Automatic segmentation of ultrafast cardiac flow scans and RMBF calculation

Analysis of electrocardiogram (EKG) gated ultrafast computed tomography (CT) cardiac flow scans for myocardial perfusion has been limited by user traced regions of interest (ROI) corresponding to the left ventricular cavity and the myocardium. Normal intravenous contrast medium transport allows pixel classification heuristics to be defined and implemented in a PC based image workstation. Pixel classification was based upon three pixel parameters: the pre-contrast HU (Hounsfield Unit) value, the maximum HU change due to contrast medium, and the time of the maximal change. The classification algorithms operate on every fourth or second pixel of a 128×128 user placed sub-image of a 256×256 image. Regional myocardial and left ventricular time density curves (TDC) were generated from a series of twenty images from a mid-left ventricular ultrafast CT slice. TDC of the automatically detected regions were derived. Results from tests using this data indicate that automatic cardiac CT flow image segmentation and TDC generation are feasible and could aid clinical measurements of myocardial perfusion by ultrafast CT