TU-H-CAMPUS-IeP2-02: Four-Dimensional Coronary Artery Visualization Based On Image Propagation




4D visualization of coronary artery using X-ray angiograms rotationally acquired on C-arm has many valuable applications, like dynamic roadmapping. In this work we propose and validate a method to visualize 4D artery based on 3D coronary image propagation.


After sorting the angiograms into different cardiac phases using nearest-neighbor ECG gating, a high quality 3D coronary image was reconstructed at one phase using a motion-compensated iterative reconstruction method. Then the reconstructed image was propagated to other cardiac phases by registering it to the angiograms of each phase (stage I registration). A third-order B-spline was used to model the vessel motion and the parameters were solved by maximizing the normalized correlation (NC) between the projections of the 3D image and the angiograms under the constraints of motion smoothness and volume preserving. Because the angiograms corresponding to one phase contain residual vessel motion and the registration method could only match the mean vessel morphology, morphological mismatch still existed between the propagated image and each angiogram. To obtain accurate vessel morphology according to each angiogram, we further registered the propagated image to each angiogram using the aforementioned registration method (stage II registration).


The proposed method was evaluated on angiograms simulated using 4D XCAT phantom. The reconstructed image at end-diastolic phase (phase 0%) was consecutively propagated to end-systolic phase (phase 40%) every 5%. After stage I registration, the mean distance between the propagated and ground truth vessel centerlines was decreased from 2.2±1.2 to 0.6±0.2 mm. The mean centerline distance was further decreased to 0.5±0.2 mm after the stage II registration.


A 4D coronary artery visualization method based on image propagation was proposed. The experimental results show that accurate 4D morphology of coronary artery could be obtained using this method.

This work was supported by the National Natural Science Foundation of China (61171005).