PR, Breast Cancer PR, Breast Cancer

Process APP




PR, Breast Cancer

According to immunohistochemistry (IHC), progesteron receptor (PR) is used to determine prognosis and as a predictive marker in breast cancer. The American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) recommends that the PR status of patients is determined on all invasive breast cancers and breast cancer recurrences, see [1]. It is, additionally, recommended that the PR status of the tumor is considered positive if there are at least 1 % positive tumor nuclei in a sample.

This protocol can be used to determine PR positivity and negativity in a tumor, and provides the number of positive nuclei as well as the total number of nuclei. In addition, the ratio of positive nuclei and the area ratio of positive nuclei are given. Tumor regions must be identified and outlined manually within a region of interest (ROI). By allowing the user to adjust the sensitivity for the detection of nuclei, and vary the threshold for differentiation between PR-positive and -negative nuclei, the protocol is functional on images with different staining intensities. 


Figure 1

Figure 1

Field of view showing invasive tumor in an image of breast tissue stained by IHC for PR.

Figure 2

Figure 2

The same field of view as in Figure 1 after detection of nuclei as either PR-positive (red label) or PR-negative (blue label). From this classification, the output variables can be calculated.

Figure 3

Figure 3

A field of view including an un-biased counting frame, ensuring that nuclei are counted only once.


Quantitative Output variables

The output variable obtained from this protocol is the H-Score.
The H-Score is calculated from the percentages of nuclei classified as 3+, 2+, 1+ (the three positive categories, where 3+ has the highest staining intensity) multiplying them with their grade:

  • H-score = (Percentage of 3+) x 3 + (Percentage of 2+) x 2 + (Percentage of 1+)

Thus the H-Score is a value between 0 and 300 (0 if there are only negative cells, and 300 if all cells are positive with an intense stain), giving an indication of the ratio of positive cells while factoring in staining intensity.

Furthermore the Positive ratio of ER cells is also calculated:

  • Neg Nuclei (#): The number of negative nuclei
  • 1+ Nuclei (#): The number of 1+ classified nuclei
  • 2+ Nuclei (#): The number of 2+ classified nuclei
  • 3+ Nuclei (#): The number of 3+ classified nuclei
  • Positive Percentage: The percentage of positive nuclei profiles.


The first image processing step involves a segmentation of all nuclei in the ROI (see FIGURE 1 and 2). This is done by assigning a label probability to all pixels in the image, resulting in a label probability image. The label probability image is found by an intensity dictionary – a dictionary with small image patches. The intensity dictionary can be coupled to a label dictionary from which the label probability image is obtained. Based on this image, segmentation of nuclei can be done by choosing the most probable label in each pixel, see [2].  Pixels belonging to both positive and negative nuclei are detected by performing a segmentation based on the red RGB color band in the image. A method for nuclei separation which is based on shape, size and nuclei probability is used, employing a fully automated watershed-based nuclei segmentation technique. The method is an extension of the method in Jung and Kim, see [3] , where an h-minima transform is used before applying the watershed. Next, pixels that contribute to positively stained nuclei are identified based on a DAB color deconvolution of the image, and from this all pixels within each nucleus are classified as belonging to either a positive of negative class. As a post-processing step, nuclei areas that are too small are removed. The image obtained after post-processing (see FIGURE 3) is the basis for quantification of the output variables.

Note on counting: Analysis of full virtual slides takes place in a tile-by-tile fashion. If not handled appropriately, nuclei that are intersecting with neighboring tile boundaries would be counted twice (or more). By using unbiased counting frames, see [4] , this can be avoided (see FIGURE 2). This principle is implemented in the present APP. Depending on the size of nuclei, the application of this principle could make an important difference.


Progesteron receptor, PRstatus, immunohistochemistry, quantitative, digital pathology, image analysis, breast cancer.


1. Hammond, M.E.H, et. al. American Society of Clinical Oncology/College of American Pathologists Guideline Recommendations for Immunohistochemical Testing of Estrogen and Progesterone Receptors in Breast Cancer, J. Clin. Oncol., 2010, 28 (16), 2784-95, DOI

2. Kårsnäs, A., et. al. Learning histopathological patterns, Journal of Pathology Informatics 2011, 2 (2), S12, DOI

3. Jung, C., et. al. Segmenting Clustered Nuclei Using H-minima Transform-Based Marker Extraction and Contour Parameterization, IEEE Transactions on Biomedical Engineering 2010, 57 (10), 2600-4, DOI

4. Howard, C.V., Reed, M.G. (2005). Unbiased Stereology: Three-dimensional measurement in microscopy. QTP Publications.

Your message has been successfully sent!