Automatic Segmentation of Perinatal Arterial Ischemic Stroke Volume

摘要: Introduction: Diffusion weighted imaging (DWI) allows rapid diagnosis of ischemic tissue following perinatal arterial stroke. Current delineation stroke is user-dependent. Aim this study was to investigate whether volume can be assessed automatically perform user-independent measurements volume. Methods: DWI scans 12 full-term infants without any cerebral abnormalities (controls) were registered the patient's scan using non-diffusion image. Subsequently, apparent diffusion coefficient (ADC) maps controls transformed ADC map patient corresponding transformations. Patient's values 2.5 standard deviations below mean value classified as ischemic. The manually delineated four times two different window widths and levels, subsequently compared with automatic segmentation. Dice similarity (DSC) calculated measure (2* (overlap segmentation1 & segmentation2)/(segmentation1+segmentation2). Results: Stroke volumes 9 (scanned day 3-6) segmented. DSCs manual segmentations lower when levels used (.89, .85-.95) than same level (.96, .92-1.0). segmentation resulted in a DSC .86(.76-.93). Of interest voxels, which not included segmentation, but on inspection did show relatively low suggesting penumbral tissue. Conclusions: Automatic approaches indices comparable may assist clinicians assessment tool evaluate therapeutic intervention