Moving beyond a single perfusion threshold to define penumbra: a novel probabilistic mismatch definition.

BACKGROUND AND PURPOSE The mismatch lesion volumes defined by perfusion-weighted imaging exceeding diffusion-weighted imaging have been used as a marker of ischemic penumbral tissue. Defining the perfusion lesion by thresholding has shown promise as a practical tool; several positron emission tomography studies have indicated a more probabilistic relationship between perfusion and infarction. Here, we used a randomized controlled trial dataset of tissue-type plasminogen activator 3 to 6 hours after stroke to: (1) quantify the relationship between severity of hypoperfusion (measured by Tmax) and risk of infarction; (2) exploit this relationship to present a novel definition of mismatch based on infarct probabilities rather than dichotomies; and (3) examine the treatment response in the subgroup of patients with mismatch by the new definition. METHODS Patients from the Echoplanar Imaging Thrombolytic Evaluation Trial (EPITHET) were included. Baseline perfusion-weighted imaging and 90-day T2-weighted imaging were coregistered. Perfusion-weighted imaging lesion volumes were divided into 10 Tmax delay strata, and infarct risk was defined as the fraction of the tissue at a given Tmax strata that progressed to infarction by day 90. RESULTS Sixty-two patients were studied. Infarct risk was an increasing function of Tmax for all subgroups, including the whole cohort. The probabilistic approach outperformed all Tmax thresholds, with exception of the Tmax ≥ 10 threshold, for which it was only favored by a trend. CONCLUSIONS Infarct risk and treatment effect increased with severity of perfusion abnormalities. This suggests that a severity-weighted mismatch definition may define penumbral tissue more accurately.