Related Experiment Video
Updated: Mar 26, 2026
A Magnetic Resonance Imaging Protocol for Stroke Onset Time Estimation in Permanent Cerebral Ischemia
Published on: September 16, 2017
Optimising MR perfusion imaging: comparison of different software-based approaches in acute ischaemic stroke
Lars-Arne Schaafs1,2, David Porter3, Heinrich J Audebert4
1Department of Radiology, Charité-Universitätsmedizin, Hindenburgdamm 30, 12203, Berlin, Germany. lars-arne.schaafs@charite.de.
Objectives:
Perfusion imaging (PI) is susceptible to confounding factors such as motion artefacts as well as delay and dispersion (D/D). We evaluate the influence of different post-processing algorithms on hypoperfusion assessment in PI analysis software packages to improve the clinical accuracy of stroke PI.
Methods:
Fifty patients with acute ischaemic stroke underwent MRI imaging in the first 24 h after onset. Diverging approaches to motion and D/D correction were applied. The calculated MTT and CBF perfusion maps were assessed by volumetry of lesions and tested for agreement with a standard approach and with the final lesion volume (FLV) on day 6 in patients with persisting vessel occlusion.
Results:
MTT map lesion volumes were significantly smaller throughout the software packages with correction of motion and D/D when compared to the commonly used approach with no correction (p = 0.001-0.022). Volumes on CBF maps did not differ significantly (p = 0.207-0.925). All packages with advanced post-processing algorithms showed a high level of agreement with FLV (ICC = 0.704-0.879).
Conclusions:
Correction of D/D had a significant influence on estimated lesion volumes and leads to significantly smaller lesion volumes on MTT maps. This may improve patient selection.
Key Points:
• Assessment on hypoperfusion using advanced post-processing with correction for motion and D/D. • CBF appears to be more robust regarding differences in post-processing. • Tissue at risk is estimated more accurately by correcting software algorithms. • Advanced post-processing algorithms show a higher agreement with the final lesion volume.
