Separation of arteries and veins from optical imaging of intrinsic signals using independent component analysis

In this paper, a classical blind source separation (BSS) method is applied to optical imaging of intrinsic signals for separation of arteries and veins in the cerebral cortex. The separation method is based on the fact that the physiological sources are different between arterial regions and venous regions. After segmenting the vascular network, independent component analysis (ICA) was applied to the vascular images for separating the mixed sources into independent components. We selected heartbeat corresponding time course, then founded coherence with every pixel time course at heartbeat frequency. The resulting value constructed a new heartbeat corresponding feature map. A new respiration corresponding feature map was obtained in the same way. Two feature maps were combined as feature vector for the fuzzy c-means clustering method. Determination of vessel type was conducted on each vessel segments divided by morphological intersection points. Finally, it was able to separate arteries and veins by optical gray image of intrinsic signal.