Processing of physiological signals for assessment of blood loss

Hemorrhage potentially impacts the chance of survival in most traumatic injuries. Thus, it would be highly desirable to maximize the survival rate in cases of blood loss by identifying the presence of hemorrhage prior to the occurrence of hemorrhagic shock with biomedical signals. Since analyzing a single physiological signal may not be enough to accurately identify blood loss and the severity of blood loss, the information contained in transcranial Doppler (TCD) signal when added to other typically used physiological signals---such as the electrocardiogram (ECG)---may be useful in identifying the presence and severity of hemorrhage.;In this study, the severity of hemorrhage is classified in two different classification types: once into three classes as mild, moderate and severe, and once into two classes severe and non-severe. The data for this study is generated using a human simulated model of hemorrhage, which utilizes lower body negative pressure (LBNP). The analysis of signals is done by applying discrete wavelet transformation (DWT). The wavelet-based features are defined using the detail and approximate coefficients and machine learning algorithms are used for classification. The objective of this study is to evaluate the impact of adding the information in the TCD signal into other physiological signals to classify the severity of blood loss. The results of this study show prediction accuracies above 85% by support vector machine, in identifying the presence and severity of blood volume loss when a combination of physiological signals including TCD are processed as input to the proposed system.