Abstract T P172: Temporal Correlation Perfusion Mapping Introduces a Robust and Rapid Method of Identifying Penumbra in Acute Ischemic Stroke
Background and Purpose: We implemented new temporal correlation perfusion (TCP) analyses that were rapid, model-free and automated. We assessed the reliability and robustness of TCP mapping compared to the standard method of time-to-peak (TTP) mapping used in the identification of perfusion deficit in acute ischemic stroke.
Methods: A model-free and automated method of detecting aberrant perfusion was developed to create TCP maps based on the maximum temporal correlations of the signal time-intensity waveforms following contrast injection. For validation purposes, the TCP and TTP maps from IV tPA treated ischemic stroke (n=20) and untreated imaging negative TIA (n=20) patients were visually read for presence of perfusion deficit by two expert readers blinded to the clinical diagnosis. Next, the perfusion lesion volumes on the TCP and TTP maps were segmented by two other experts and were compared for inter-rater reliability, volumetric and spatial concordance, and contrast to noise ratio. Pearson’s r correlation and t-test for TCP and TTP maps were used with degrees of freedom indicated by parentheses. Additionally, inter-rater reliability and contrast to noise ratios were compared.
Results: Both readers identified all of the subjects (40/40) correctly as either having perfusion deficits or not based on the TCP maps alone. Improved rater inter-reliability in lesion volume segmentation was found in TCP compared to TTP maps, with high correlation in lesion volume (r(18) = 0.85) and spatial overlap (91.1+/-9.5%) between TCP and TTP maps. Better contrast to noise ratio in lesions segmented with TCP compared to TTP maps (t(19) = 14.9, p<0.0001) was demonstrated which led to increased lesion conspicuity.
Conclusions: We conclude that temporal correlation mapping holds promise as a reliable and more robust method for perfusion deficit identification as demonstrated by correct identification and improved lesion segmentation compared to current standard methods.
Author Disclosures: S. Song: None. M. Luby: None. S. Shah: None. M. Edwardson: None. T. Brown: None. L.L. Latour: None.
- © 2015 by American Heart Association, Inc.