Help Privacy Policy Disclaimer
  Advanced SearchBrowse


  Independent component analysis of resting-state hemodynamics in acute stroke: A new approach for identifying hypoperfusion

Khalil, A., Kirilina, E., Nierhaus, T., Villringer, K., Villringer, A., & Fiebach, J. (2016). Independent component analysis of resting-state hemodynamics in acute stroke: A new approach for identifying hypoperfusion. Poster presented at ESMRMB 2016 Congress, Vienna, Austria.

Item is


show Files




Khalil, Ahmed1, 2, Author
Kirilina, Evgeniya3, Author              
Nierhaus, Till2, Author              
Villringer, Kersten1, Author
Villringer, Arno2, Author              
Fiebach, Johann1, Author
1Center for Stroke Research Berlin (CSB) Charité - Universitätsmedizin Berlin, ou_persistent22              
2Department Neurology, MPI for Human Cognitive and Brain Sciences, Max Planck Society, ou_634549              
3Department Neurophysics (Weiskopf), MPI for Human Cognitive and Brain Sciences, Max Planck Society, ou_2205649              


Free keywords: -
 Abstract: Purpose/Introduction: Hemodynamically compromised tissue following ischemic stroke produces changes in the BOLD signal that include localized temporal delays [1] and alterations in signal frequency and amplitude [2]. We propose a new method to localize hypoperfused tissue based on a combination of spatial independent component analysis (sICA) and temporal correlation analysis of resting-state functional MRI (rsfMRI) data and compared it to an established method of perfusion analysis. Subjects and methods: Five patients were scanned within 24 hours of stroke symptom onset with rsfMRI (TR = 2.3 s, TE = 30 ms, 150 volumes) in addition to a standard stroke protocol. Diffusion weighted imaging (DWI) was used to localize the acute infarct and gadolinium-based dynamic susceptibility contrast MRI (DSC-MRI, TR = 1.39 s, TE = 29 ms) was used to identify hypoperfused tissue. Spatial ICA was performed on the rsfMRI data using FSL’s MELODIC (fsl.fmrib.ox.ac.uk/fsl/fslwiki/MELODIC) after volume realignment, spatial smoothing, and slice time correction (no temporal filtering was applied). Spatial correlation between independent components (ICs) and DSC-based maps of time-to-maximum bolus delay (Tmax) was calculated. The time courses of the ICs were cross-covaried with the whole brain reference signal and the mean Tmax delay in the areas covered by the ICs was also calculated. Results: For each patient, we identified ICs corresponding to the area of hypoperfusion and ICs predominantly located in the same vascular territory on the healthy hemisphere (see Figure 1), the latter being used as a control. Mean Tmax delay in the areas covered by the hypoperfusion ICs ranged from 3.23 to 4.91 s (median = 4.62 s) compared to 1.74 - 2.51 s (median = 2.42 s) for the healthy hemisphere ICs. Spatial correlation between hypoperfusion ICs and Tmax (>2 s) lesions ranged from 0.2 to 0.6 (median = 0.51). Delay between the ICs’ time courses and that of the whole brain (defined as the time shift value at maximum cross-covariance; negative values indicate delay) ranged from -9.7 to -5.6 s (median = -7.9 s) for hypoperfusion ICs and -4.8 to 0.4 s (median = 0.2 s) for healthy hemisphere ICs (see Figure 2). Discussion/conclusion We demonstrate that maps with a similar spatial distribution as hypoperfused tissue and temporal features suggestive of hemodynamic compromise can be extracted using sICA of rsfMRI data. In addition, these ICs can be distinguished from components unrelated to tissue hypoperfusion on the healthy hemisphere. This data-driven method could potentially be a diagnostic tool for perfusion imaging without contrast agents.


Language(s): eng - English
 Dates: 2016-09
 Publication Status: Not specified
 Pages: -
 Publishing info: -
 Table of Contents: -
 Rev. Type: -
 Identifiers: -
 Degree: -


Title: ESMRMB 2016 Congress
Place of Event: Vienna, Austria
Start-/End Date: 2016-09-28 - 2016-10-01

Legal Case


Project information