Texture analysis and optical anisotropy measurements of leukocytes for early diagnostics of diabetes mellitus

DNA damage had been suggested to contribute to the pathogenesis of insulin dependent diabetes mellitus (IDDM). In this work we present a method for detection and discrimination of such DNA changes by means of analysis of light microscope images of anisotropically stained leukocytes nuclei. Several features of the images have been evaluated, including integrated optical density, degree of polarisation, and textural features. A genetic algorithm, coupled with a neural classifier, has been used to find the best features for identification of the pathology. Reported results indicate the best set is able to achieve an 83% correct classification ratio.