Assessing the Neurological Sequelae of COVID-19 and Acute Respiratory Distress Syndrome
Abstract
COVID-19 infection leading to acute respiratory distress syndrome (ARDS) has been associated with impaired neurocognitive function and is known to increase the risk for endothelial dysfunction and coagulopathy affecting the vasculature of the brain. Recent studies have reported a higher concentration of cytokine and glutamate receptors along white matter tracts which may increase susceptibility to inflammatory-induced damage, further affected by hypoxemia due to direct and indirect lung damage. We conducted a systematic review and meta-analysis which suggests that the combination of ARDS and COVID-19 doubles the risk of developing intracranial hemorrhage and increases vulnerability to cerebral white matter injury as opposed to ARDS of other causes. Given the findings of the review, we then assess image analysis methods to detect these injuries, particularly white matter hyperintensities (WMH). Although various deep learning techniques have been proposed to automatically quantify WMH, the influence of image preprocessing steps on segmentation accuracy has been underexplored. We examine the impact of five intensity normalization methods on deep learning segmentation accuracy, emphasizing the importance of careful consideration for WMH analysis in COVID-19 ARDS populations. Further exploration of intensity normalization approaches using the Neuro-SAVE ICU data is underway to determine the optimal method for WMH analysis.