As the proximate pathologic cause of ischemic stroke, focal deficits of cerebral blood flow are central to understanding stroke evolution and developing therapies for its treatment. Over the several years of this project we performed serial perfusion MRI studies in ischemic stroke patients. We previously reported that reduction of the volume of ischemia by at least 30% within 2 hours from the start of tPA therapy predicted good clinical outcome. From a further analysis of serial perfusion scans, we developed the pharmacodynamic concept of reperfusion half-life as a measure of thrombolytic drug activity. We modeled the time-reperfusion relationship for untreated and IV tPA-treated ischemic stroke patients to quantify the reperfusion rate characteristic of each condition. We hypothesized that the probability of reperfusion relates to time according to an exponential decay function, analogous to plasma concentration half-life, and that reperfusion half-life following IV tPA therapy would indicate more rapid rates of reperfusion than those in the untreated condition of spontaneous reperfusion. Blinded to clinical and treatment status, we evaluated serial perfusion MRI scans for evidence of reperfusion in 148 patients. We analyzed IV tPA-treated (n=45) and untreated (n=103) patients separately, and used Kaplan-Meier survival analysis to calculate the cumulative probability of reperfusion. We then fit the data with exponential decay functions. The reperfusion half-life (t), the rate at which 50% of the sample reperfuses, was the parameter used to compare the groups. In untreated patients (spontaneous reperfusion), a monoexponential decay function described the data well (R-squared = 0.95) with t = 29.1 hours. In tPA-treated patients, a biexponential decay function, with fast and slow components, was required to describe the data (R-squared = 0.99). The slow component was similar to that of the untreated condition; t=29.21 hours. The fast reperfusion component, attributable to tPA therapy, had t=0.7 hours. By approximately 3.5 hours after start of treatment, the effect of tPA on the probability of reperfusion was negligible. Our data show that modeling reperfusion as an exponential decay function distinguishes spontaneous from tPA-associated reperfusion and provides a measure of the speed and duration of thrombolytic activity. The probability of spontaneous reperfusion over time is a mono-exponential function with a half-life of approximately 29 hours. Treatment with tPA adds a fast reperfusion component with a half-life of 0.7 hours. Determination of reperfusion half-life is a promising approach to evaluate the relative potency and reperfusion effects of different thrombolytic regimens. A paper describing these results has been submitted to a medical journal. Prospective clinical and pre-clinical studies are planned to further study the utility of reperfusion half-life in identifying single or combination thrombolytic therapy that may ? ? In 2004, we published the our discovery of an imaging marker of early blood brain barrier in ischemic stroke and its hypothesized relationship to reperfusion, hemorrhagic transformation, thrombolytic therapy, and worse clinical outcome from a sample of 214 stroke patients. For simplicity of reference we described the observation by the acronym HARM (Hyperintense Acute Reperfusion injury Marker). Several follow-up studies to further elucidate the mechanism and features of HARM have completed enrollment the during current year; data analysis and study report preparations are in progress and we expect to publish these results in the upcoming year. Preliminary results confirm that the localization of HARM is in the CSF space rather than in the parenchyma, prove an association of HARM with serum concentrations of matrix metalloproteinase-9, and demonstrate that HARM can be detected on immediate post-contrast, prior to treatment with tPA. The ability to detect of BBB disruption prior thrombolytic therapy may aid in the development of new therapies to protect the barrier and minimize the potential risks of thrombolytic therapy.? ? Imaging based predictors of stroke outcome and response to therapy are necessary for the utility and validation of imaging biomarkers in drug development. Useful models are those that can distinguish patients destined for good outcomes versus poor outcomes, those who received effective therapy from those who did not, and treatment responders from non-responders. We are investigating several predictive models. These prediction models may be useful for the development, selection and use of acute therapies.? ? A change in acute-to-chronic lesion volume has been proposed as a biomarker for stroke therapies. We determined the magnitude of lesion volume change in 53 patients treated with standard tPA and determined specific volume change thresholds that discriminated clinical responders from non-responders. The mean acute-to-chronic lesion volume increase was 11.7 (7.7) cm-c. In 23 patients, the chronic lesion was smaller than the baseline lesion. At 3 months, 32 patients had an excellent clinical outcome. Dichotomous volume change variables associated with outcome include decrease in volume 30% (p=0.004) and volume increase 5 cm3 (p=0.002). We conclude that change in lesion volume can discriminate between patients destined for good and poor outcomes when treated with effective acute stroke therapy. Thus, lesion volume change may be a useful marker of clinical response in the stroke therapy development.? ? The more accurate prediction of final infarct volume was investigated by combining clinical and imaging variables in a generalized regression neural network model. The model was trained with data from 99 patients that presented with stroke symptoms within 3 hours of onset, underwent acute MRI scans, and were treated with IV tPA therapy using the following variables: sex, age, National Institutes of Health Stroke Scale (NIHSS), stroke onset time to MRI scan time and baseline lesion volume, baseline ischemia volume. The model was then tested in 13 tPA-treated and 30 untreated stroke patients to assess whether the model could differentiate treated from untreated patients. The measured values for final lesion volume did not significantly differ from the predicted values for the tPA-treated patients, but did differ for the untreated patients, indicating that this model, indicating that neural net modeling of infarct volume can discriminate between patients receiving clinically effective stroke therapies and those who do not.? ? We have begun to develop and evaluate other approaches to using baseline imaging volumetric data to predict clinical outcomes. As part of a global project aimed at creating a large scale stroke-specific predictive brain atlas, we correlated acute DWI lesion location with the presence of individual sub-item scores from the NIHSS. Inclusion criteria were: acute ischemic stroke; brain MRI (including DWI) performed within 24 hours of symptom onset; and baseline NIHSS score available. Acute DWI images from all subjects were aligned to a common neuroanatomic coordinate system. Chi square images were calculated on a voxel-by-voxel basis. A total of 163 patients met inclusion criteria. Maps were created that show color-coded chi square values, using a false discovery rate of 5%, for anatomic regions in which there was a significant association between symptom presence and DWI hyperintensity. This is the first voxel-based stroke atlas correlating NIHSS sub-items employing DWI data. The maps illustrate the neuroanatomic representation of the NIHSS in standardized space. The ultimate goal is to build an atlas that uses acute diffusion and perfusion MR imaging to predict long-term outcome for two scenarios: untreated vs. treated with recanalization therapy.
