This poster is published under an
open license. Please read the
disclaimer for further details.
Keywords:
Ischaemia / Infarction, Imaging sequences, Contrast agent-intravenous, CT-Quantitative, CT, Interventional vascular, Computer applications
Authors:
G. Van Eyndhoven1, J. Sijbers2, J. Batenburg3; 1Wilrijk/BE, 2Antwerp (Wilrijk)/BE, 3Amsterdam/NL
DOI:
10.1594/ecr2014/C-0282
Results
Projection data was simulated with a realistic brain perfusion phantom [5],
which enables accurate validation and reproducibility.
A region of moderate and severely reduced perfusion was indicated on the phantom,
which manifests itself by a reduced blood flow and volume in the ground truth CBV and CBF maps as shown in Fig.
6 and Fig.
7,
respectively.
Poisson distributed noise was applied to the projection data.
For every experiment,
the CBF and CBV were calculated with the truncated singular value decomposition approach [2].
The results were validated with the root mean squared error (RMSE) with respect to the ground truth TCCs,
CBF map and CBV map.
The results are displayed in function of the number of projections per 180° rotation (i.e.,
per reconstructed time point) in Fig.
3,
Fig.
4 and Fig.
5,
respectively.
The RMSE plots (Fig.
3-5) reveal that for the same dose,
our method significantly improves the quality of TCCs,
CBV maps and CBF maps.
In terms of RMSE of the TCCs,
our method maintains image quality while reducing radiation exposure to 32% in comparison to SIRT,
i.e.,
RMSE of TCCs for SIRT at 370 projections per time point is equal to RMSE of TCCs for our approach at 120 projections per time point.
Fig.
6-7 reveal that the overall structure of CBF and CBV map are better recovered with our approach in comparison to SIRT.