In article a possibility of application neural network methods of an estimation of a degree of surgical risk is surveyed. The process of the analysis of the medical data is stage by stage with the help of neural network methods described. As the most suitable, it is offered to use two architectures – probability networks and the multilayer perceptrons. Matching the given architectures of networks is represents, possible limitations on input datas and interpreting of results of data processing is describes. Obtained earlier, results of an estimation of a degree of surgical risk for patients by the mechanical icterus by means of nonparametric probability-statistical methods also are given.