Super resolution convolutional neural networks for increasing spatial resolution of1H magnetic resonance spectroscopic imaging

Abstract

Proton magnetic resonance spectroscopic imaging (1H-MRSI) provides noninvasive information regarding metabolic activity within the tissues. One of the main problems of1H-MRSI is low spatial resolution due to clinical scan time limitations. Advanced post-processsing algorithms, like convolutional neural networks (CNN) might help with generation of super resolution1H-MRSI. In this study, the application of super resolution convolutional neural networks (SRCNN) for increasing the spatial resolution of1H-MRSI is presented. Fluid Attenuated Inversion Recovery (FLAIR), T1-weighted, T2-weighted magnetic resonance imaging (MRI) data and a fused MRI, which contained the three different structural MR images in each RGB channel, were used in training the SRCNN scheme. The spatial resolution of1H-MRSI images were increased by a factor of three using the models trained with the anatomical MR images. The results of the proposed technique were compared with bicubic resampling in terms of peak signal to noise ratio and root mean square error. Our results indicated that SRCNN would contribute to reconstructing higher resolution1H-MRSI.