|
Spectral imaging is an active area of investigation for interventional applications. We have previously proposed a joint processing strategy that leverage both temporal and spectral information to obtain digital subtraction angiograms (DSA). This strategy allows for the isolation of iodine signal using measurements from just two energy channels and can yield a noise advantage compared to alternative spectral acquisitions. Previous investigations imposed fairly restrictive assumptions on patient motion (at least one material map is stationary pre- and post-contrast). In this work, we investigate the performance of the joint processing strategy without such assumptions. Specifically, we included a simple affine registration prior to material decomposition to mitigate gross patient motion. Assuming such registration can compensate motion in one tissue type (e.g., bone which exhibits locally rigid motion), we then use the total of four energy measurements to solve for four unknowns: water pre- and post-contrast, calcium, and iodine. This method accommodates registration mismatches in the pre- and post-contrast water images. To evaluate this method, we simulated a head phantom with a backward tilt about the lateral axis between the pre- and post-contrast acquisitions, as well as an abdomen phantom with cardiac, breathing, and a vertical translation. Results from the joint processing strategy was compared with that using temporal subtraction. Following four-material decomposition, both water images and calcium image show a small bias around 5% from ground truth. The iodine image contains bias primarily concentrated around bone edges. Compared to temporal subtraction, motion artifact from soft tissue is largely eliminated while that from bone is reduced in magnitude. As a result, improvement in motion artifact and iodine visibility is improved across the image. The improvement is especially obvious in the abdomen where soft tissue motion is dominant. We demonstrated in this work that the joint processing strategy is robust against motion artifact in the presence of gross patient motion and can present advantages compared to temporal subtraction, providing further support for the clinical translations of spectral DSA.
|
