Prediction of histopathological outcome using averaged multimoidal information in rat

Commonly used autoradiographic image analysis for stroke research has been restricted to the assessment of local cerebral blood flow (LCBF), glucose utilization (LCMRglc) or messenger RNA (mRNA) (by in situ hybridization) in individual brains, with repeated measurements from the same animal. Histopathological analysis of perfusion-fixed paraffin-embedded brain material has long been regarded as the “gold standard” for the quantitative assessment of tissue injury in experimental models of cerebral ischemia. It is desirable to reveal the interrelationship between signals measured at the time of ischemia and the final distribution of cell damage. In this study averaged images for each modality were derived from individual rats and infarction frequency distribution maps were generated. A single-layer perceptron learning machine with linear and non-linear discriminant functions were constructed to function as a predictive model. The original infarction frequency distribution map and predicted frequency map are presented