Purpose: To develop and validate a dual-contrast image subtraction (DCIS) strategy for eliminating the flow artifacts in black-blood carotid MRI. Methods: Twelve patients with carotid stenosis and eight healthy volunteers were imaged using the black and grayblood dual-contrast imaging based on the relaxationenhanced compressed sensing three-dimensional motion-sensitizing driven equilibrium prepared rapid-gradient-echo (RECS-3D MERGE) sequence. Subtraction of black-blood images (BBIs) and gray-blood images (GBIs), together with a preweighting procedure, was performed to eliminate the residual blood signal in BBIs. A wavelet denoising procedure was applied to offset the noise amplification. In addition to the lumen signal-to-noise ratio (SNR) and wall-lumen contrast-tonoise ratio (CNR), the signal variance ratio (SVR) and contrast variance ratio (CVR) were also used to evaluate the blood suppression efficiency. Results: By choosing the weighting factor of one, the lumen SNR of DCIS images was approximately 1% of that of the original BBIs, and the CNR showed a 91.4% improvement as compared with the BBIs. The median of the lumen SVR decreased to zero, and the CVR increased to 123% of that of the BBIs. Conclusions: DCIS is demonstrated to be an effective strategy for sufficiently removing the residual flow signal from black-blood carotid MRI. (C) 2016 International Society for Magnetic Resonance in Medicine