(An) efficient echo time encoding technique for metal artifact correction in MRI

Abstract

When imaging near metal implants in magnetic resonance imaging (MRI), the large difference of susceptibility between human tissue and metal implants causes the local field inhomogeneity. These field inhomogeneity results in severe artifacts such as bulk displacement and through-plane distortion. Many sequences such as SR-FPE and MAVRIC-SL have been developed to correct metal artifacts, but their drawback is that the imaging time takes too long. It is inconvenient for patients to take images without moving for a long time in a closed space. Therefore, the sequence that reduces the imaging time with minimal metal artifact is needed. In this paper, the metal artifact is minimized by using the echo time encoding technique (ETE) in 3D turbo spin sequences without slab selectivity gradient. Also, we propose a reconstruction method to reduce the imaging time through the machine learning. To reduce the imaging time, only the limited number of ETE steps is obtained, and the remaining ETE steps are reconstructed by machine learning through the measured ETE steps and all of the ETE images are combined to get the corrected image. For various sampling patterns, we compared and analyzed the reconstructed results from the regression and deep learning.