Scientist Center for Neuroscience and Regenerative Medicine, HJF Bethesda, Maryland
Abstract Text: Introduction Peri-infarct depolarizations (PIDs) are waves of spreading depolarization (SD) in cortical gray matter that have been associated with worse outcome following ischemic brain injury1,2. However, the role of PIDs in the expansion of infarct volume remains unclear, likely because clinical monitoring of PIDs is currently performed using electrocorticography, which lacks the spatial resolution to measure the 3D trajectory of the propagating wave. Diffusion MRI is a clinically feasible imaging method that can be used to assess cellular edema and restriction of extracellular space3,4. Dynamic diffusion measurements at sufficient temporal resolution allow for the assessment of the spatiotemporal dynamics of PIDs and their role in infarct growth. Dynamic diffusion abnormalities (DDAs), occurring as propagating waves of transiently reduced apparent diffusion coefficient (ADC), have been observed in animal models of brain injury that elicit SDs5,6, but these observations have typically been performed using manual region of interest (ROI) based methods to detect and map how they propagate in a manner corresponding to electrophysiologically measured PIDs. We describe an automated methodology to detect DDAs and demonstrate its utility in an animal model of ischemic injury. Methods Previously acquired dynamic diffusion data (TR/TE = 2500/45 ms, b = 1000 s/mm2, 0.5mm isotropic voxels) from five male Sprague Dawley rats following ischemic stroke injury7 were processed using TORTOISE to correct for motion and eddy current distortions8. ADC time series maps were computed with a temporal resolution of ten seconds. To identify DDAs likely representing PIDs, a sliding window analysis was used to measure a peak correlation coefficient (Rpeak) of the dynamic ADC within each intracranial voxel with a test kernel representing a one minute transient reduction in ADC consistent with previous measurements of dynamic diffusion during triggered SD9. Voxels with Rpeak greater than 0.80 were defined as falling within the trajectory of at least one DDA. The dynamic ADC of voxels within spatially contiguous ROIs with detected DDAs was smoothed temporally by using a Savitzky-Golay filter and applying the findpeaks function in Matlab (Mathworks, 2022) to the smoothed signal to detect local minima corresponding to a reduction in ADC of at least 15%. Evolution of the infarct volume was assessed using a threshold-based approach by measuring the volume of intracranial voxels with dynamic ADC < 500 μm2/s10 across time. Results Regions of interest without detected DDAs were primarily located in the contralesional hemisphere and the temporal ADC in these voxels remained above the infarct threshold during the imaging time. The majority of voxels with detected DDAs were located along with peri-infarct boundary, and multiple DDAs occurred with a recoverable ADC that remained above the infarct threshold. Infrequently, DDAs were detected in the contralesional hemisphere in voxels where the ADC tended to remain above the infarct threshold. Although DDA positive voxels were present in infarcted tissue they occurred near to the lesion boundary and, the fluctuations in ADC made their relationship to a presumptive unclear. In all animals, infarct volume increased over time and plateaued towards the end of the imaging time course. The volume increases were punctuated by transient bursts in lesion volume that temporally corresponded with DDAs in perilesional and contralesional regions where the ADC in voxels were transiently reduced below the defined infarct threshold. Across the five animals, the lesion volume increased relative to the baseline by a mean of 17±9% per hour. Discussion The proposed analysis not only detected DDAs in ipsilesional tissue, both within the developing infarct and along the perilesional boundary, but also in contralesional tissue. Consistent with literature of electrophysiologically measured PIDs11, DDAs in perilesional tissue were preferentially located along the infarct boundary. The mechanism of spread of DDAs to the contralesional hemisphere has yet to be determined and is worth further investigation. An ROI based analysis of the spatiotemporal relationship between the trajectory of perilesional DDAs and the expansion of infarct volume would determine whether the incremental increase in lesion volume that is associated with PIDs occurs as part of PID initiation or during their subsequent propagation.