Quantification of histological stain

Hello everyone,

I am a research student and taking initiative to do some quantitative histological stain analysis with Fiji-imageJ. I have performed histological stain- masson trichrome with aniline blue (that demonstrates) collagen fibres as blue. My aim is to determine the density of the collagen present in a defined ROI. I have been able to set scale, RGB colour space conversion, colour image image deconvolution using ‘masson trichrome’ vector. Now my challenges are as follows:

  1. Is it necessary to colour ‘Threshold’ for quantification as there is already colour separation during image deconvolution?

2.In selecting a ROI to measure area, IntDen and RawInDen, am I to say IntDen=colour intensity per that selected area=density of fibres in that selected area? EG 1 Area=17667.98 Mean= 130.95 StdDev= 61.53 Min7 Max=245 IntDen= 2313621.99 RawIntDen= 80539752.00

  1. How can I deduce the density of collagen fibres from the data obtained if my scale set was at 5.87pixels/um.

Please do help as I am a beginner to imageJ analysis but finding the software interesting and very useful.
Thanks
Nick

0123141708963.tif (17.1 MB)

@gabriel Might have some thoughts on this, but I have only ever used MTC to measure area, not stain intensity.
Regarding the built in vectors:
" Using the built-in vectors ‘blindly’ is not a good idea because recipes for histological stains often vary, involving alternative chemicals or procedures that result in colour variations."

If you have separated your image by color vectors, they should be effectively a gray scale image, measuring intensity only along a color axis.

I don’t see how you could estimate fiber density from a stain like MTC. The stain itself is variable, which is why you are often best off using your own stain vectors.

This paper notes some steps that should be taken if using MTC for quantitation.

1 Like

Thank you for your kind suggestions.I have spent a week on working out the process of measuring intensity of collagen fibres with MTC using 6 different criteria on imageJ. Upon careful optimisation using the same image, same scale bar and exact area selection, conversation of image to 3 grayscale images: image>colour>split channels and by using optimised vector for MTC colour deconvolution, applying threshold yields the similar result. Background noise was completely eliminated and no IntDen and RawIntDen values were given. See below
Split channels results in pixels
Area: 1250
IntDen: 198780.32
RawIntDen: 6686970

Colour Deconvolution with MTC vector
Area: 1244
IntDen: 197725.96
RawIntDen: 6728175

Both minimum and maximum values for those were 0 and 255 respectively.

Thanks for your help
Nick