A spin labeling method to measure cerebral blood flow without a contrast medium was developed and applied clinically to obtain a non-invasive perfusion-weighted image. The purpose of this study is to compare the non-invasive perfusion-weighted image using FAIR with the well-established PWI using a bolus injection of Gd-DTPA. Of 41 lesions which revealed decreased perfusion, 13 were shown to be low signal intensity areas on FAIR. Therefore, detection rate of FAIR for hypoperfusion was 32%.0 Of 8 lesions which revealed increased perfusion, 7 demonstrated high intensity on FAIR. Therefore, detection rate of FAIR for hyperperfusion was 88%.0 7 lesions were found to have a mean pixel value of zero on PWI. Of these lesions, 5 lesions could be detected as high signal intensity area on FAIR. The rCBV- and rCBF index ratios of hypoperfused lesions detected on FAIR were significantly lower than those of lesions which were not detected on FAIR (p=0.007,p=0.01). As concerns the lesions detected of FAIR, there were positive correlation between rCBV- or rCBF index ratio and FAIR signal ratio (rCBV ratio : ρ=0.873,p=0.0002,rCBF index ratio : ρ=0.858,p=0.0003). FAIR is valuable clinical tool to detect perfusion abnormality semi-quantitatively without contrast medium, although it showed relatively low detection rate for hypoperfused lesions.