Histopathology

This page describes the algorithm and the use of the program for nucleus and whole cell segmentation and nucleus/cellular scale biomarker quantification. The algorithms have been implemented in c++ and IDL and instructions to use the program are also included.

Contents 1 Background and motivation 2 Algorithm 3 User Guide 4 References

Background and motivation

Detecting molecules of interest and quantifying them plays an increasingly important role in modern medical diagnosis and treatment planning[1]. Currently, most systems are visually based qualitative descriptors which suffer from the consequent limitations. There are other algorithms that quantify biomarkers on a pixel-by-pixel basis or a regional basis[2-5]. Segmenting the cells and nuclei allows us to classify the different cell types based on antigens that mark specific cell. This enables us to quantify the biomarker of interest in the relevant cell type. This is a more biologically relevant measure and can be more useful in medical diagnosis.
Reminder: Scale and add images to show multiplexed and unmixed histological samples

Algorithm

The algorithm involves the following steps:

• Segment(delineate) nuclei
  • For this we use the 2-D nucleus segmentation algorithm described here[[1]]
• Delineate cell boundaries
  • If reliable membrane or cytoplasm staining is available then:
    • Binarize the image by graph cuts[6]
    • Create a gradient weighted distance map with the segmented nuclei as the initialization points
    • Use a seeded watershed algorithm with the gradient weighted map as the feature input and the segmented nuclei as the markers and segment(delineate) the cells[7]
  • If neither the cytoplasm nor the membrane staining is reliable then use a purely geometric approach to estimate the cytoplasmic domains:
    • Draw dynthetic boundaries by redefining the dominance region described by Nath,et. al.[8] as
      

where (x,y)is any point in the neighborhood of the nucleus N_i and r_max is the user defined maximum radius of of the synthetic boundaries. r_max is in terms of the number of pixels from nucleus N_i

• Quantify biomarker in the region of interest
• Load training set and classify cells as biomarker positive or negative using a Bayesian classifier

User Guide

As of May 2009 the the program is written as a combination of c++ and IDL. To run the program, you will need the IDL virtual machine which is available online for free[[2]]. Soon, you should see parts or all of the code on the Farsight repository and it should be incorporated into the Farsight build by the end of the 2009 Summer.

A custom GUI was built using IDL. You should be able to use the program with the following screen shots and description. If you can not, sign up to the Farsight mailing list and post your problem there.

References

1. Hammerschmied CG, Walter B, Hartmann A. [Renal cell carcinoma 2008 : Histopathology, molecular genetics and new therapeutic options.]. Pathologe 2008;29(5):354-63.}
2. Camp RL, Chung GG, Rimm DL. Automated subcellular localization and quantification of protein expression in tissue microarrays. Nat Med 2002;8(11):1323-7.
3. Mulrane L, Rexhepaj E, Penney S, Callanan JJ, Gallagher WM. Automated image analysis in histopathology: a valuable tool in medical diagnostics. Expert Rev Mol Diagn 2008;8(6):707-25.
4. Taylor CR, Levenson RM. Quantification of immunohistochemistry--issues concerning methods, utility and semiquantitative assessment II. Histopathology 2006;49(4):411-24.
5. Tamai S. Expert systems and automatic diagnostic systems in histopathology--a review. Rinsho Byori 1999;47(2):126-31.
6. Al-Kofahi Y, Lassoued W, Lee W, Roysam B. Improved Automatic Detection & Segmentation of Cell Nuclei in Histopathology Images. IEEE Trans Biomed Eng 2009
7. Beare R., Lehmann G. The watershed transform in ITK - discussion and new developments. The Insight Journal - 2006 January - June [[3]]
8. Nath SK, Palaniappan K, Bunyak F. Accurate Spatial Neighborhood Relationships for Arbitrarily-shaped Objects using Hamilton-Jacobi GVD. Lect Notes Comput Sci 2007;4522 LNCS:421-431.