KIKI‐net: cross‐domain convolutional neural networks for reconstructing undersampled magnetic resonance images

• Published in: Magnetic resonance in medicine Vol. 80; no. 5; pp. 2188 - 2201
• Main Authors: Eo, Taejoon, Jun, Yohan, Kim, Taeseong, Jang, Jinseong, Lee, Ho‐Joon, Hwang, Dosik
• Format: Journal Article
• Language: English
• Published: United States Wiley Subscription Services, Inc 01.11.2018
• Subjects: Algorithms; Aliasing; Alzheimer's disease; Artificial neural networks; Consistency; convolutional neural networks; cross‐domain deep learning; Data acquisition; Data recovery; Datasets; Fourier transforms; Image processing; Image reconstruction; k‐space completion; Magnetic resonance imaging; Medical imaging; MRI acceleration; Neural networks; Neurodegenerative diseases; Neuroimaging; Neurology; Performance evaluation; Restoration; cross-domain deep learning; k-space completion; convolutional neural networks; image reconstruction; MRI acceleration
• ISSN: 0740-3194; 1522-2594; 1522-2594
• DOI: 10.1002/mrm.27201

Purpose To demonstrate accurate MR image reconstruction from undersampled k‐space data using cross‐domain convolutional neural networks (CNNs) Methods Cross‐domain CNNs consist of 3 components: (1) a deep CNN operating on the k‐space (KCNN), (2) a deep CNN operating on an image domain (ICNN), and (3) an interleaved data consistency operations. These components are alternately applied, and each CNN is trained to minimize the loss between the reconstructed and corresponding fully sampled k‐spaces. The final reconstructed image is obtained by forward‐propagating the undersampled k‐space data through the entire network. Results Performances of K‐net (KCNN with inverse Fourier transform), I‐net (ICNN with interleaved data consistency), and various combinations of the 2 different networks were tested. The test results indicated that K‐net and I‐net have different advantages/disadvantages in terms of tissue‐structure restoration. Consequently, the combination of K‐net and I‐net is superior to single‐domain CNNs. Three MR data sets, the T2 fluid‐attenuated inversion recovery (T2 FLAIR) set from the Alzheimer's Disease Neuroimaging Initiative and 2 data sets acquired at our local institute (T2 FLAIR and T1 weighted), were used to evaluate the performance of 7 conventional reconstruction algorithms and the proposed cross‐domain CNNs, which hereafter is referred to as KIKI‐net. KIKI‐net outperforms conventional algorithms with mean improvements of 2.29 dB in peak SNR and 0.031 in structure similarity. Conclusion KIKI‐net exhibits superior performance over state‐of‐the‐art conventional algorithms in terms of restoring tissue structures and removing aliasing artifacts. The results demonstrate that KIKI‐net is applicable up to a reduction factor of 3 to 4 based on variable‐density Cartesian undersampling.