Thin Layer Chromatography

KNIME Analytics Platform
1

Hello,

Is there a node or has anyone developed a workflow to calculate Rf values from a jpg/png/bmp... image of a TLC?

If you are interested, here is a description of the simple calculation performed:

http://www.uwplatt.edu/chemep/chem/chemscape/labdocs/catofp/chromato/tlc/tlcq.htm

 

Any ideas on how to develop a workflow to do this?

 

Cheers,

Fred

2

Hi Frederic,

should the detection of the spots on the image also be part of the workflow? If yes, maybe you could try the KNIME Image Processing Plugin (http://tech.knime.org/community/image-processing) to detect the spots and measure there centroids and afterwards the Math Formula Node to make your calculations?

If you like, you could provide us one example image and we can help you with the spot detection (if possible).

Cheers,

Christian

3

Hello Christian,

 

Thanks for your reply.

 

Yes, the detection should be part of the worklfow. I had never used any of the image-processing nodes so would definitely appreciate your help here.

I am using the Image Reader to import a png. In the future, the input file could be a jpg or a png if it's a picture of a TLC taken from a camera or a scan, or a bmp generated from a third party software.

Please, find attached the file I'm using and an example of a bmp (you will have to remove the _.jpg extension as it was added to be able to upload the file).

Regarding the png file, two spots should be identified in the left column.

 

Cheers,

Fred

 

4

Hi Fred,

I'm not sure if I understand correctly, but in the .bmp image the spots als should be detected? Is not possible that the image is provided without additional information like axis, Rt etc?

For the .png image: In the left column I can hardly see two spots. But you mean the thing slightly above the very obvious spot on the left of the two obvious spots?

Cheers,


Christian

 

5

Hello Christian,

 

You are correct, all spots should be detected in the bmp image. However, no text will be present, only the spots, the base line and the max line.

 

You are also correct about the png, the two spots are the obvious one and the other one that can hardly be seen. However, this is an extreme case: the spots usually are easier to see. This was just to test the detection of the obvious and of the more difficult spots.

 

Thanks,

Fred

6

Hello Christian,

 

I've managed to extract the four spots on the real TLC I posted. What do I do from there?

Image Reader>Max Homogeneity>Global Thresholder (Threshold value=145).

 

Cheers,

Fred

7

Hi Fred,

I attached a workflow, where I do the analysis after global thresholding.

Some restrictions:

  • I cropped the input image (see settings of imgcropper), because we can't detect the spot which you can hardly see by eye.
  • The analysis works for arbitrary number of columns, as long as you detect at least two spots per column
  • You need to install the following packages: DistanceMatrix and MathFormula from the KNIME Update-Site
  • I suggest to update to the nightly build of knime image processing, just as we make use of some of the ewer dialog components (e.g. image cropper), which will be released stable pretty soon

The workflow works as follow: We used your workflow (Max Homogenity and GlobalThresholder) to detect the spots, followed by a CCA to identify the invidiual connected components. We calculate some features on this labeling afterwards (centroid x, centroid y and numberofpixels) and filter the segment by size, such that we only have the big spots in the end.

Afterwards we check with the similiarty nodes for spots, which are close to each other on the x-axis and simply calculate the difference in the y-axis on them.

I hope this helps, if you have any more questions we would be happy to help you.

Cheers,

Christian

 

 

 

 

8