Abstract
BACKGROUND: The National Institutes of Health Stroke Scale (NIHSS) is widely used to assess stroke severity. While prior studies have identified subcortical regions where infarcts correlate with NIHSS scores, stroke symptoms can also arise from hypoperfusion, not just infarcts. Understanding the potential for neurological recovery post-reperfusion is essential for guiding treatment decisions. The goal of this study was to identify brain regions where hypoperfusion correlates with NIHSS scores, using computed tomography perfusion (CTP) scans in cases of acute ischemic stroke. METHODS: In this prospective observational study, we analyzed CTP scans and NIHSS scores from 89 patients in the acute phase. We employed a unique support vector regression approach to overcome limitations of traditional mass univariate analyses. Additionally, we used stability selection to identify the most consistent features across subsets, reducing overfitting and ensuring robust predictive models. We verified the consistency of results through nested cross-validation. RESULTS: Both cortical and subcortical areas, including white matter tracts, showed associations with NIHSS scores. These regions aligned with functions such as language, spatial attention, sensory, and motor skills, all assessed by the NIHSS. CONCLUSIONS: Our findings reveal that hypoperfusion in specific brain regions, including previously underreported cortical areas, contributes to NIHSS scores in acute stroke. Moreover, this study introduces a novel brain mapping approach using CTP imaging and stability selection, offering a more comprehensive view of acute stroke impairments and the potential for recovery before structural reorganization occurs.