2005 Conference Objects

Ground Truth for Evaluation of Ischemic Stroke Hybrid Segmentation in a Rat Model of Temporary Middle Cerebral Artery Occlusion

Imielinska, Celina Z.; Rosiene, J.; Jin, Y.; Liu, Xin; Udupa, Jayaram K.; Zacharia, B.; Komotar, R.; Mocco, J.; Sughrue, M.; Grobelny, B.; Sisti, A.; Silverberg, J.; Khandji, J.; Cohen, H.; Connolly Jr., E. Sander; D'Ambrosio, Anthony

In vivo rodent models of focal cerebral ischemia have been developed to investigate stroke therapy. Typically these models require rapid quantification of cerebral infarct volumes using vital stains with tetrazolium salts to delineate the extent of neuronal death. To avoid animal sacrifice, we sought a study with MR acquired volumetric rata data where surrogate of ground truth is obtained by repeated manual delineation by experts, and an automated hybrid segmentation is evaluated for accuracy. We propose a rating system for the expert delineations that captures intra- and inter-expert discrepancy. Our preliminary results show that surrogate ground truth derived from MR data is at least as good as the one derived from histologic stained slices. Hence animal sacrifice is not necessary to evaluate ischemic stroke automated segmentation in a rat model of temporary middle cerebral artery occlusion.