Abstract TP38: Improved Infarct Prediction After Acute Ischemic Stroke: A Priori Infarct Probability Defined By Site Of Vessel Occlusion Can Be Combined With Infarct Likelihood-ratios Based On Ct Perfusion Imaging
PURPOSE: CT-perfusion (CTP) is frequently used in acute stroke imaging, however, there is need for further standardization for reliable quantification of tissue at risk. Using a global threshold to predict infarct is not precise. Proximal branch occlusion, is an important predictor for outcome. The purpose was to present a holistic stroke imaging approach by combining CTP maps with a priori infarct probability defined by site of vessel occlusion to improve prediction of infarct.
METHOD AND MATERIALS: A novel method is presented to derive CTP based images displaying voxel-wise probability of infarction: 1) In 105 MCA occlusive stroke, voxel-wise sensitivity and specificity for infarction for incremental perfusion thresholds was measured with respect to tissue lesion maps at 48h. CT perfusion parameters in CTP images were converted to likelihood ratios defined by sensitivity/(1-specificity). Thus, the discriminative power for every perfusion value with respect to tissue outcome was directly encoded within the perfusion image. 2) A probabilistic a priori map in standard space based on 103 non-recanalized distal M1-MCA occlusive strokes at 48h after onset was generated. Lesion-masks were registered to standard MNI-152 space. 3) The pre-test probability map for M1-occlusive strokes was transformed to CTP space and multiplied by the CTP based likelihood ratio to obtain a post-test map displaying voxel-wise probability of infarction.
RESULTS: In 10 sample cases, infarct probability in admission imaging was displayed and compared to follow up imaging at 48h. Areas of >95% infarct probability correlated very well with actual tissue outcome. In contrast to original CTP maps, the resulting probability map exclusively displays voxels at risk of infarction. Noise was significantly reduced.
CONCLUSION: Infarct prediction can be improved combining CT perfusion maps with a priori infarct probability defined by site of vessel occlusion. Further studies are warranted.
- © 2012 by American Heart Association, Inc.