Vessel segmentation in DCE MR imaging

A computer program product stored on a non-transitory computer readable medium for automated vessel segmentation in a perfusion related map derived from perfusion MR images comprising:

• retrieving, for each pixel on the perfusion MR images, one or more curve parameters representing the degree to which the tissue in the pixel entirely consists of blood from perfusion curves;
• retrieving, for each pixel on the perfusion MR images, one or more curve parameters representing different measures of contrast arrival at the tissue in the pixel from perfusion curves;
• performing a k-class cluster analysis based on the retrieved curve parameters, where n≧3, comprising clustering the pixels into n cluster classes, the cluster analysis comprising determining:
• (i) cluster(s) whose pixels consist of blood to the highest degree;
• (ii) cluster(s) whose pixels have the earliest contrast arrival;
• and
• (iii) cluster(s) whose pixels have the latest contrast arrival;
• and
• generating a vessel mask based on pixels that are in both of clusters (i) and (ii) and pixels that are in both of clusters (i) and (iii).

A method for automated normalization of cerebral blood volume, CBV, maps based on dynamic susceptibility contrast, DSC, images, comprising:

• identifying and excluding noisy pixels with a mean baseline value below a predefined noise threshold in a pre-contrast DSC image;
• identifying and excluding tumor pixels;
• identifying, in all DSC slices and for all remaining pixels, first-pass curves thought to represent normal-appearing brain tissue pixels by excluding all confounding vessel-like and otherwise abnormal first-pass curves using one or more of the following criteria:
• identifying and excluding vessel pixels using a previously described DSC-based vessel segmentation technique;
• or
• identifying and excluding spur pixel values from first-pass curves with a peak value in the global baseline area;
• or
• identifying and excluding from first-pass curves erroneous pixels with evident T1-leakage effects or T2-leakage effects compared to a mean global first-pass curve.