Iterative image reconstruction for cerebral perfusion CT using a pre-contrast scan induced edge-preserving prior

Abstract

Cerebral perfusion x-ray computed tomography (PCT) imaging, which detects and characterizes the ischemic penumbra, and assesses blood-brain barrier permeability with acute stroke or chronic cerebrovascular diseases, has been developed extensively over the past decades. However, due to its sequential scan protocol, the associated radiation dose has raised significant concerns to patients. Therefore, in this study we developed an iterative image reconstruction algorithm based on the maximum a posterior (MAP) principle to yield a clinically acceptable cerebral PCT image with lower milliampere-seconds (mA s). To preserve the edges of the reconstructed image, an edge-preserving prior was designed using a normal-dose pre-contrast unenhanced scan. For simplicity, the present algorithm was termed as 'MAP-ndiNLM'. Evaluations with the digital phantom and the simulated low-dose clinical brain PCT datasets clearly demonstrate that the MAP-ndiNLM method can achieve more significant gains than the existing FBP and MAP-Huber algorithms with better image noise reduction, low-contrast object detection and resolution preservation. More importantly, the MAP-ndiNLM method can yield more accurate kinetic enhanced details and diagnostic hemodynamic parameter maps than the MAP-Huber method.

Figure 1

Three Shepp-Logan phantoms used in the studies. (a) the pre-contrast unenhanced phantom; (b) the enhanced phantom wherein the large bright region represents the enhanced tissue; and (c) the enhanced phantom with a low-contrast lesion indicated by an arrow. The display window option: width is 220 HU, level is 670 HU.

Figure 2

Shepp-Logan phantom reconstructions by different methods. (a) the noise-free phantom image; (b) the image reconstructed by the FBP method from the noisy sinogram data; (c) the image reconstructed by the MAP-Huber method from the noisy sinogram data (β = 8.0×10³); and (d) the image reconstructed by the MAP-ndiNLM method from the noisy sinogram data (h =40.0; β = 1.5×10⁻³). The display window option: width is 220 HU, level is 670 HU.

Figure 3

Vertical profiles located at the pixel positions x =103 and y from 60 to 180 of images in figure 2. The ‘blue line’ is image reconstructed by the FBP method while the ‘red line’ is from the image reconstructed by the MAP-ndiNLM and MAP-Huber methods, respectively, and the “dotted black line” is from the noise-free phantom which acts as a ground-truth.

Figure 4

The noise-resolution tradeoff curves of the FBP, MAP-Huber and MAP-ndiNLM methods. The resolution was measured by the FWHM in pixel unit.

Figure 5

The ROC curves of the FBP, MAP-Huber, and MAP-ndiNLM methods.

Figure 6

Cerebral PCT image reconstructions by different methods from the simulated low-dose singoram data. (a) the pre-contrast unenhanced image reconstructed by the FBP method from the pre-contrast normal-dose scan, which acts as the reference image; (b) the image reconstructed by the FBP method from the normal-dose scan; (c) the image reconstructed by the FBP method from the simulated low-dose sinogram data; (d) the image reconstructed by the MAP-Huber method from the simulated low-dose sinogram data (β = 1.0×10³); and (e) the image reconstructed by the MAP-ndiNLM method from the simulated low-dose sinogram data (h =80.0, β = 5.0×10⁻²). The display window option: width is 160 HU, level is 56 HU.

Figure 7

Horizontal profiles through the center of the images in figure 6. The “blue line” is from the FBP reconstruction while the “red line” is from the reconstructions with the MAP-ndiNLM and MAP-Huber methods, and the “dotted black line” is from the normal-dose image which acts as the ground-truth for comparison.

Figure 8

TDC accuracy of the AIF, VOF and tissue perfusion of dynamic images reconstructed from the noisy sinogram data. (a) TDCs of the AIF (the 3×3 ROI indicated by a red square in Figure 6(b)); (b) TDCs of tissue 1 (the 3×3 ROI indicated by a magenta square in figure 6(b)); (c) TDCs of tissue 2 (the 3×3 ROI indicated by a yellow square in figure 6(b)); and (d) TDCs of the VOF(the 3×3 ROI indicated by a blue square in Figure 6(b)).

Figure 9

The MTT (column one), CBF (column two), and CBV (column three) maps calculated from the different brain PCT images. The first row was calculated from the normal-dose images; the second, third and fourth rows were calculated from the simulated low-dose images reconstructed by the FBP, MAP-Huber, and MAP-ndiNLM methods, respectively. The radiation dose in the low-dose sinogram data simulation is about one-seventh of the normal dose.

Figure 10

Zoomed ROIs of the MTT (column one), CBF (column two), and CBV (column three) maps in figure 9. The first row was calculated from the normal-dose images; the second, third and fourth rows were calculated from the simulated low-dose images reconstructed by the FBP, MAP-Huber, and MAP-ndiNLM methods, respectively. The radiation dose in the low-dose sinogram data simulation is about one-seventh of the normal dose.

Figure 11

The correlation (left column) and Bland–Altman plot (right column) between the MTT computed from the normal-dose images and the low-dose images reconstructed by different methods. Plots (a) and (b) represent the results obtained from the normal- and low-dose FBP reconstructions. Plots (c) and (d) represent the corresponding results obtained from the normal-dose images and the low-dose MAP-Huber reconstructions. Plots (e) and (f) represent the corresponding results obtained from the normal-dose images and the low-dose MAP-ndiNLM reconstructions.

Figure 12

The correlation (left column) and Bland–Altman plot (right column) between the CBV values computed from the normal-dose images and the low-dose images reconstructed by different methods. Plots (a) and (b) represent the results obtained from the normal- and low-dose FBP reconstructions. Plots (c) and (d) represent the corresponding results obtained from the normal-dose images and the low-dose MAP-Huber reconstructions. Plots (e) and (f) represent the corresponding results obtained from the normal-dose images and the low-dose MAP-ndiNLM reconstructions.

Figure 13

The correlation (left column) and Bland–Altman plot (right column) between the CBF values computed from the normal-dose images and the low-dose images reconstructed by different methods. Plots (a) and (b) represent the results obtained from the normal- and low-dose FBP reconstructions. Plots (c) and (d) represent the corresponding results obtained from the normal-dose images and the low-dose MAP-Huber reconstructions. Plots (e) and (f) represent the corresponding results obtained from the normal-dose images and the low-dose MAP-ndiNLM reconstructions.