Related Experiment Video
Updated: Jun 27, 2026
Quantitation of Protein Expression and Co-localization Using Multiplexed Immuno-histochemical Staining and Multispectral Imaging
Published on: April 8, 2016
A comparison of pixel intensity-based and object-based image analysis software algorithms for assessing
Sarah E Kimambo1, Josh Overton2, Nicole A Bouffard2
1Department of Pathology and Laboratory Medicine, Larner College of Medicine, University of Vermont, Burlington, VT, 05405, USA.
Insights
Computer-aided pathology software shows potential for analyzing immunohistochemical (IHC) markers in cancer samples. Object-based analysis offers more consistent results than pixel-based methods, though neither fully meets comprehensive assessment needs.
Area of Science:
- Digital Pathology and Computational Analysis
- Oncology and Histopathology
- Biomarker Quantification
Background:
- Histopathological diagnosis relies on expert interpretation of immunohistochemical (IHC) markers.
- Computer-aided detection (CAD) systems are emerging as decision support tools in pathology.
- Automated analysis software must capture sample information for complex IHC marker interpretation.
Purpose of the Study:
- To technically assess two common automated analysis approaches: positive pixel count and cell-by-cell.
- To compare the performance of these algorithms in commercially available digital pathology software.
- To evaluate their utility in quantifying IHC marker positivity in tumor samples.
Main Methods:
- Analyzed 37 whole slide images of immunohistochemically stained breast, colon, and endometrial tumor samples.
- Utilized two software platforms (ImageScope and HALO) as proxies for pixel-based and object-based analysis.
- Employed three sampling methods to record percentage of antibody marker positivity.
Main Results:
- Pixel-based software excelled at identifying color intensity and offered grading options for IHC markers.
- Object-based software demonstrated superior consistency in positivity estimates across different sampling methods.
- Neither software's metrics provided comprehensive IHC marker assessment as required by the study's criteria.
Conclusions:
- Object-based analysis shows promise for more consistent IHC marker quantification in digital pathology.
- Current automated analysis software can aid in specific quantitative expression questions for tumor markers.
- Further development is needed to meet the demands of comprehensive histopathological assessment.
Abstract:
Histopathological diagnosis relies on careful and expert assessment of tissue as guided by multiple criteria relevant to specific immunohistochemical (IHC) markers. Computer-aided detection or diagnosis systems have recently been deployed to detect abnormalities in histological samples, transforming many areas of research and medicine such as pathology. These software packages can provide a helpful decision support tool for accelerating analysis, but they would need to capture information from the sample in a manner that facilitates the multicriteria assessment/interpretation demanded by the IHC markers and other histochemical stains. As a result of this potential, and the limited assessment of the performance of software utilized for automated analysis of histological samples, we conducted this study. We aimed to provide a technical assessment of two analysis approaches that are utilized in two commercially available image analysis software platforms, namely positive pixel count analysis approach and cell-by-cell analysis approach. These two approaches are used in many digital histopathological slide analysis software packages including ImageScope (Leica Biosystems) and HALO (Indica Labs), which respectively deploy the aforementioned algorithms and thus were used as proxies for the comparison in this study. Thirty-seven whole slide images of immunohistochemically stained tumor samples from breast, colon, and endometrium were analyzed using three different sampling methods recording percentage of antibody marker positivity. The pixel-based software was better able to identify color intensity, offering the option for grading the IHC marker. However, the object-based software outperformed pixel-based software, having more consistent positivity estimates across the three sampling methods. These results are limited by the small number of clinical samples, IHC marker heterogeneity, and the lack of ground-truth data. Nonetheless, neither of the software packages' metrics performed in a manner required for comprehensive assessment of the IHC markers in this study, yet they can be used to address specific questions related to quantitative expression of tumor diagnostic markers.
