Hi Feng Li,

I have done the analysis of amount of protein in plasma membrane in terms of fluorescence using ImageJ. You have to subtract the background and then measure the fluorescence after thresholding it. You get a fluorescence value. I have not done by counting the punctaes. May be you can try this method if you are comparing two groups.

Regards,

Ekta

hi Ekta, 

Can you explain it more exactly?

Hello Feng Lin,

I think I know what you want but if you could maybe post an example Image, I could try to find a suitable solution for you. I think if you want to compare groups then you need a normalization.

Martin

Hi Martin,

here is a image. Could you help me have a look at it? I just want to measure density rather than comparing.

Cheers,

Feng

As I already told you via mail, here a summary for a step-by-step solution how i would do it:

1) Open the image and select it

2) Optional: Apply a “Gaussian Blur Filter”

3) Subtract the background

4) Select cells and add each cell one-by-one to the ROI manager

5) Make sure your measurements in the Results Table show the “IntDen” (integral density)

6) Select all ROIs (=cells) in the ROI Manager and measure

7) Copy or save the Results Table from ImageJ

When reading your question again, I realize that I may have misunderstood it.

In order to have a good representation of your tubules you probably would want to consider making maximum z-projections of Image stacks from your fluorescently labelled cells.

What you need to do is to segment the tubules by using a proper threshold, which can be very tricky due to unequal signal intensities. However, there may be some tricks  including image filters, image calculations, or local contrast enhancements that could come in handy. I recommend to use the otsu thresholding method which is implemented in imageJ. Then you could see how accurate you can segment your tubules.

You also may have to apply different thresholds for each cell separately in order to obtain accurate binary Images. From the binarized Images you could measure the area of your thresholded signal.

If you calculate the area of the whole cells and divide it by the area of the thresholded images you get some kind of a (density) ratio.

However, I have the impression that this method will give you very inaccurate and probably non-reproducible results because of the segmentation process.

Alternatively you could also represent the tubules as topological skeletons using your segmented Images. Then you could try to estimate how many tubules there are. This may be more accurate.

The best results, if you don't need high troughput data, you will get by manual countings or manual selections as suggested. However, I understand your concern of the bias caused by the experimentators' expectations.

If you want to give the skeleton method a try, let me know then I can explain it in more detail.

Martin