisitions are typically ordered for vascular interventions, whereas in\ntypical scenarios, MRA sequences can be absent in the patient scans. This\nmotivates the need for a technique that generates inexistent MRA from existing\nMR multi-contrast, which could be a valuable tool in retrospective subject\nevaluations and imaging studies. In this paper, we present a generative\nadversarial network (GAN) based technique to generate MRA from T1-weighted and\nT2-weighted MRI images, for the first time to our knowledge. To better model\nthe representation of vessels which the MRA inherently highlights, we design a\nloss term dedicated to a faithful reproduction of vascularities. To that end,\nwe incorporate steerable filter responses of the generated and reference images\ninside a Huber function loss term. Extending the well- established\ngenerator-discriminator architecture based on the recent PatchGAN model with\nthe addition of steerable filter loss, the proposed steerable GAN (sGAN) method\nis evaluated on the large public database IXI. Experimental results show that\nthe sGAN outperforms the baseline GAN method in terms of an overlap score with\nsimilar PSNR values, while it leads to improved visual perceptual quality.\n