A new hybrid Bayesian-variational particle filter with application to mitotic cell tracking

Abstract

Tracking algorithms are traditionally based on either a variational approach or a Bayesian one. In the variational case, a cost function is established between two consecutive frames and minimized by standard optimization algorithms. In the Bayesian case, a stochastic motion model is used to maintain temporal consistency. Among the Bayesian methods we focus on the particle filter, which is especially suited for handling multimodal distributions. In this paper, we present a novel approach to fuse both methodologies in a single tracker where the importance sampling of the particle filter is given implicitly by the optimization algorithm of the variational method. Our technique is capable of outlying nuclei and tracking the lineage of biological cells using different motion models for mitotic and nonmitotic stages of the life of a cell. We validate its ability to track the lineage of HeLa cells in fluorescence microscopy.

Cell tracking using particle filters with implicit convex shape model in 4D confocal microscopy images

“…We address tracking during the nonmitotic stages of cells. Mitosis detection is a separate complex problem [5, 6] that is not addressed in this paper. The ground truth consists of a rough estimate of the centroid of 14 lens cells and 19 retinal cells across 227 time frames.…”

“…Particle filters infer the hidden states of a dynamic system x 1: t = { x 1 , x 2 ․… x t }, using a sequence of noisy observations z 1: t = { z 1 , z 2 ․… z t } [4]. Particle filter methods, along with variational methods, have been used to track spherical and elliptical objects [5, 6, 7, 8]. In variational methods, the hidden states are estimated by optimizing a cost function that depends on the observation using standard optimization algorithms [9].…”

Large‐scale tracking and classification for automatic analysis of cell migration and proliferation, and experimental optimization of high‐throughput screens of neuroblastoma cells

“…However, this requires that corresponding objects in subsequent frames overlap. For applications with large interframe displacements where corresponding objects do not overlap (i.e., low temporal resolution) such methods have been combined with other methods to determine correspondences, for example, probabilistic methods such as particle filters or interacting multiple models filters . Also in combination with other methods for cell localization probabilistic approaches (e.g., Kalman filters ) have been applied.…”

“…An important requirement for successful cell tracking is the robust detection of cell divisions. Mitosis detection is either directly integrated into the tracking approach or performed as a separate step, e.g., based on morphological features or event classification techniques . After cell tracking the cellular migration behavior can be analyzed in detail, for example, based on trajectory classification .…”