Automated acute ischemic stroke lesion delineation based on apparent diffusion coefficient thresholds

Purpose Automated lesion segmentation is increasingly used in acute ischemic stroke magnetic resonance imaging (MRI). We explored detail the performance of apparent diffusion coefficient (ADC) thresholding for delineating baseline diffusion-weighted (DWI) lesions. Methods Retrospective, exploratory analysis prospective observational single-center 1000Plus study from September 2008 to June 2013 ( clinicaltrials.org ; NCT00715533). built a fully automated algorithm using fixed ADC threshold (≤620 × 10–6 mm 2 /s) delineate DWI and analyzed its compared manual assessments. Diagnostic capabilities best possible thresholds were investigated receiver operating characteristic curves. Influential patient factors on techniques’ studied by conducting multiple linear regression. Results 108 patients selected analysis. The median Dice was 0.43 (IQR 0.20–0.64). Mean values β = −0.68, p < 0.001) volumes 0.29, predicted performance. Optimal individual differed between subjects with ≤691 10 −6 /s ≤660–750 /s). −0.96, mean brain parenchyma 0.24, associated thresholds. Conclusion lesions varies substantially patients. It influenced such as size well parenchymal values. Considering inherent noisiness maps, threshold-based delineation very small not reliable.