Deformable trellis: open contour tracking in bio-image sequences

Abstract

This paper presents an open contour tracking method that employs an arc-emission Hidden Markov Model (HMM). The algorithm encodes the shape information of the structure in a spatially deformable trellis model that is iteratively modified to account for observations in subsequent frames. As the open contour is determined on the trellis of an HMM, a dynamic programming procedure reduces the computational complexity to linear in the length of the structure (or contour). The method was developed for tracking general curvilinear structures, and tested on subcellular image sequences, where microtubules grow, shrink and undergo lateral motion from frame to frame. Microtubule length changes are modeled by the addition of appropriate transient and absorbing states to the HMM. Our results provide experimental evidence for the proposed algorithm's capability to track non-rigid curvilinear objects in challenging environments in terms of noise and clutter.

Extracting Dynamic Microtubule Features from Image Sequences

“…In this section, we discuss two tracking methods: [3] uses active contours for extracting the MT length as open ended curves, and [10] uses a hidden Markov model based approach for tracking the deformation of a curve after initially tracing it on the first frame. From the usage point of view, the two methods provide different user interaction stipulations.…”

“…Therefore, it is sensitive to frequent variations of intensity caused by additive fluorescence in crowded areas. The method discussed in this section, [10], addresses this issue by using an exploratory tracing algorithm which constraints a search space around the MT tip.…”

Variable Length Open Contour Tracking Using a Deformable Trellis

“…HMM provides globally optimal solutions for contour adjustment with the Viterbi decoding which is based on dynamic programming. This critical advantage was exploited in our preliminary work to perform open contour tracking using a deformable trellis with ad hoc parametrization [23]. This paper subsumes [23] and provides a comprehensive method including model parameter learning schemes and overall optimization of the probabilistic model without recourse to ad hoc or manually tuned parameters.…”

“…In the same table, we further compare the accuracy of the HMM-based approaches with three representations for the transition matrix. In the Gibbs and tilted-Gibbs representations, the transition probabilities are modeled with a Gibbs distribution (23) and a tilted-Gibbs distribution (15), respectively. Finally, in the Viterbi learning-based representation, the transition probabilities are estimated by counting the occurrences of the state transitions on the training data.…”

Evaluation and benchmark for biological image segmentation

“…Due to the limitations in biological sample preparation and fluorescence imaging, typical images in live cell studies exhibit severe noise and considerable clutter and automatic microtubule tracing becomes a hard task. Our benchmark includes an automatic method [5] for extracting curvilinear structures from live cell fluorescence images. The data also include ground truth for microtubule tip location and microtubule bodies that could be useful for evaluating image segmentation and tracking meth- ods.…”