We have described and developed the imaging marker Hyperintense Acute Reperfusion Marker (HARM) as a measure of early blood-brain barrier (BBB) disruption in acute stroke. We are developing clinical trials, using HARM as the biomarker, to limit BBB disruption and associated hemorrhagic transformation in patients treated with tPA. This marker has been defined as delayed gadolinium enhancement of the cerebrospinal fluid in the subarachnoid space on fluid-attenuated inversion recovery (FLAIR) MRI. HARM has been associated with risk of hemorrhagic transformation, thrombolytic therapy, and worse clinical outcome. HARM has also been associated with increased age, but it is unknown whether this association is confounded by impaired gadolinium clearance due to age-related decreases in renal function. Patients (n=187) were selected from the LESION project if they had the following features: treatment with IV-tPA, gadolinium enhanced MRI at baseline, and FLAIR MRI at 2 hr or 24 hr post tPA. HARM was defined as enhancement of sulcal spaces in 10 or more FLAIR slices. The estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation (CKD-EPI) using admission plasma creatinine values. Age, race, sex, NIHSS at presentation, hypertension history, diabetes history, plasma creatinine, eGFR, number of gadolinium doses and MRI field strength were entered into a conditional stepwise logistic regression as potential covariates of HARM. HARM was independently associated with age (p = 0.003;odds ratio, 1.047;confidence interval, 1.016-1.079), after adjustment for eGRF, number of gadolinium doses, MRI field strength. The other factors were not significant predictors of HARM. This study indicates that age is an independent predictor of BBB disruption seen as HARM even after adjusting for the effects of the eGFR. These findings suggest that older patients are at higher risk for BBB disruption following treatment with IV-tPA, independent of their kidney function. We have investigated whether ischemic brain lesions after heart surgery may have a high enough frequency to serve as potential imaging markers with which to assess novel neuroprotective stroke therapies. A total of 19 patients were imaged 3 days (1.1) after coronary artery bypass graft (CABG) or valve replacement surgery. We studied two types of ischemic brain lesions: parechymal infarction on high-resolution diffusion MRI (DWI) and breakdown of the blood brain barrier seen on on post-gadolinium FLAIR as enhancement in the subarachnoid space, a phenomenon known as HARM (hyperintense reperfusion injury marker). HARM reflects breakdown of the BBB at the time of gadolinium administration, and persists for at least a week. Gadolinium contrast was administered 24 hours after surgery to assess the integrity of the BBB at that time. DWI and FLAIR imaging was done 24-72 hours after surgery. Two stroke neurologists evaluated the DWI for new ischemic lesions, and the FLAIR for evidence of HARM. None of the patients had a clinical stroke and all had a normal neurological examination. Nine patients (47%) had HARM, 14 (74%) had lesions on DWI, and 7 (37%) had both. Not all patients with DWI lesions had HARM, and not all patients with HARM had DWI lesions: the association between DWI lesions and HARM was not significant (p=0.56). Most (16 of 19) of these patients undergoing cardiac surgery have imaging signs of ischemic brain injury post-operatively, despite absence of gross clinical signs. This is the first report of MRI-demonstrated BBB disruption after cardiac surgery, approximately half of whom have evidence of BBB disruption within 24 hours of surgery. Further studies are needed to determine if BBB disruption in the cardiac surgery patient occurs because of focal or global ischemia, reperfusion injury, changes in the level of matrix metalloproteinases (MMPs), or inflammatory mechanisms. The high incidence of brain lesions after cardiac surgery supports the potential of this population for proof of principle therapeutic trials of acute neuroprotective therapies in stroke. Another aim of our research into MRI markers of stroke is to evaluate and compare methods for threshold based segmentation (TBS) of ischemic and penumbral areas without lengthy processing and reader expertise. We tested specific Apparent Diffusion Coefficient (ADC) and perfusion Tmax thresholds for identification of ischemic and penumbral areas in acute stroke patients. Seventy-two patients (69 +/- 15 years, 39 females) were selected from the LESION project if scanned within 3 hours from onset, prior to acute intervention, and high quality diffusion and perfusion MRI. Patients had a median NIHSS of 10 (IQR 6-18) and median onset to MRI time of 124 minutes (IQR 83-163). TBS of MRI scans were determined by generating the perfusion maps and then measuring the ADC and Tmax volumes using an ADC value of 615 mm2/seconds and Tmax value of >6 seconds. Relative to the standard qualitative assessment of the images, identification of lesion by ADC TBS agreed in 80% of the cases, with a positive predictive value of 96%. Tmax TBS decisions were in agreement with the qualitative read of perfusion deficit in 90% of the cases, with a positive predictive value of 97%. TBS, using an ADC value of 615 mm2/seconds and Tmax value of >6 seconds demonstrates potential for use in interpretation of MRI for acute intervention decision making. The choice of PWI maps, such as first moment method MTT or deconvoluted Tmax, and segmentation approach can impact the estimated diffusion-perfusion (penumbral) volume and thus the selection of patients for inclusion in clinical trials. We compared the PWI lesion and mismatch volumes determined using standard MTT maps segmented manually, to those obtained using Tmax and an automated segmentation algorithm. Twenty-seven iv tPA treated patients having pre- and 2hr post-treatment MRI were included. A trained imaging scientist manually segmented lesions on DWI and MTT maps to form the gold standard (GS). Tmax was computed by AIF deconvolution and automated lesion segmentation and mismatch classification was performed using: i) an automated algorithm (Auto-Tmax) and ii) basic thresholding Tmax maps at 4s, 6s, 8s and 10s, with no additional processing. Patients were classified as mismatch=true if mismatch volume exceeded a range of cutoff values of 25cc, 50cc, 75cc, or 100cc. Agreement with GS was compared across technique and cutoff values.For both pre and post tPA treatment MRI, Auto-Tmax and basic thresholding Tmax >=6s were found to provide similar volume estimates to those of GS. With Auto-Tmax, the mismatch agreement with GS was stable (83%), independent of cutoff value. For the basic Tmax >=6s, the best agreement was 81% at >25 cc and dropped off with increasing cutoff value. The error rates for mismatch agreement were higher with basic threshold segmentation compared to Auto-Tmax and worsened as the mismatch volume cutoff value was increased. Both Auto-Tmax and basic Tmax threshold of >=6s provides a similar estimate of perfusion volume and mismatch detection to that of manual segmentation. However, Auto-Tmax resulted in a more robust classification independent of cutoff value, providing an automated method for mismatch detection that could eventually be used to select patients prior to therapy.
