Abstract W P47: Developing an Acute Lesion Atlas from Patients with Middle Cerebral Artery Stroke

• Published in: Stroke (1970) 2015-02, Vol.46 (suppl_1)
• Main Authors: Annasamudram, Abhijith, Esterquest, David, Marin, Bianca Maria, McSweeney, Melissa, Brandstatt, Kelly, Voss, Joel, Prabhakaran, Shyam
• Format: Article
• Language: English
• ISSN: 0039-2499; 1524-4628
• DOI: 10.1161/str.46.suppl_1.wp47

Abstract only Background and purpose: Chronic lesion locations are associated with neurologic outcomes. However, voxel-based acute infarct maps have not been used to predict long-term deficits and outcomes. The combination of ischemic lesion volume and location may aid in acute outcome prediction models. We evaluated the feasibility of voxel-based lesion mapping in acute ischemic stroke patients. Methods: In a retrospective study of patients with suspected middle cerebral artery (MCA) territory strokes, we identified acute infarcts on the diffusion-weight imaging (DWI; b=1000) sequences and confirmed them on the apparent diffusion coefficient (ADC) sequences using a semi-qualitative approach. Only patients with all three sequences available were included in the study. We then manually created a lesion map for each patient. Lesions were overlaid on T1-weighted images and transformed to standardized stereotactic space. We overlaid the volumetric lesion masks from all patients to create a composite acute lesion atlas. Results: Among 164 patients, 5 (3.1%) patients had no clear acute lesion and 9 (5.5%) had technical problems with normalization or overlaying, leaving 150 (91.5%) patients for analysis (mean age 66.1 years; 52.0% male; 66.7% white; median NIHSS scale 4). A composite acute lesion atlas was developed (Figure) with regions of high overlap easily visible, suggesting high statistical power for acute lesion-outcome prediction modeling. Conclusion: Acute ischemic lesions can be reliably mapped and used to create aggregated infarct atlases. The considerable overlapping regions in stroke patients might provide sufficient power to detect highly correlated lesion-outcome relationships. The proposed technique may provide a time-sensitive method for identifying acute infarct location, predicting the course of recovery, and selecting patients for early interventions.