Pipeline Validation and next steps

Hello,

After adjust this pipeline (in attach)for this kind of images I need to validate it. What can you advise me to do?
For example, for counting objects I will ask for some people to do it manually, and compare results.

OTHER ISSUES
I check the option “Discard Objects touching the border” but it looks to be counting it in the top of the image. Why?
Why the module “MeasureObjectSizeShape” and “MeasureObjectIntensity” are not exported to a csv file? In fact I cannot see all the data in the displayed window!

In the displayed window “IdentifyPrimaryObjects” there’s a table with some info; what is the “Area Covered by Objects”? Is it different from the Module “Occupied Area” right?
What measured features should I relate in order to extract info about Cell Shape?
Can I monitor Cell position variation along a sequence of images, right? How?

After validate my pipeline I need to go forward: Data Storage, Image Archiving, Analysis. I’m thinking to use CPAnalyst. I saw some of the videos, but I’ve some questions about “HOW TO START”:

a) Setup MySQL Database
This should be done on my computer or in a server (at my research center)?
I would like to learn it How to do it, as I have some basic knowledge on this and I would like to Improve it :smile: (is it a difficult or sensible step?)
In case of ask for help at the I&T department what info should I give them?

b) Manage Data (Images)
After setup MySQL database where should the images be saved and managed?

c) Run CPAnalyst
What about the properties file? I read in the tutorial that the CP Module “ExportToDatabase” will have nearly complete properties. But does this module export a properties file?

d) Data Analysis
How do I extract the most accurate values of each image for each feature? Should I use all the output data from CellProfiler and manage it with CP Analyst? I mean, we can obtain some outliers when measuring each cell feature in each image, thus we would need to “eliminate” some cells to not be considered in some feature measurement.

With Best Regards,
Ricardo M.
Pipeline4.zip (393 KB)

Hi Ricardo,

[quote=“Ricardo”]After adjust this pipeline (in attach)for this kind of images I need to validate it. What can you advise me to do?
For example, for counting objects I will ask for some people to do it manually, and compare results.[/quote]

While time-consuming, this is the most reliable means of comparing CP results against ground truth.

You are using a per-object thresholding method even though you are not masking the image with objects. However, even so, the edge objects should excluded anyway, so this behavior is a bug. The workaround is to use Otsu global instead.

The per-object measurements created by these two modules are exported to the per-object .csvs of that object’s name; the module window just shows the per-image aggregate statistics for display purposes, but these are not exported unless you request them (and it looks like you requested the mean) and you export a per-image .csv (which is not the case for your pipeline).

The % area statistic in IdentifyPrimaryObjects is shown for display purposes, and is not actually recorded. However, MeasureImageAreaOccupied will get you the same measurement, although if you want a percentage, you’ll need to include a CalculateMath module to explicitly compute it as a new measurement.

MeausureObjectSizeShape can be used to obtain cell morphology measurements. These measurements are exported in the per-Cell .csv.

If you are referring to object tracking, you can use the TrackObjects module to do this.

[quote=“Ricardo”]a) Setup MySQL Database
This should be done on my computer or in a server (at my research center)?
I would like to learn it How to do it, as I have some basic knowledge on this and I would like to Improve it :smile: (is it a difficult or sensible step?)
In case of ask for help at the I&T department what info should I give them?[/quote]

You can use the SQLite database option in ExportToDatabase to create a database of your measurements on your computer, without requiring a remote server to be set up by someone else. Very convenient!

The images can be placed wherever you like. The per-image measurements saved to the database include the file name and path to each of the input images, so as long as you don’t move them after creating the database, you should be fine.

Correct, ExportToDatabase will create a properties file for you to be used as input for CPA. The module will attempt to fill in as many fields as it can based on the information available.

[quote=“Ricardo”]d) Data Analysis
How do I extract the most accurate values of each image for each feature? Should I use all the output data from CellProfiler and manage it with CP Analyst? I mean, we can obtain some outliers when measuring each cell feature in each image, thus we would need to “eliminate” some cells to not be considered in some feature measurement[/quote]

The answer to this depends on what features characterize a cell that you want to discard. If there is a specific measurement that can be used to identify a “good” cell from a “bad” cell, you can use FilterObjects to get rid of the “bad” cells. If you can then create a set of “good” nuclei, cells and cytoplasm, can then export just these objects to the database, and need not worry about the other objects.

The alternative is to use the machine-learning tool in CPA to discriminate between the cells you want to keep and those you don’t. If you are already decided on using a database to store your measurements, this may be an option to explore. Also, keep in mind that we have a separate forum for CPA-related questions if you get stuck.

Regards,
-Mark