Advanced Imaging


Advanced Imaging Magazine

Updated: January 12th, 2011 09:49 AM CDT

Imaging in Cancer Outcome Predictions

New techniques for quantitative analysis of biomarker images at the molecular level
Definiens Life Sciences
An immunoflourescence multiplex pseudo-color image. (a) DAPI (blue) and CK18 (green) in pseudo-colors; (b) segmented epithelial (blue) and stroma (pink) nuclei, CK18+ (green); (c) CK18+ AMACR+ (red), CK18+ AMACR- (cyan) objects superimposed on the AMACR image; and (d) AR+ (red) and AR- (cyan) epithelial nuclei objects superimposed on the AR image.

By Mikhail Teverovskiy, PhD
Definiens Life Sciences

Protein biomarkers are widely used in histopathology for cancer diagnosis and therapeutic response prediction. They provide information on expression levels of proteins in the cells, allowing for the detection of particular cell activities associated with the disease state. Their role has been significantly amplified since image analysis and machine learning methods have been used to create multivariate statistical models for personalized predictions of cancer outcomes. These models combine clinical and pathological variables with quantitative features extracted from tumor-containing areas of imaged tissue sections or tissue microarray cores (TMA). Methodologies that integrate data from different sources into a mathematical model are known in life sciences as “systems biology.” 1, 2

The most common outcomes measured in biomarker diagnostic assays are 4 : a) cancer susceptibility, b) cancer recurrence, and c) cancer survivability. Cancer susceptibility refers to an estimation of the likelihood of cancer developing before the disease is diagnosed. When addressing cancer recurrence the goal is to estimate the chances of the disease relapsing over a five or 10 year period after treatment. Cancer survival or clinical failure models estimate the likelihood of metastasis or death caused by cancer.

Cancer biomarkers are identifiable, molecular level features. One of the key factors in developing successful predictive models is the selection of prognostic biomarkers, which are strongly associated with the disease progression. Another important factor is accurate and consistent quantitation of protein expression. Combining these two elements together guarantees accurate results.

Two prevailing techniques for quantitation of protein expression are immunofluorescence (IF) and immunohistochemistry (IHC). In both cases, a protein is localized by inducing an antibody labeled with either a fluorescent dye (IF) or with a special chromogen (IHC) into the tissue, where it binds to a target antigen.7 The stained IF slide is illuminated under a fluorescent microscope by a light with specific wavelength8 and the fluorescent dye emits light of a longer wavelength. The intensity of the emitted light is a measure of the concentration of the target protein. IF images are gray scale images with pixel intensity proportional to the radiance at a given point. In the IHC, a chromogen produces a colored reaction when the antibody binds to the antigen. Most common chromogens impart a brown, red, or blue color with saturation proportional to protein concentration. Unlike IF, IHC images are color images where positive and negative objects have different color (i.e. brown and blue nuclei).

In IF multiplexing, the tissue is labeled with several antibodies at one time, each paired to a fluorescent dye with distinct spectral characteristics. Separation of multiple biomarkers is accomplished via multispectral imaging of the tissue followed by spectral unmixing to obtain gray scale images that represent the expression of individual antibodies. In IHC multiplexing, several antibodies are stained in a single slide. In this case, the antibodies are labeled with chromogens producing distinct staining colors. Multiplexed IHC image is a color image where individual antibodies are distinguished by their staining colors.

1 2 3 4 5 6 next

Subscribe to our RSS Feeds