Edge Preserved Herringbone Artifact Removal from MRI Using Two-Stage Variational Mode Decomposition

Magnetic Resonance Imaging (MRI) is an efficient and non-invasive method for analyzing the structural features and functional behaviors of internal organs and tissues for medical diagnosis. The artifacts present in MRI mislead the diagnostic procedure. Herringbone artifact is a hardware artifact generated from the outlier in k-space measurement. In realtime MRI, the herringbone artifact has non-stationary noise characteristics. The non-stationary noise characteristics affect the high-frequency characteristics which in turn results in an improper estimation of structural details of the image in the processing stage. The objective of the present work is to exploit the properties of the variational mode decomposition (VMD) in reducing the effective herringbone noise at selected spectral regions (high-frequency regions in particular) of the given MRI data. In the present work, the given herringbone artifact affected image is subjected to VMD in two stages. The reconstructed image by removing the higher frequency VMD modes in two-stages found to enhance the noisy MRI data. In the second stage of processing, the discarded higher frequency VMD mode in the first stage is further decomposed into component modes in order to preserve the high-frequency details. The enhanced image is later reconstructed by adding low-frequency modes obtained in both decomposition stages. The effectiveness of the proposed two-stage VMD based enhancement is confirmed from the improved scores obtained from the non-reference quality measures such as Blind Referenceless Image Spatial Quality Evaluator (BRISQUE) and Naturalness Image Quality Evaluator (NIQE).