Cytoskeletal structures can be affected by external factors including ultrasound. Our task was to develop a structure analysis method to evaluate these changes quantitatively. We exposed HeLa cells to continuous ultrasound (1 MHz, 1 and 2 W/cm2, 10 min at 37 degrees C). The microtubules were detected by the monoclonal antibody TU-01/SwAM/FITC, observed in a fluorescence microscope and photographed digitally. The images were processed by "FFT magic" software. The structure analysis is based on frequency domain filtering using discrete Fourier transform. The basic idea is to design filters to extract information describing best the structural changes. The properties of the filter can be enhanced by direction filtering, i.e., extraction of a symmetric angular segment in the frequency domain centered on a zero frequency. The final image is a normalized sum of inverse FFT's of such segmented spectra. We needed a method yielding a single number assigned to the structure, e.g., the ratio of the area of microtubules to the total cell area. Assuming that the image background intensity is constant, we can use thresholding to detect areas occupied by the cells. The information about the area of the microtubules is contained in a wide range of higher intensities. Therefore, we use a gamma correction. The area occupied by microtubules is then considered an area with intensities above the selected threshold. There were tested three different filters to extract information about microtubules. The mathematical method chosen seems sensitive enough for quantitative assessment of changes of the microtubular network. Copyright 2005 Wiley-Liss, Inc.

Institute of Mathematics Department of Computer Graphics and Geometry Faculty of Mechanical Engineering Brno University of Technology Czech Republic