epsis mortality prediction with the quotient basis kernel

Abstract

Sepsis is a common clinical syndrome at the intensive care unit (ICU). This condition may lead to severe sepsis, or to a more severe state of septic shock and multiorganic failure, which entails a substantial risk of death. The extreme realtime demands of the ICU require informed decision making on the basis of the available evidence, and such decision making can be supported by semi-automated methods for quantitative mortality prediction. These methods should be robust and feasible within the constraints of the domain. In this paper, we describe a novel sepsis mortality prediction method that first embeds the available data in a suitable feature space, and then uses algorithms based on linear algebra, geometry and statistics for inference. A simplified version of the Fisher kernel (practical Fisher kernel for multinomial distributions) as well as a novel kernel, namely the Quotient Basis Kernel (QBK), are defined and used as the basis for mortality prediction using softmargin support vector machines. The results compare favorably with those obtained using alternative kernels and the standard clinical prediction method based on the basal SAPS score.