Classification of trichostrongyle eggs in ruminant fecal samples using a back propagating neural network

Abstract

This article examines the use of a back-propagating neural network to count the number of parasite eggs present in a given fecal sample. If possible this would save hours of trained labor currently used for the task and potentially improve the accuracy of the procedure. The preliminary results of this study showed that the procedure could be performed with an error rate of less than five percent by a properly trained and configured network. Further study is needed to determine whether the method is viable for more expansive data sets, and whether the current configuration is optimal.