Microscope Image Processing

of interest, the objects to be tracked have widely differing and even time-varying

appearances. As a consequence, full automation usually can be achieved only by

developing very dedicated algorithms that are tuned for the specific application.

This explains why existing commercial tracking software tools, which are devel-

oped for broad, general applicability, often fail to yield satisfactory results for

specific tracking tasks.

In this section we discuss published approaches to automated object tracking

in time-lapse microscopy images. A distinction is made between cell tracking and

particle tracking. Two different strategies exist for both problems. The first

consists of the identification of the objects of interest in the entire image

sequence, separately for each frame, followed by temporal association, which

tries to relate identified objects either globally over the entire sequence or from

frame to frame. In the second strategy, objects of interest are identified only in

a first frame and are subsequently followed in time by image matching or by

model evolution. In either case, the algorithms usually include a detection or

segmentation stage and a temporal association stage. Both are essential to

performing motion analysis of individual objects. Alternative approaches

based on optic flow have also been studied [17, 19, 40, 41], but these are limited

to computing collective cell motion and intracellular particle flows, unless

additional detection algorithms are applied.

15.4.1 Cell Tracking

Cell motility and migration are of fundamental importance to many biological

processes [4, 10, 42, 43]. In embryonic development, for example, cells migrate

and differentiate into specific cell types to form different organs. Failures in this

process may result in severe congenital defects and diseases. In adult organisms

as well, cell movement plays a crucial role. In wound healing, for example,

several interrelated cell migration processes are essential in regenerating dam-

aged tissue. The immune system consists of many different proteins and cells

interacting in a dynamic network to identify and destroy infectious agents.

Many disease processes, most notably cancer metastasis, depend heavily on

the ability of cells to migrate through tissue and reach the bloodstream. Because

of its importance for basic cell biology and its medical implications, cell migra-

tion is a very active field of research. Automated methods for segmenting cells

and following them over time (Table 15.1) are becoming essential in quantifying

cell movement and interaction under normal and perturbed conditions.

15.4.1.1 Cell Segmentation

The simplest approach for separating cells from the background is intensity

thresholding (see Chapter 9). This involves a single threshold parameter that can

415

15.4 Image Analysis