Supervised classification of cerebral blood vessels

Abstract

The x-ray angiograms are frequently performed before a cerebral intervention on diseased vessels. The clinicians complain about the difficulty to find small vessels on images with grey level intensities and to differentiate veins from arteries. The main problem in classification resides in finding the right parameters which can completely characterize the patterns of pixels' intensity time course from x-ray projection image series. A basic signal processing was made: a low pass filter for eliminating the undesired information, as long as the application of fast Fourier transform for investigation of signal spectral characteristics. In the presented article, a classification method of blood vessels from the cerebral angiograms based on temporal signals is presented, with a successful rate of identification of arteries of 78% and veins of 65%.