Analysis of active feature selection in optic nerve data using labeled fuzzy C-means clustering

Describes an iterative analysis technique that aids in the process of searching for an optimal set of features for classification, and its application to detection of early glaucoma from optic nerve data in an evolving data acquisition system. The selection and evaluation of features were done using fuzzy C-means clustering and support vector machines. The clustering method was updated using a semi-supervised process. The search space for feature selection was reduced using an active feature selection algorithm. Data samples from different stages of the evolving system are analyzed and evaluated.