Imaging Selection in Ischemic Stroke: Feasibility of Automated CT-Perfusion Analysis

You better hope you are one of the strokes that meets these cherry picked patients so you can get proper interventions. Real world difficult strokes need not apply for interventions. I would leave no stroke patient hanging with no decent interventions possible. This cherry picking crap is what happens when you don't have stroke survivors running the strategy and research. 

Imaging Selection in Ischemic Stroke: Feasibility of Automated CT-Perfusion Analysis

1. Bruce C.V. Campbell1,2,*
2. Nawaf Yassi1
3. Henry Ma3
4. Gagan Sharma2
5. Simon Salinas2
6. Leonid Churilov3
7. Atte Meretoja1,3
8. Mark W. Parsons4
9. Patricia M. Desmond2
10. Maarten G. Lansberg5
11. Geoffrey A. Donnan3
12. Stephen M. Davis1

1. ¹ Department of Medicine, Royal Melbourne Hospital, Melbourne, Australia
2. ² Department of Radiology, Royal Melbourne Hospital, Melbourne, Australia
3. ³ Florey Institute of Neuroscience and Mental Health, University of Melbourne, Melbourne, Australia
4. ⁴ Priority Research Centre for Brain and Mental Health Research, John Hunter Hospital, University of Newcastle, Newcastle, Australia
5. ⁵ Department of Neurology and Neurological Sciences, Stanford Stroke Center, Stanford University Medical Center, Stanford, CA, USA

1. ↵* Correspondence: Bruce Campbell, Department of Neurology, Royal Melbourne Hospital, Grattan Street, Parkville, VIC 3050, Australia. E-mail: bruce.campbell@mh.org.au

Next Section

Abstract

Background Advanced imaging may refine patient selection for ischemic stroke treatment but delays to acquire and process the imaging have limited implementation.

Aims We examined the feasibility of imaging selection in clinical practice using fully automated software in the EXTEND trial program.

Methods CTP and perfusion-diffusion MRI data were processed using fully-automated software to generate a yes/no ‘mismatch’ classification that determined eligibility for trial therapies. The technical failure/mismatch classification error rate and time to image and treat with CT vs. MR-based selection were examined.

Results In a consecutive series of 776 patients from five sites over six-months the technical failure rate of CTP acquisition/processing (uninterpretable maps) was 3·4% (26/776, 95%CI 2·2–4·9%). Mismatch classification was overruled by expert review in an additional 9·0% (70/776, 95%CI 7·1–11·3%) due to artifactual ‘perfusion lesion’. In 154 consecutive patients at one site, median additional time to acquire CTP after noncontrast CT was 6·5 min. Subsequent RAPID processing time varied from 3–10 min across 20 trial centers (median 5 min 20 s). In the EXTEND trial, door-to-needle times in patients randomized on the basis of CTP (n = 47) were median 78 min shorter than MRI-selected (n = 16) patients (P < 0·001).

Conclusions Automated CTP-based mismatch selection is rapid, robust in clinical practice, and associated with faster treatment decisions than MRI. This technological advance has the potential to improve the standardization and reproducibility of interpretation of advanced imaging and extend use to practice settings beyond highly specialized academic centers.