Data mining of patients on weaning trials from mechanical ventilation using cluster analysis and neural networks

摘要: The process of weaning from mechanical ventilation is one the challenges in intensive care. 149 patients under extubation (T-tube test) were studied: 88 with successful trials (group S), 38 who failed to maintain spontaneous breathing and reconnected F), 23 test but that had be reintubated before 48 hours R). Each patient was characterized using 8 time series 6 statistics extracted respiratory cardiac signals. A moving window statistical analysis applied obtaining for each a sequence patterns features. Applying cluster two groups majority dataset obtained. Neural networks discriminate between S, F R. best performance obtained 84.0% well classified linear perceptron trained feature selection procedure (that selected 19 features) taking as input main centroid. However, classification baseline 69.8% could not improved when original set instead centroids classify patients.