We aim to investigate whether dynamic perfusion image time series instead of static perfusion image can offer extra insight to bipolar disorder (BD). Average perfusion and perfusion fluctuation maps were compared between patients with BD and controls using customized Statistical non-Parametric Mapping (SnPM). Perfusion decrease in the posterior lateral regions of the default mode network and increase of perfusion fluctuations in the parahippocampus and amygdala regions were observed. The abnormal perfusion fluctuations may be supported by perturbed functional connectivity at the same regions. These results indicate that dynamic perfusion image series may serve as potential novel neuroimaging biomarkers for BD.