Page 8 - Live-cellanalysis handbook

Live-Cell Analysis Handbook — Third Edition

Performing these steps on individual images to generate sufficient to assess differences, and hypotheses evaluated. Scaling this to
statistical power to support a hypothesis can be a tedious the analysis of live-cell experiments allows for the evaluation of
process. However, when operating on large numbers of images temporal data, and extending this to microplate microscopy means
which have been collected in a substantially similar manner, the that population data may be studied with ease. This basic workflow
series of operations performed to clean up the data, extract is the subject of countless tutorials and books, and the domain
desired information, and compare images may be recorded and of numerous software packages that offer a cornucopia of tools
automatically applied to many images in a single experiment. Once intended to answer a broad range of scientific questions.
this data has been extracted, treatment groups may be compared

Image Processing to Remove Systematic or Sample-Induced Artifacts

The image data we have described above is typically captured by or non-biological signal sources (i.e. shading or patterns arising
detectors that convert analog information, specifically photons, from sample matrices, micro-fluidic channels, or non-uniform
into digital signals. This analog information is collected in a matrix illumination effects in microwells) must be removed before usable,
fashion, spatially rendered according to location in the sample. replicable information can be extracted.
Ideally, the signal undergoing analog to digital conversion would
come only from photons produced by the sample of interest, In order to perform these corrections, one must be aware of the
and in perfect focus. However, this is not the usual case. There effects of each process, and manipulations on the raw images
are multiple sources of confounding signal present in an image, must be repeatable to ensure faithful capture of the true biological
each needing correction, removal, or cleaning in order to reveal signal across images. There are many tutorials and software
information which has been generated by the sample elements of toolkits available to process images, however systems that perform
interest. Corrections are needed due to systematic aberrations in these corrections as a matter of course provide consistency and
an imaging system stemming from multiple sources. For example, ease of use, particularly when coupled with standardized assays,
detector anomalies (e.g. detector bias, dark current variability, reagents and consumables which normalize the experimental
field flatness and thermal or gamma-ray noise), optical issues process (e.g. the IncuCyte Live-Cell Analysis System, and the assays
(non-flat optical components and illumination imperfections) and reagents available from Sartorius). The consistency with which
or undesired signal introduced by the sample are common images are acquired and processed will influence the ability to
issues. Autofluorescence from cellular componenets or media, analyze the collected data.

Identifying Biology of Interest via Image Masking or “Segmentation”

Once an image has been appropriately processed to remove do exist, and more complex interactions can be performed with
aberrant signal, the next step is to identify the biology of multiple masks, and Boolean operations (e.g., AND, OR, NOT) in
interest. Image segmentation is a binary process, meaning pixels order to hone in on the exact pixels of scientific interest. Again,
are classified as either “in” and are included in any enumeration this can be a time-consuming task, and purpose-built software
process, or “out” and not considered as part of the sample. The that presents only the tools necessary for a specific scientific
simplest method for determining which pixels are in or out is by question can remove what can be a significant hurdle in the
thresholding, or setting a boundary above which all pixels are image analysis workflow.
“in”, and below which, all pixels are “out”. More complex tools

6