SPECT-CT fusion in cardiac imaging: towards a specific sectorization of myocardial perfusion
Myocardial perfusion SPECT (MPS) and coronary CT angiography (CTCA) are both established techniques for the noninvasive evaluation of coronary artery disease. MPS provides a 3D representation of myocardial perfusion. CTCA provides anatomical information related to the arterial tree and detects the presence of coronary stenosis. The combination of functional and anatomical information give nuclear physicians anatomical landmarks and help them to isolate the stenosis responsible of the patient coronary artery disease.
Segmentation of coronary arteries is an important preliminary step for visualization of vessel anatomy with myocardial perfusion. We propose an algorithm for extraction of vessel walls without detection of the centerline based on mathematical morphology and region growing. It is a good compromise between accuracy and computational cost.
The second part of this work concerns the registration of MPS and CTCA. Our process is adapted to the anatomical and functional modalities.Mutual information was estimated more reliably on 3D points randomly selected from MPS by computing full rank bandwidth matrices for Parzen density calculations. The problem was solved using a stochastic optimization algorithm combining particle filtering and a quasi-Newton method.
MPS and coronary arteries are projected and fused onto a polar map to allow a combined analysis of stenosis location and myocardial perfusion defects. The novel proposed quantification integrates fused functional and anatomical information: each left ventricular segment is specifically assigned to a coronary artery with the lowest geodesic distance measurement over myocardial surface.
The proposal application is adapted for a routine clinical use in nuclear cardiology. As no gold standard exists for cardiac data registration, a reference database composed of brain MRI was used to evaluate the proposed framework. Then, it was evaluated on a cardiac phantom and clinical datasets. Coronary artery segmentation can be performed on artery diameters down to 1 mm. Registration between MPS and CTCA has proven to be precise with an accuracy in the order of MPS resolution. Patient-specific quantification of myocardial perfusion allows a better discrimination between coronary territories.
Keywords: multimodal registration, segmentation, perfusion quantification, myocardial perfusion territories, coronary arteries, myocardial perfusion SPECT, CT coronary angiography.