It may be valuable to explore the olympus resource for image analysis since many principles of image analysis are discussed within this on-line resource

http://olympusmicro.com/primer/resources/imageanalysis.html

When determining the mean gray value in an image, a boundary area is generally defined around a center pixel and the mean value of gray scale is determined within this area over a sub-set of pixels. The gray scale area used to average gray scale may not extend to the true edges of the image itself. The gray scale image pixel count therefore may not correspond in full to the image area pixel count. 

You are also selecting a gray scale "threshold" to define area for protein expression (using threshold--> binary image--> analyse particle--> summary= total area pixels) that is very likely different from the mean gray scale.

Try cycling your value for mean gray scale calculated from the image into the threshold value that you are using for defining area for protein expression and look at the correlations coming from the analysis. 

Image analysis can be quite an art and practice makes perfect. Does the freeware ImageJ suite have on-line primers and suggestions for "tweaks" and best practices to use in certain circumstances?

https://imagej.nih.gov/ij/docs/examples/index.html

Hi Stephen, thanks for the reply is there a good way of directly selecting a protein expression regions directly from an original or inverted image other than creating an ROIs election for each staining we see, what I did was used the roi from the threshold image create selection and applied it to the inverted gray scale image

How bad or good choice did I do?

Your definition of the RIO was a good one, your grayscale average is therefore derived from the RIO of the image for your region of interest. 

Did the grayscale average return values for min and max grayscale and mean and median grayscale. You can generally cycle an analysis a second time by selecting the min value obtained from your analysis as cut off value for defining the ROI's, track the values you are obtaining from analysis of each protein expression region and average these in excel spreadsheet. You can generate a specific image analysis method for your samples this way and build a validated test method that applies to your analysis at your laboratory. generating your own criteria for a specific analysis will help you in future image analyses your laboratory will do. 

Image analysis is highly subjective and biased to the operator. Building a set of criteria for applying to your protein expression ROI's will assist you alot

I think that the density measurement portrays how densely a protein is concentrated in an area and the %area is how widely spread the protein is

if I am not wrong