Assessing neuronal density in peri-infarct cortex with PET: Effects of cortical topology and partial volume correction

The peri-infarct cortex (PIC) is the site of long-term physiologic changes after ischemic stroke. Traditional methods for delineating the peri-infarct gray matter (GM) have used a volumetric Euclidean distance metric to define its extent around the infarct. This metric has limitations in the case of cortical stroke, i.e., those where ischemia leads to infarction in the cortical GM, because the vascularization of the cerebral cortex follows the complex, folded topology of the cortical surface. Instead, we used a geodesic distance metric along the cortical surface to subdivide the PIC into equidistant rings emanating from the infarct border and compared this new approach to a Euclidean distance metric definition. This was done in 11 patients with [F-18]-Flumazenil ([18-F]-FMZ) positron emission tomography (PET) scans at 2 weeks post-stroke and at 6 month follow-up. FMZ is a PET radiotracer with specific binding to the alpha subunits of the type A γ-aminobutyric acid (GABAA) receptor. Additionally, we used partial-volume correction (PVC) of the PET images to compensate for potential cortical thinning and long-term neuronal loss in follow-up images. The difference in non-displaceable binding potential (BP_ND) between the stroke unaffected and affected hemispheres was 35% larger in the geodesic versus the Euclidean peri-infarct models in initial PET images and 48% larger in follow-up PET images. The inter-hemispheric BP_ND difference was approximately 17–20% larger after PVC when compared to uncorrected PET images. PET studies of peri-infarct GM in cortical strokes should use a geodesic model and include PVC as a preprocessing step. Hum Brain Mapp 38:326–338, 2017.