Session S81.2

Assessment of Myocardial Perfusion with Multi-Detector Computed Tomography

G Coppini*, R Favilla, B Barbagli, S Diciotti, S Lombardo, M Schlueter,
L Salvatori, C Canapini, D Neglia, P Marraccini

Institute of Clinical Physiology
Pisa, Italy

Current Multi-detector Computed Tomography (MDCT) systems allow the visualization of the entire heart volume at high spatial and density resolution. MDCT imaging is thus opened to a wide spectrum of cardiac applications. Particularly, the spatial distribution of contrast medium in the myocardium may provide valuable information about muscle perfusion and viability. In this view, we have developed a method to quantify the regional contrast enhancement as imaged during standard CT coronary angiography.
Our computational procedure includes the following steps: 1) Long axis reformatting of cardiac volume and interactive selection of ventricular volume; 2) Normalization of image gray levels by suppressing the myocardium baseline and scaling the HUs by the average contrast medium. Myocardium baseline is estimated in a slice acquired before contrast agent injection and average contrast medium density is measured at the coronary tree root; 3) Extraction of the LV cavity by gray level thresholding and blob growing; 4) LV segmentation using the Fast Marching Level Set algorithm: the evolving border originates from boundary of the LV cavity and propagates towards the outer boundary of the LV. The velocity map needed by eikonal equation is obtained from original CT values as transformed with a Gaussian function; 5) Processing of the segmented volumes to build polar maps similar to those adopted in PET and SPECT imaging and representing the spatial distribution of contrast medium in the myocardium.
The potential of the method was tested on a dataset including the scans from 10 patients with different perfusion defects and undergoing CT angiography (GE Discovery VCT CT/PET, with 64-slices CT). First of all, the quality of obtained segmentations was evaluated. In all the cases, expert observers confirmed that segmentations were anatomically correct with no artifact. In addition, manual segmentations by two observers was compared with the computed ones: sensitivity was about 95 ± 4%, with both false positives and negatives less than 4 ± 4%. Finally, in 5 infarcted patients PET Ammonia scans were also performed: polar maps by CT imaging demonstrated well-defined regional perfusion defects consistent with PET imaging results.

(Abstract Control Number: 275)