Pneumonia Prognosis During Acute Poisoning as a Problem of Pattern Recognition

Early prognosis of pneumonia is an urgent problem of modern medicine. Application of methods of the applied mathematics can significantly increase efficiency of its solution. In the present work the problem of pneumonia prognosis is solved by pattern recognition on the example of acute poisoning by psychotropic and hypnotic drugs. It is established, that the reliable prognosis can be made, when values of only two parameters of the blood formula are known. 2-D approximations of conditional data distributions of these parameters are constructed on the basis of two approaches: approximation by 2-D normal distribution and approximation by Pearson curves product constructed in orthogonal coordinates related to a direction of data correlation minimum. On the basis of founded approximations, Byes classifier for pneumonia prognosis is obtained. The results are based on the information about 129 patients treated at Poisoning Center of N.V.Sklifosovsky Emergency Medical Aid Scientific Research Institute.