Reconstruction of 3D Cardiac MR Images from 2D Slices Using Directional Total Variation

Cardiac MRI allows for the acquisition of high resolution images of the heart. Long acquisition times of MRI make it impractical to image the full heart in 3D at high resolution. As a result, multiple 2D images are commonly acquired with a slice thickness greater than the in-plane resolution. One way of achieving isotropic high-resolution images is to apply post-processing techniques such as super-resolution to produce high resolution images from low resolution input. We use short-axis stacks as well as orthogonal long-axis views in a super-resolution framework, constraining the reconstruction using the contrast independent directional total variation algorithm to produce a high resolution 3D reconstruction with isotropic resolution. The 3D reconstruction retains the contrast of the short-axis stack, but incorporates the edge information from both the short-axis and the long-axis stacks. Results show improved reconstructions, with a segmentation voxel misclassification rate of 3.51% as opposed to 4.27% using linear interpolation.

[1]  D. Donoho,et al.  Sparse MRI: The application of compressed sensing for rapid MR imaging , 2007, Magnetic resonance in medicine.

[2]  Arthur W. Toga,et al.  MRI resolution enhancement using total variation regularization , 2009, 2009 IEEE International Symposium on Biomedical Imaging: From Nano to Macro.

[3]  Jan Sijbers,et al.  General and Efficient Super-Resolution Method for Multi-slice MRI , 2010, MICCAI.

[4]  Konstantinos Kamnitsas,et al.  Anatomically Constrained Neural Networks (ACNNs): Application to Cardiac Image Enhancement and Segmentation , 2017, IEEE Transactions on Medical Imaging.

[5]  J. Shewchuk.  An Introduction to the Conjugate Gradient Method Without the Agonizing Pain , 1994 .

[6]  Marta M. Betcke,et al.  Multicontrast MRI Reconstruction with Structure-Guided Total Variation , 2015, SIAM J. Imaging Sci..

[7]  Konstantinos Kamnitsas,et al.  Multi-input Cardiac Image Super-Resolution Using Convolutional Neural Networks , 2016, MICCAI.

[8]  Jerry L. Prince,et al.  Reconstruction of High-Resolution Tongue Volumes From MRI , 2012, IEEE Transactions on Biomedical Engineering.

[9]  Eric Van Reeth,et al.  Super-resolution in magnetic resonance imaging: A review , 2012 .

[10]  Wiro J Niessen,et al.  Super‐resolution methods in MRI: Can they improve the trade‐off between resolution, signal‐to‐noise ratio, and acquisition time? , 2012, Magnetic resonance in medicine.

[11]  Charl P. Botha,et al.  Super-resolution reconstruction of whole-body MRI mouse data: An interactive approach , 2012, 2012 9th IEEE International Symposium on Biomedical Imaging (ISBI).

[12]  Xiaoou Tang,et al.  Learning a Deep Convolutional Network for Image Super-Resolution , 2014, ECCV.

[13]  L. Rudin,et al.  Nonlinear total variation based noise removal algorithms , 1992 .

[14]  Daniel Rueckert,et al.  Dictionary Learning and Time Sparsity for Dynamic MR Data Reconstruction , 2014, IEEE Transactions on Medical Imaging.

[15]  Freddy Odille,et al.  Motion-Corrected, Super-Resolution Reconstruction for High-Resolution 3D Cardiac Cine MRI , 2015, MICCAI.

[16]  Hayit Greenspan,et al.  MRI inter-slice reconstruction using super-resolution , 2002 .