Mathematical evaluation of diagnostic informativity of laboratory values in serous and purulent pyelonephritis

From a practical point of view, the recognition of different forms of acute pyelonephritis should be most effective in a short time and through the use possible of a limited number of routine laboratory methods of the study. Purpose: development of forecasting the severity of acute pyelonephritis system using neural networks. Materials and methods. We used the neural network approach of the analysis of the results of examination of patients with acute pyelonephritis, based on the use of self-learning neural structures. The main advantage of neural network classifier is a high degree of confidence of predicting the correct diagnosis in a patient with acute pyelonephritis based on the use of a minimum number of subjective and objective index, medical history data and laboratory parameters. Results and their discussion. The use of laboratory immune parameters (TNF, factor H and С3а) advanced neural network classifier in patients with acute pyelonephritis increases the accuracy of prediction of serous or purulent pyelonephritis to 99%. Conclusion. The possibility of obtaining verbal descriptions with high prediction accuracy provides additional information for diagnosis and determination of treatment strategies in these patients.