Visual assessment of magnetic resonance imaging perfusion lesions in a large patient group.

PURPOSE: Few magnetic resonance imaging (MRI) studies of stroke have evaluated the value of visual assessment of perfusion/diffusion mismatch, which is crucial for routine application. In this study an attempt was made to visually assess perfusion lesions resembling the acute clinical situation and identify parameters with the highest interobserver reliability when used to define a perfusion/diffusion mismatch and the highest accuracy for prediction of infarct growth.
METHODS: Magnetic resonance imaging was performed within 6 h of symptom onset and again 1-11 days thereafter in 86 consecutive stroke patients who received intravenous thrombolytic therapy. The MRI protocol included diffusion-weighted imaging apparent diffusion coefficient (DWI/ADC), fluid-attenuated inversion recovery (FLAIR) and perfusion imaging (PI). Maps for different perfusion parameters, e.g. cerebral blood volume (CBV), cerebral blood flow (CBF), mean transit time (MTT) and time to peak (TTP) were calculated. Areas of perfusion deficits of all perfusion parameters were visually compared to corresponding ADCs and final infarct size by two independent observers.
RESULTS: The final infarct size was overestimated by TTP (in 81/83 patients by raters 1 and 2, respectively), MTT (82/83) and CBF (65/74) lesions. The ADC lesions were rated smaller than the final infarct size in 43/38 cases by raters 1 and 2 and the CBV decrease was rated to underestimate final infarct size in 40/31 cases. The only significantly increased OR of 3.883 (95 % CI 1.466-10.819, p = 0.004, rater 1)/5.142 (95 % CI 1.828-15.142, p = 0.001, rater 2) for predicting infarct growth was observed for the presence of a CBV > ADC mismatch, which also showed the highest kappa value of 0.407.
CONCLUSIONS: All mismatch patterns were prone to high interrater variability when assessed under conditions resembling the clinical setting. Of all tested mismatch patterns the CBV > ADC mismatch was the strongest predictor of lesion growth while visual assessment of TTP and CBF generally resulted in an overestimation of infarct sizes and the presence of a TTP > ADC or CBF > ADC mismatch was not significantly predictive for lesion growth. Visual inspection of these most commonly used mismatch patterns has a low value for the prediction of infarct growth and thus the estimation of the penumbra in ischemic stroke patients.