N.V. Orlov, D.M. Eckley, L. Shamir, and I.G. Goldberg (USA)
Global features, image transforms, and biomedical images
We present a method for automatic classification of biomedical images. The method exploits a machine vision approach to analyze image content based on a global set of image descriptors combined with image transforms. High dimensional feature spaces were used, and Fisher ranking was applied for weighting and truncating dimensions of the original feature space. Weighted nearest distance method was used as a basic classifier. The approach was applied to a variety of imaging problems including automatic classification of lymphoma malignancies and data from high-content screening experiments. The large degree of applicability of the proposed global feature set allows a search for patterns in a wide range of image types, across different domains. Importantly, we demonstrate certain capabilities of the machine vision approach in the biomedical domain.
Important Links:
Go Back
