This article presents a method to explore the flexibility of channel reduction in k-domain parallel imaging (PI) with massive arrays to improve the computation efficiency. In PI, computation cost increases with the number of channels. For the k-domain methods requiring channel-by-channel reconstruction, this increase can be significant with massive arrays. In this article, a method for efficient k-domain PI reconstruction in large array systems is proposed. The method is based on the fact that in large arrays the channel sensitivity is localized, which allows channel reduction through channel cross correlation. The method is tested with simulated and in vivo MRI data from a 32-channel and 64-channel systems using the multicolumn multiline interpolation (MCMLI) method. Results show that the proposed algorithm can achieve similar or improved reconstruction quality with significantly reduced computation time for massive arrays with localized sensitivity.
