Figure 1

Visualization of the algorithmic workflow. A vessel segmentation is obtained by matching the histogram of the input image to the one of a reference image and equalizing the resulting histogram followed by a Frangi vesselness filter and a subsequent hysteresis thresholding operation. From this segmentation the vessel density can be directly computed as relative frequency. Moreover the intercapillary areas can be obtained by identifying the connected components. For the proposed method, a perfusion distance map is computed from which the perfusion distance values can be obtained. This perfusion distance map is subsequently binarized. From these binarized perfusion deficit areas, the geometric perfusion deficit percentage can be calculated. Moreover, the distinct perfusion deficit areas can be identified using connected components. Information about those regions can be summarized using features of the probability density function of their areas.