Objective. To work out the techniques of forecasting the development of postoperative cognitive dysfunction in early and late periods after coronary bypass surgery using fully connected multilayer neural networks. Material and methods. Based on the study of 264 patients with stroke, 9 neural network models for prognosis of cognitive dysfunction after coronary bypass surgery and 7 models for prognosis of cognitive dysfunction 12 months after surgery were built. Fully connected multilayer neural networks of direct propagation were used for working-out the prognostic techniques. Results and conclusion. The high prognostic significance of neural networks in relation to cognitive disorders in early and late periods after coronary bypass surgery under extracorporeal circulation was demonstrated. Possibilities of creating neural networks make their application in cardiac surgery clinics available and promising.