A registration based approach for 4D cardiac micro-CT using combined prospective and retrospective gating

Abstract

Recent advances in murine cardiac studies with three-dimensional cone beam micro-computed tomography (CT) have used either prospective or retrospective gating technique. While prospective gating ensures the best image quality and the highest resolution, it involves longer sampling times and higher radiation dose. Sampling is faster and the radiation dose can be reduced with retrospective gating but the image quality is affected by the limited number of projections with an irregular angular distribution which complicate the reconstruction process, causing significant streaking artifacts. This work involves both prospective and retrospective gating in sampling. Deformable registration is used between a high quality image set acquired with prospective gating with the multiple data sets during the cardiac cycle obtained using retrospective gating. Tests were conducted on a four-dimensional (4D) cardiac mouse phantom and after optimization, the method was applied to in vivo cardiac micro-CT data. Results indicate that, by using our method, the sampling time can be reduced by a factor of 2.5 and the radiation dose can be reduced 35% compared to the prospective sampling while the image quality can be maintained. In conclusion, we proposed a novel solution to 4D cine cardiac micro-CT based on a combined prospective with retrospective gating in sampling and deformable registration post reconstruction that mixed the advantages of both strategies.

Figure 1

Cardio-respiratory gating can be performed with a prospective (a) or retrospective approach (c). Projection images are acquired during end-expiration. Physiologic data acquired via LabVIEW [(a), (c)] show ECG, ventilation, x-ray exposure, and pulse commands to rotate to the next angular setting. In case of prospective gating (a) the images are triggered by a predefined phase of the cardiac cycle, e.g., in this case the R peak of the QRS complex in the ECG signal. Prospective gating creates an uniform angular distribution of projections (b). In retrospective gating projection images are acquired in the same time point of the end-expiration periods but at various cardiac phases. The retrospective gating results in an irregular angular distribution of projection data (d). In (b) and (d) we used a representation of projections as radial lines in Fourier space that is based on the central slice theorem.

Figure 2

A flow chart of the proposed procedure combining prospective and retrospective gating in acquisition. The approach is based on the registration of a high quality data set acquired with the prospective gating over the lower quality data sets acquired with retrospective gating.

Figure 3

Simulations using the mouse phantom without Poisson noise added to projections. (a) We display a region of interest in an axial slice showing the heart during nine phases of cardiac cycle and reconstructed with data from the retrospective data sets. (b) The results after the deformable registration of the prospective data set over the retrospective sets. (c) The difference between the prospective and each phase retrospective before registration. Note the artifacts around the heart due to the difference in cardiac phase. (d) Difference between the deformed prospective and each phase retrospective. Note that the structured errors around the heart have been significantly reduced. (e) Maps showing the deformation field between the prospective and each phase of the retrospective sets.

Figure 4

(a) A surface representation of the errors (for each cardiac phase) and (b) a plot of the percentage of voxels vs the error distances between the deformed prospective data set and the true data sets for each time point. Note that the majority percentage of voxels on the surfaces is registered with an error that is less than 100 μm, i.e., the voxel size. (c) The RMS errors given in CT numbers and computed between the prospective set and each retrospective set before and after registration.

Figure 5

Simulations using the mouse phantom with Poisson noise added to the projections. (a) We display a region of interest in an axial slice showing the heart during nine phases of cardiac cycle and reconstructed with data from the retrospective data sets. (b) The results after the deformable registration of the prospective data set over the retrospective sets. (c) The error difference between the prospective and each phase retrospective before registration. Note the artifacts around the heart due to the difference in cardiac phase. (d) Difference between the deformed prospective and each phase retrospective. Note that the structured errors around the heart have been significantly reduced. (e) Maps showing the deformation field between the prospective and each phase of the retrospective sets.

Figure 6

(a) A surface representation associated with the noisy case of the errors (for each cardiac phase) and (b) a plot of the percentage of voxels vs the error distances between the deformed prospective data set to the true data sets for each time point. Note that now ∼4% of voxels on the surface have errors higher than 100 μm. (c) The RMS errors given in CT numbers and computed between the prospective set and each retrospective set before and after registration.

Figure 7

A live animal experiment. One prospective image set (a) is deformed over a lower quality retrospective data sets (b) at a different cardiac phase. The result is shown by (c). The error between the prospective and retrospective data sets before registration (d) and the error between the deformed prospective and the retrospective after registration (e). The image shows results for both systole and diastole. The arrows indicate some errors that appear in the heart regions before the deformation. Note how these have been significantly reduced after registration. (f) The RMS error given in CT numbers and computed between the prospective set and each retrospective set before and after registration.