Reader reliability for our method was excellent. We found high levels of agreement for lesion measurement between experienced and inexperienced readers (ICC 0.89). Interestingly, agreement was slightly less for attenuation measurement of the contralateral normal brain (ICC 0.55), but this did not significantly negatively impact agreement for the calculated attenuation ratio (ICC 0.85) thus estimated time is unlikely to be greatly different between readers.
Such consistency and reliability of lesion measurement are reassuring for the potential future clinical application of our
CT Clock Tool. On the other hand, we did observe greater error for time estimation when scans were performed late versus early. This may reflect fewer attenuation measures in our study at these later times but is also likely indicative of the natural heterogeneity of lesion progression. Rocha and Jovin describe
fast and slow progressors among patients with proximal large vessel obstruction being considered for thrombectomy and suggest that variability in individual patient trajectories relates to differences in their collateral circulation [
20]. Those with good collateral supply are likely to sustain viable brain tissue for longer and this concept is backed by pre-clinical evidence [
21]. In this context, it is possible that some of the apparent discrepancies between our
CT Clock Tool and
elapsed time reflect real CT measurable inter-patient differences, rather than error in our technique. We might be estimating time since infarct rather than time since stroke symptom onset for patients with good collateral supply, but this would require further testing. Kucinski and colleagues demonstrated in a small group (25) of patients with early ischemic stroke (< 5.4 h) that apparent diffusion coefficient values (ADC, a component of diffusion weighted MRI and often considered to represent infarct) of ischemic lesions correlate with measurable non-enhanced CT attenuation changes, albeit weakly [
22]. In a small cohort of patients (41) undergoing thrombectomy, Mokin and colleagues similarly estimated that attenuation ratios of 0.94–0.96 most consistently predicted tissue death [
23]. These attenuation ratios fit with the precipitous decline in attenuation demonstrated during the first hour after stroke onset in our study (Fig.
3). Ultimately, patient-specific lesion measurements are likely to be more powerful predictors of brain tissue viability and thus favourable treatment response than standard measures of elapsed time in stroke and might be used to offer treatment to patients who present late. An interesting and potentially novel observation in our analysis was that the attenuation of affected brain briefly
increased in the very early stages after ischemic stroke onset immediately prior to the described rapid decline. This may reflect increased regional blood volume, which is a recognized perfusion-imaging feature of ischemia rather than infarct and likely a physiological attempt to maximize nutrient delivery to the injured but still viable brain tissue. Whether measuring this increase in CT attenuation could provide a meaningful assessment of brain tissue viability remains untested. Associations between specific lesion measurements thought to indicate tissue viability and improved clinical outcome following treatment, even at extended
elapsed times, are beginning to emerge [
6‐
9]. However, all of these methods require the use of additional or advanced imaging in the acute setting, which for the purpose of broad clinical applicability we were keen to avoid.