Velocity and feature estimation of actin filaments using active contours in noisy fluorescence image sequences

We present a new approach for determining particle fea tures such as length, curvature and hence flexibility in addition to the velocity of moving single actin filaments in noisy fluorescence image sequences. The reliable deter mination of these features is essential for the analysis of the elementary force generation process of single motor molecules including heart and skeletal muscle myosins. First, the image sequence is preprocessed with the 3D structure tensor - where the third dimension is the time t in the image sequence - in order to eliminate noise and to obtain a measure for extracting coherently moving par ticles. Secondly, we determine the contour of the actin filaments with subpixel accuracy using active contours. Thereafter, we readjust a local coordinate system to elim inate inner movements of the active contour. In the fourth step, we estimate the initial position of the active contour in the next frame from the displacement vector field cal culated by the 3D structure tensor. The accuracy of the method is verified on synthetic test data, a prerequisite for the quantitative use of this method on experimentally obtained data. Finally this is demonstrated for fluorescence image sequences of actin filament movement in the in vitro motility assay.