Using Articulated Models for Tracking Multiple C. elegans in Physical Contact

Abstract

We present a method for tracking and distinguishing multiple C. elegans in a video sequence, including when they are in physical contact with one another. The worms are modeled with an articulated model composed of rectangular blocks, arranged in a deformable configuration represented by a spring-like connection between adjacent parts. Dynamic programming is applied to reduce the computational complexity of the matching process. Our method makes it possible to identify two worms correctly before and after they touch each other, and to find the body poses for further feature extraction. All joint points in our model can be also considered to be the pseudo skeleton points of the worm body. It solves the problem that a previously presented morphological skeleton-based reversal detection algorithm fails when two worms touch each other. The algorithm has many applications in the study of physical interactions between C. elegans.

Multi-Environment Model Estimation for Motility Analysis of Caenorhabditis elegans

“…In practice, features are extracted from binary images, or segmentations , separating the nematode from its environment, or background . Several analysis systems compute binary images by applying a simple intensity-based threshold at each pixel location [30] , [31] , [35] , [36] , [40] , [41] . Most commonly, this involves having the user manually select an appropriate range of intensities which characterizes the nematode.…”

Algorithmic Mapping and Characterization of the Drug-Induced Phenotypic-Response Space of Parasites Causing Schistosomiasis

“…Other strategies for analysis of cellular phenotypes include fitting of elliptical models [46] and statistical combination of manually constructed models [47]. For C. elegans, recent methods include the use of articulated models [48] and probabilistic shape models [49]. Explicit models of size, shape and/or intensity however, are impractical for schistosomula due to the natural variation of individuals (schistosome clones do not exist) and the extreme and unpredictable effects of drug exposure.…”

Segmenting the Etiological Agent of Schistosomiasis for High-Content Screening

“…Methods employing articulated models [8] and path searching on probabilistic shape models [9] have been used for C. elegans, while cells have been segmented by fitting of elliptical models [10] and statistical merging of manually constructed models [11]. However, explicit models of size, shape and/or intensity are impractical for schistosomula due to the natural variation of individuals and the extreme and unpredictable variation induced by drug exposure A number of methods attempt to separate touching objects based on the watershed transform, including region merging [12], [13], iterative erosions for marker localization [14] and minima merging [15].…”

“…Frequency information (energy and phase angle ) is then extracted by comparing responses to the symmetric filter and antisymmetric filter via (6) (7) Log-Gabor wavelets, characterized by a Gaussian transfer function on a logarithmically scaled frequency axis, are chosen because they are psychophysically justified [36], and because they possess zero mean at arbitrary bandwidths. The log-Gabor transfer function with center frequency is given by (8) Analysis at multiple scales is conducted by summing the responses to a bank of log-Gabor filters. This sense of scale, defined over a range by the extent of the tail of the log-Gaussian distribution in frequency space, is attractive because it derives not from the spatial extent of a feature, but from the spatial extent of its constituent frequency components.…”