This paper described a z-score mapping method on the basis of a voxel-by-voxel analysis to visualize hypoattenuation regions of hyperacute stroke on unenhanced CT images. The algorithm of this method consisted of 5 major steps,
i.e., anatomic standardization, construction of a normal reference database, calculation of z score, elimination of false positive areas, and extraction of hypoattenuation areas. The performance of this method in detection of hypoattenuation with a series of 21 patients with infarction (< 3 hours) showed the AUC =0.834 for distinction between hypoattenuation and normal regions. The result of an observer study with five neuroragiologists, also, demonstrated the average AUC for detection oh hypoattenuation regions significantly improved from 0.883 to 0.925 (
P< 0.01). Therefore, the z-score mapping method could effectively extract hypoattenuation regions that showed a high z-score value on the z-score map. The use of this method has the potential to help neuroradiolgists detect hypoattenuation regions of hyperacute stroke on unenhanced CT images.
View full abstract