Parallel Imaging for Efficient Spike Noise Detection and Correction

Introduction Spike noise is a term used to describe broadband electrical interference in an MRI system. The result of spike noise can be seen in k-space as a bright dot. After the image reconstruction processing, the spike noise produces an undesired corduroy or wavy pattern over the magnetic resonance image. Typical sources of spike noise include unclosed door, a failed receiver channel, electrostatic discharge (ESD), and metal-on-metal vibration caused by loose mechanical fasteners or metal debris. Usually, the corrupted data has to be abandoned and the scan needs be repeated. Several methods have been proposed to remove spike noises using thresholds [1] or image space post processing [2] to avoid the repeated scan. As discussed in Ref [3], however, it is difficult to apply these techniques for some applications. In this work, a partially parallel imaging based technique, COnvolution and Combination OperAtion [4], is applied to efficiently detect and correct the spikes.