Quantifying Branching Density in Rat Mammary Gland Whole-mounts Using the Sholl Analysis Method

The overall goal of this computer-assisted method is to objectively quantify mammary gland branching density and branching decay, two important developmental indicators in rodents by applying a modified sholl analysis to mammary gland whole-mount images. This method can help answer questions in the field of mammary gland biology, such as how chemical exposures can alter mammary gland development. The main advantage of this technique is that it utilizes open source software to objectively quantify fundamental characteristics of mammary gland development.

For this procedure, have images of whole-mounts showing the glandular epithelium at a fixed magnification. To begin, open an image in ImageJ. Next, select the freehand tool.

Trace around the glandular epithelium and clear the surrounding area. Then, trace around the lymph node and remove this part of the image with the cut command. If the background of the image is black, it must be converted to white.

Click Edit, then Options, then Colors. In the dialog box, select white from the dropdown menu of the background options. Although the background may remain black, it will be treated as white when processing the image.

Now, separate the color channels into three eight-bit grayscale images. Then, select the channel with the best contrast, which is typically the blue channel. Now, subtract the background from the selected color channel.

Then, proceed with choosing the desired parameters and previewing the changes. Check light background and preview in the dialog box and then adjust the number of pixels until the desired contrast is obtained. If additional contrast is needed, use the Unsharp Mask option.

Check Preview in the dialog box and adjust pixel radius and mask weight until desired contrast is obtained. After applying the unsharp mask filter, remove the border that was created by the filter. Click on the double red arrow button in the toolbar and select Drawing Tools.

Then, choose the Eraser tool. Adjust it's diameter by right-clicking the eraser tool button, then hold down the left mouse button to erase the border. Only one level of Undo is allowed while erasing, so be careful.

After applying one of the filters, remove the remaining noise manually. Refer to a copy of the original image and remove identified noise by first clicking on the double red arrow button in the toolbar and selecting the drawing tools. Then, choose the eraser tool.

Adjust it's diameter by right-clicking the eraser tool button. Then, hold down the left mouse button to erase noise. Only one level of undo is allowed while erasing, so be careful.

Next, adjust the image threshold. Move the sliders to adjust the minimum and maximum threshold values to improve the image of the gland. Then, click Apply.

Remove additional noise automatically using either of two methods. The first option is to use the despeckle command. The second option is to use the remove outliers command.

This process only replaces a pixel with the median of it's immediate surroundings if the pixel deviates from the median by more than a user-defined threshold value. Now, using the original image for comparison, reconstruct portions of the glandular epithelium that were removed by thresholding and noise removal. When reconstructing the image, be sure to do it carefully and on a minimal basis to maintain the integrity of the image.

Select the spray can tool from the drawing tools. Right-click on the spray can tool button to adjust the tools size, then carefully fill in the missing sections of the gland using the left mouse button. Next, create a skeletonized image of the gland for the sholl analysis.

First, put the image into a binary format. Then, skeletonize the image. This repeatedly removes pixels from the edges of the binary image until it is reduced to a single pixel-wide shape.

Now, dilate the image once to close the gaps created by thresholding and skeletonizing. Then, save the image as a new file using the save as command. Finally, check the accuracy of the skeleton image against the original by overlaying the two images.

In the add image dialog box, select the skeleton image and set it's opacity at 30%Then, save the overlay as a new image for future reference. To begin, open a skeleton image and convert the image into a binary format. Now, define the image scale.

Enter the number of pixels per millimeter, then set both the known distance and pixel aspect ratio to one, then enter millimeters for the unit of length and toggle the global option to apply the scale to all the images. Next, determine the ending radius for the sholl analysis using the line drawing tool. Draw a line from the start of the primary duct to the most distal point of the glandular epithelium.

Then press the M key and the distance of the line will be reported. Now, input the settings for the sholl analysis. First, select options in the definition of shells section, set the starting radius to 0.00 millimeters and set the radius step size to 0.1 millimeters.

Set the number of samples to one. The length of the line just drawn is automatically entered in the ending radius field. For the multiple samples per radius options, choose mean for the integration.

For the descriptors and curved fitting options, set the enclosing radius cutoff to one. Choose infer from starting radius for the number of primary branches and select the linear and best fitting degree options. Then, under profiles without normalization, select most informative.

Now, under the sholl methods section, choose area for normalized profile. Lastly, in the output section, select create intersections mask to make a heat map of the intersections. Now, open a preview window of the image with rings and confirm the area of analysis.

If everything looks good, click okay to run the analysis. Running the analysis will create a sholl profile linear window, a sholl profile semi-log window, a sholl results window, which contains all sholl output values and a file name sholl profiles window, which provides the number of intersections at each radius as determined by the polynomial. To measure the MEA, work from the skeleton image of the gland in ImageJ.

Open the polygon tool and draw a polygon around the perimeter of the gland. Use mouse clicks to add line segments. When the entire epithelial area has been circumvented, double-click to close the polygon.

Then, press the M key to open a results window. The value in the area column corresponds to the MEA. If the glandular epithelium is extended beyond the lymph node, subtract the lymph node area from the MEA when calculating the branching density.

The lymph node area can be measured in the same manner as the MEA. Using the described methods, two age matched groups of rats were analyzed. The animals in group B were treated with a chemical known to cause precocious mammary gland development.

Plotting the number of branch intersections along the radius from the gland illustrates the variation in the results. The sholl analysis results revealed a clear difference in the mean number of intersections between the groups. However, the intersection data must be normalized to the mammary gland area to draw conclusions.

The sholl analysis and the mammary gland epithelial area measurements provide two physiological parameters useful for making comparative analysis. Clearly, the size of the glands was much greater in the treatment group. The sholl analysis returns the value for Sum N, which is used for calculating the branching density and the value for k.

In the treatment group, the higher mean branching density and the lower mean, k, which reflects a lower rate of branching decay indicate that the treated animals had more well developed glands than the untreated animals. While attempting this procedure, it is important to remember that the removal of noise is the most crucial step. Removing too much or too little can disrupt the integrity of the gland and skew the results.

When utilizing this method, it's important to begin with a ristine mammary whole mound. An accurate skeletonized image cannot be obtained if there are imperfections, such as tears or folds. Creating an accurate skeletonized image takes patience and practice.

Once mastered, this technique can be done in 15 to 20 minutes per gland analyzed if it is done properly.