Feasibility of using brain attenuation changes on CT to accurately predict time of ischaemic stroke onset

Grant Mair, Peter Sandercock, Joanna Wardlaw

Research output: Contribution to conferencePoster

Abstract

Introduction Following ischaemic stroke, CT attenuation of affected brain reduces with time. We piloted whether attenuation of infarct can be used to predict time of stroke onset. Methods We identified patients from the Third International Stroke Trial with large cerebral infarct on thin-slice (<2.5 mm) non-contrast CT. We selected a range of stroke onset to scan times (time). A neuroradiologist manually applied regions of interest within the infarct and an equivalent contralateral location and derived an attenuation ratio. We allocated cases into development and testing datasets (75/25%) blind to attenuation ratio. Attenuation ratios in the development dataset were plotted against time, a best-fit logarithmic function determined, and used to estimate time in the test dataset. Estimates for time were correlated with true time in the test dataset and differences assessed. Results We included 120 scans (75 patients), of time range 22 minutes to 8 days, median 11.75 hours. Development (n=90) and test (n=30) datasets had equal distributions of time and attenuation ratio, p>0.18. The logarithmic model estimated an attenuation ratio of 0.8 at 6 hours. Estimated and true time values in the test dataset were highly correlated (0.93, p<0.0001). Time estimation errors were greatest at extended times: 90% (9/10) scanned <5 hours had error <100 minutes; versus 46% (11/24) scanned <48 hours. Conclusion This pilot analysis suggests that it is feasible to estimate time after ischaemic stroke onset using only CT brain attenuation, particularly during the early, most clinically relevant period. Analyses are ongoing to further develop this technique.
Original languageEnglish
Publication statusPublished - May 2017

Fingerprint Dive into the research topics of 'Feasibility of using brain attenuation changes on CT to accurately predict time of ischaemic stroke onset'. Together they form a unique fingerprint.

Cite this