Classification of microwave scattering data based on a subspace distance with application to detection of bleeding stroke

Abstract

This paper demonstrates the usefulness of a classifier based on a subspace distance for the detection of bleeding stroke based on microwave scattering measurements from an antenna array placed around the skull. This discriminating classifier is suitable for high dimensional data applications when the number of training data samples is less than the data dimension. The proposed classifier was tested on both clinical and experimental data to separate bleeding subjects from non-bleeding ones. A pseudo-inverse Mahalanobis distance based classifier and a classifier based on the Euclidean distance were used on clinical data for the purpose of comparison with the proposed classifier.