The possibility of the classification about Ht values using the AR model

Abstract

There are various blood purifications. Plasma Exchange(PE) is one of them. It is very important for medical workers to classify the changing Hematocrit(Ht) values during PE. Ht values have been obtained by CLIT-LINE monitor. Ht values mean the circulating blood volume. At present, the medical worker classifies the changing Ht values during the PE. The classification is depended on their experiment. Therefore, the classification is subjective. If we can find the information of the changing Ht values using the AR model, and can predict the changing Ht values, PE is treated more safety. Hence, our purpose is the classification into the rest of Ht values using the Ht values for less than 20 minutes from the beginning of PE by AR model. Initially, we have classified the Ht data into 3 groups. The 3 groups are falling, constant, and rising Ht values. The coefficients of ARmodel have updated by the delta rule using the Ht values of less than 20 minutes from the beginning of PE. AR model has captured the character of Ht values for less than 20 minutes from the beginning of PE, and rest of Ht values have been classified by the coefficients about the AR model. As the result, it is difficult to classify the rest of Ht value using this method. However, we have found that Ht values for less than 20 minutes from the beginning of PE must include some information about the rest of Ht values. © 2009 Springer-Verlag.