Fourier and fractal analysis of cytoskeletal morphology altered by xenobiotics

Abstract

The cytoskeletal microtubules (<i>MT</i>s) of rat hepatocytes treated by <i>Benomyl</i> (a fungicide) were imaged by means of immunofluorescent staining and optical microscopy. Images of untreated, or control (<i>C</i>), and of treated (<i>T</i>) cells were processed both by fractal and Fourier analysis. The <i>C</i>-<i>MT</i>s had contour fractal dimensions higher (&ge; 1.4) than those of <i>T</i>-<i>MT</i>s (&le;1.3). Fourier analysis included computation of the anisotropy of power spectral density, angle averaging and "spectrum enhancement," which corresponds to the application of a (pseudo)differential operator to the image. Enhanced spectra were interpolated by a polynomial, <i>q</i>, of degree 39, from which morphological descriptors were extracted. Descriptors from Fourier analysis made image classification possible. Principal components analysis was applied to the descriptors. In the plane of the first two components, {z₁,z₂}, the minimum spanning tree was drawn. Images of <i>T</i>-<i>MT</i>s formed a single cluster, whereas images of <i>C</i>-<i>MT</i>s formed two clusters, <i>C</i>1 and <i>C</i>2. The component z₁ correlated positively with the local maxima and minima of <i>q</i>, which reflected differences between <i>T</i> and <i>C</i> in power spectral density in the 1 to 2000 cycles/mm spatial frequency band. The difference between <i>C</i>1 and <i>C</i>2 was ascribed to anisotropy of the <i>MT</i> bundles. The implemented image classifier is capable of telling differences in cytoskeletal organization. As a consequence the method can become a tool for testing cytotoxicity and for extracting quantitative information about intracellular alterations of various origin.