Abstract W P34: Voxel-Based Mapping of C-arm CT Cerebral Blood Volume to Infarct Probability in a Canine Model of Ischemic Stroke
OBJECTIVE: Advances in C-arm cone-beam CT (CBCT) have opened up opportunities for imaging CBV during endovascular therapy. However, lack of quantitative and voxel-based evaluation of CBCT-CBV maps precludes widespread adoption in the angiography suite. The purpose of this study was therefore to determine the feasibility of voxel-based infarct probability mapping using CBCT-CBV.
METHODS: Adult beagle dogs (n=10) received stroke by unilateral ICA-injection of an autologous blood clot. Baseline and contrast-enhanced CBCT images were acquired 4.5h after stroke onset to calculate CBV maps, immediately followed by ADC-mapping at 3T. One animal was excluded because of acquisition problems.
CBV maps were normalized by the value in the vertebral arteries. After co-registration, a supervised learning approach modeled infarct probability as a function of CBV. A generalized linear model (GLM) discriminated ischemic lesion (ADC < 5.3 * 10^-4 mm^2 s^-1) from healthy brain tissue voxels. Only subjects with significantly reduced CBV in the lesion (vs. contralateral control area) were evaluated (n=7) using leave-one-out cross-validation.
RESULTS: The area under the ROC-curve for all predictions was 0.80, with a sensitivity and specificity of 73% and 75%, respectively, at the optimal operating point. The figure displays input and prediction images in one coronal slice for each animal, A-G. ADC maps [0-0.003] display a clearly defined lesion, whereas the normalized CBV maps [0-0.25] are more heterogeneous. Infarct probabilities [0-100%] were high in the lesion territories for subjects A-D and G. The discrepancy between ADC and CBV in subject E is likely due to a spontaneous complete recanalization around 4 hours post-stroke.
CONCLUSION: In conclusion, our results underscore the feasibility of voxel-level infarct probability mapping using angiographic CBV imaging.
Author Disclosures: K. van der Marel: None. J. Chueh: None. A.K. Wakhloo: Research Grant; Significant; Philips Healthcare. M.J. Gounis: Research Grant; Significant; Philips Healthcare.
- © 2015 by American Heart Association, Inc.