Cardiovascular disease remains the leading cause of death in the western world, placing an ever-increasing burden on both private and public health services. The electrocardiogram (ECG)-gated cardiac CT imaging is a promising non-invasive technique for early detection of fatty vulnerable plaque in coronary arteries. However, there are two major problems with the current technique: large patient radiation dose and insufficient temporal resolution (TR). Currently, the typical radiation dose is 10-15 mSv, which is 3-5 times as large as a standard chest CT scan. The current TR is merely 80-165 ms in contrast to the minimum requirement of 10-30 ms to observe the beating heart motion without motion artifact.

Current technique (Fig. 2.1) uses the ECG-signals to select projection data acquired in a time window that is placed within the “quiet” portion of the cardiac cycle (e.g., mid-diastole). Then, images are reconstructed by neglecting the cardiac motion within the time window resulting in blurring and artifacts in the reconstructed images. Also, this technique uses only 10-30% of the acquired data and throws away the rest of “off-phase” data, resulting in unnecessary radiation dose to the patient.

The long-term goal of this research is to develop the time resolved, low dose cardiac CT imaging (e.g., Fig. 2.2).

We will develop algorithms that estimate the time-dependent motion vector field of the heart from the measured data and integrate it into the image reconstruction process. We will develop motion estimation methods based on image-to-image matching (Pub. 2) or projection-to-projection matching. We will also develop motion compensation methods based on image processing schemes (Pub. 3) or a novel reconstruction process (Pub. 4).

In the final form, the motion will be estimated by maximizing the agreement between the acquired 4D projection data and the reconstructed time-resolved 4D images (Fig. 2.3).

The quality of the image will be significantly improved since the motion is compensated. In addition, lower tube current could be utilized since all of the acquired data will be used to reconstruct any cardiac phase of interest. We estimate the radiation dose to the patient will be reduced to 25-50% of the current level (Ref. 2).

We will then conduct the quantitative and qualitative evaluation of the performance of the new algorithms with various factors with patients and parameters used in the algorithms.

The proposed methods will not only solve the current problems of motion blur and excessive radiation dose, but also enable future cardiac applications (e.g., correlation between the motion, perfusion and stenosis) that are not possible with the current techniques.

Figure 2.4 shows estimated motion vector field; and Fig. 2.5 shows image-based compensation.
We recently derived a novel reconstruction formula which provides very good approximation to non-rigid deformation problem (Pub. 4). Figure 2.6 shows an example of images reconstructed by the proposed method.
 

Figure 2.4: Image based motion estimation from end-systole to mid-diastole separated by 400 ms.

Figure 2.5: Image based motion compensation. (Left) An image at end-systole (ES); (middle) an image at 200 ms from ES; (right) a pseudo image created with a estimated motion vector field and the image at ES (left) which would be obtained at 200 ms with superior temporal resolution.

Figure 2.6: (a) Projection data with motion; (b) a reconstructed image by Parker weighting with ramp filtering (Refs. 3-4); (c) a reconstructed image by the proposed DAxBPF algorithm (Pub. 4).