Separation of perfusion signals from diffusion-weighted image series enabled by independent component analysis.

BACKGROUND AND PURPOSE An important task in diagnostic imaging of acute ischemic stroke is to identify the so-called diffusion-perfusion mismatch area. We aimed to investigate the possibility of facilitating the identification process by combining independent component analysis (ICA) and diffusion-weighted MRI (DWI), with the expectation that this would eliminate the need for additional perfusion imaging to delineate perfusion lesion. METHODS Simulations were performed to confirm the utility of an intuitively determined sequence of 14 b-factors ranging from 0 to 1,000 seconds/mm2 for ICA separation of perfusion lesion. Corresponding DWI data from 2 stroke patients, 1 in the acute and 1 in the subacute phase, were decomposed into independent component (IC) maps, and their b-dependent amplitude decay profiles were subjected to multiexponential fitting. RESULTS Low-perfusion areas were successfully delineated on IC maps in both patients. Comparison with the areas of diffusion lesion identifiable on relatively high b-factor images in the DWI data, for example, those at b= 1,000 seconds/mm2, allowed the mismatch to be identified. CONCLUSION This study demonstrates that combining ICA and DWI enables noninvasive mapping of sluggish perfusion provided an appropriate b-sequence is applied, and that it thereby facilitates the identification of diffusion-perfusion mismatch.