Super-resolution MRI through Deep Learning

摘要: Magnetic resonance imaging (MRI) is extensively used for diagnosis and image-guided therapeutics. Due to hardware, physical physiological limitations, acquisition of high-resolution MRI data takes long scan time at high system cost, could be limited low spatial coverage also subject motion artifacts. Super-resolution can achieved with deep learning, which a promising approach has great potential preclinical clinical imaging. Compared polynomial interpolation or sparse-coding algorithms, learning extracts prior knowledge from big produces superior images low-resolution counterpart. In this paper, we adapt two state-of-the-art neural network models CT denoising deblurring, transfer them super-resolution MRI, demonstrate encouraging results toward two-fold resolution enhancement.