Optical microscopy is widely used
to quantify single cell characteristic such as cell size or intracellular
density. Accurate quantification is highly dependent on the cell segmentation
results in the microscope image. The cell segmentation algorithms do not
converge to a single solution with good performance and is developed with
various algorithms depending on the characteristics of the target cell. The
imaging hardwares and analysis software platforms 1 have developed rapidly,
but such cell segmentation studies have been relatively lagging behind.
Cell segmentation is challenging
due to the following three reasons. First, various experimental configurations,
such as cell types or imaging protocols, produce images with different shapes
or brightness characteristics. Second, since cells generally have dynamically
changing shapes over time, we can not mathematically define the cell shape. Third,
the boundaries of some cells that are in contact with each other during cleavage
or migration may be unclear, and experts may have different opinion whether one
cell or more cells are connected.
In order to overcome the various brightness problem of cells
appeared by the image condition, representative image binarization methods such
as Otsu method 2-4 or Watershed transformation 5 were improved to
brightness-invariant localization or adaptive binarization. These methods are
simple to use without any additional parameters, but it is difficult to expect
good performance in complicated backgrounds or splitting overlapping or
adjacent objects. Energy-minimization based image segmentation techniques show
better results than the intensity-based techniques in the above-mentioned
difficult environmental conditions. ACM (Active Contour Model) 6-9 is a
representative energy-minimization technique that generates appropriate results
on noise images based on initial points defined by a user. GC (graph cut) 10-14,
another segmentation method based on energy minimization, finds a global
optimal solution for a given initial value.
Machine learning-based methods typically show more than a certain
level of segmentation performance in various datasets 15-19. Especially,
unsupervised-learning based cell segmentation method using blob detector
produces boundaries similar to those perceived by humans 2021.
Conventional cell segmentation studies usually consist of seed
detection to find the approximate location of a cell and cell split to divide the
region. In the conventional studies, due to the structure in which the result
of seed detection affects the accuracy of cell division, precise seed detection
must be preceded. Furthermore, since these techniques require a large number of
parameters and the parameters should be appropriately selected depending on the
type of the target cell and the imaging condition, the segmentation results are
sensitive to the parameter configuration.
In this paper, we propose a cell segmentation method using a cell
region discriminator R that detects a cell region and a multi-cell
discriminator M that determines whether a cell region is divided by an
Expectation-maximization algorithm (see Fig. 1). R identifies regions of
interest for cells using linear regression analysis and features of statistical
cell imaging and distribution characteristics for image brightness. The region
of interest (ROI) is divided into two cells using expectation-maximization to
the coordinates of the detected ROI and their local maximum point coordinate
feature. In the process of dividing the region, M determines whether to
re-segment the region by finding a hyperplane for the surface error in the cell
area and the area of the segment boundary. The research has the following two
? The proposed method does not require seed detection because it
divides the cell independently of the seed detection.
? Using various learning techniques trained from each image data, ROI
detection and cell division show high accuracy without changing parameters
according to data.