ImmunoRatio: a publicly available web application for quantitative image analysis of estrogen receptor (ER), progesterone receptor (PR), and Ki-67
© Tuominen et al.; licensee BioMed Central Ltd. 2010
Received: 9 March 2010
Accepted: 27 July 2010
Published: 27 July 2010
Accurate assessment of estrogen receptor (ER), progesterone receptor (PR), and Ki-67 is essential in the histopathologic diagnostics of breast cancer. Commercially available image analysis systems are usually bundled with dedicated analysis hardware and, to our knowledge, no easily installable, free software for immunostained slide scoring has been described. In this study, we describe a free, Internet-based web application for quantitative image analysis of ER, PR, and Ki-67 immunohistochemistry in breast cancer tissue sections.
The application, named ImmunoRatio, calculates the percentage of positively stained nuclear area (labeling index) by using a color deconvolution algorithm for separating the staining components (diaminobenzidine and hematoxylin) and adaptive thresholding for nuclear area segmentation. ImmunoRatio was calibrated using cell counts defined visually as the gold standard (training set, n = 50). Validation was done using a separate set of 50 ER, PR, and Ki-67 stained slides (test set, n = 50). In addition, Ki-67 labeling indexes determined by ImmunoRatio were studied for their prognostic value in a retrospective cohort of 123 breast cancer patients.
The labeling indexes by calibrated ImmunoRatio analyses correlated well with those defined visually in the test set (correlation coefficient r = 0.98). Using the median Ki-67 labeling index (20%) as a cutoff, a hazard ratio of 2.2 was obtained in the survival analysis (n = 123, P = 0.01). ImmunoRatio was shown to adapt to various staining protocols, microscope setups, digital camera models, and image acquisition settings. The application can be used directly with web browsers running on modern operating systems (e.g., Microsoft Windows, Linux distributions, and Mac OS). No software downloads or installations are required. ImmunoRatio is open source software, and the web application is publicly accessible on our website.
We anticipate that free web applications, such as ImmunoRatio, will make the quantitative image analysis of ER, PR, and Ki-67 easy and straightforward in the diagnostic assessment of breast cancer specimens.
Immunohistochemical staining of the estrogen receptor (ER), progesterone receptor (PR), and proliferation antigen Ki-67 are routinely used in the diagnostic assessment of breast cancer. Positive ER status of a tumor is considered necessary for patients to be eligible for post-surgical hormonal therapies. ER and PR assays are based on immunohistochemistry performed on formalin-fixed and paraffin-embedded tumor tissue blocks . Although the analytical quality of ER and PR assays has been debated for decades, recent results of interlaboratory quality assurance studies provide convincing evidence for the high reproducibility of these laboratory staining procedures [2, 3]. The tumor cell proliferation antigen level, as defined by Ki-67 immunostaining, is an auxiliary tool for defining patient prognosis. Patients with rapidly proliferating tumors are predicted to endure poorer outcomes than patients with tumors exhibiting low proliferation . Meta-analyses confirming the role of Ki-67 as a prognostic factor have included more than 15,000 patients .
Common practice in pathology laboratories is to score ER-, PR-, and Ki-67-stained slides visually (also termed manually) using light microscopy at medium power magnification (10× or 20× objectives). For ER and PR evaluation, a tumor is scored as negative or positive or, as is currently recommended, by evaluating the percentage of positively stained tumor cell nuclei . A threshold of 10% total stained tumor cells is commonly used as a cut-off for defining positive ER and PR status. A combination of the stained cell percentage and the staining intensity is applied in histoscore and Allred-score methods [7, 8]. Whichever scoring method is used, it is well known that microscopic evaluation of ER- and PR-stained slides is subjective and can lead to significant inter-observer variability. For example, in an extensive inter-laboratory study of 172 pathologists, 24% of ER-positive slides were interpreted as falsely negative . Interpretation of Ki-67 staining can be even more difficult, mainly owing to the lack of uniformly accepted cut-off points for defining low- and high-risk patient groups. In most of the published studies included in the meta-analyses, Ki-67 has been evaluated in a single center or by one or very few observers, thereby failing to address the problem of possible inter-observer variability [4, 5]. The magnitude of inter-observer variability for Ki-67 scoring is largely unknown, but there is no reason to believe that it would be less than that of ER and PR.
In clinical practice, ER-, PR- and Ki-67-stained slides are interpreted by a pathologist. Careful estimation of the percentage of positively stained cells (labeling index) is not only prone to inter-observer variation, but is also tedious and time-consuming. To overcome this, various digital image analysis methods have been described [10, 11]. The principles behind quantitative immunohistochemistry analyses are based on differentiation of the staining components by using, for example, the color deconvolution algorithm . The color deconvolution algorithm detects and separates multiple stains by analyzing their absorption spectra and relative contributions to areas containing two or more overlapping stains. Although the optical density of the immunoreaction product (brown diaminobenzidine (DAB) precipitate) may not accurately reflect the abundance of the antigen (ER, PR, or Ki-67 protein), systems discriminating between negatively and positively stained cells have turned out to be useful . Unfortunately, the image analysis software described in the literature is seldom released for public use. Likewise, the commercially available software is usually proprietary and/or bundled with dedicated analysis equipment or virtual microscopy scanners, making it difficult to compare them .
In order to become widely accepted and utilized by pathologists, a digital image analysis system should be easily accessible, not require dedicated equipment or software installation, and be compatible with existing microscope setups. For this purpose, we developed an image analysis application, named ImmunoRatio, which is accessed and used within a web browser. ImmunoRatio supports all modern web browsers and operating systems, requiring no software installation. The application segments immunostained and hematoxylin-stained cellular areas from the user-submitted image and calculates the labeling index (percent of DAB-stained area out of the total nuclear area). ImmunoRatio is free, open source, and publicly available on our research group website .
Materials and methods
Formalin-fixed and paraffin-embedded tissue sections from invasive breast cancers were derived from the archive of the Department of Pathology, Seinäjoki Central Hospital. The study has been approved by the Scientific Committee of Seinäjoki Central Hospital, Finland. According to the Finnish national ethics committee regulations, informed consent was not considered necessary for this study. Immunohistochemical stainings of ER, PR, and Ki-67 tissue sections followed the recommended staining protocols . The slides were stained using the BondMax staining robot (Leica Microsystems, Wetzlar, Germany). In brief, ER was detected using monoclonal antibody 6F11 (diluted 1:300, Leica Biosystems, Newcastle, UK), PR was detected using monoclonal antibody PgR636 (diluted 1:600, Leica Biosystems, Newcastle, UK), and Ki-67 was detected using monoclonal antibody MIB-1 (diluted 1:100, Dako, Carpinteria, CA, USA). Antigen retrieval was performed in Tris-EDTA buffer (pH 9, 100°C for 40 minutes). Bound antibodies were visualized using Bond Refine Detection kit (Leica Biosystems, Newcastle, UK). Immunoreaction was intensified using 0.5% copper sulfate (5 minutes). Hematoxylin counterstaining (1 minutes in ChemMate diluted 1:6, Dako, Carpinteria, CA, USA) was performed using PBS as bluing reagent. The samples were cleared with ethanol and xylene and mounted using standard procedures.
Samples from 123 primary breast cancer patients were derived from the archives of the Department of Pathology at Tampere University Hospital, with the permission from National Supervisory Authority for Welfare and Health (Köninki et al.: Analysis of PIK3CA mutations and protein expression in breast cancer, submitted). Survival rates of all patients were calculated by the method of Kaplan and Meier. Data on breast cancer-specific mortality was obtained from Finnish Cancer Registry. Up to 20-year follow-up was available for this patient cohort (cancers diagnosed between 1988 and 1992). The immunohistochemical staining for Ki-67 was carried out as described above, except that PowerVision+ kit (ImmunoVision, Springdale, AZ, USA) was used for antibody detection and LabVision Autostainer (LabVision, Fremont, CA, USA) for staining automation. Informed consent in very old retrospective patient cohorts was deemed unnecessary, because the study was approved by the local hospital ethics committee and the National Supervisory Authority for Welfare Health.
Digital images were captured using a Leica DM3000 microscope (Leica Microsystems, Wetzlar, Germany) equipped with 10×, 20×, and 40× objective lenses, a 1× phototube, and a Scion CFW-1612C digital color camera (Scion Corporation, Frederick, MD, USA; 24-bit color depth; resolution 1,600 × 1,200 pixels, pixel size 4.40 μm). The images were stored using an uncompressed image file format (bitmap). For every imaging session, an image from empty slide background area was acquired (blankfield image), which was used to correct image color balance and uneven illumination. Optimal image brightness and contrast were determined by using the Camera Adjustment Wizard, which was developed as an incorporated function of ImmunoRatio. The Camera Adjustment Wizard measures the brightness of the blankfield image and performs a contrast analysis using an image containing hematoxylin-stained cells. An optimal brightness (mean gray intensity) of the blankfield image is considered to be in the range of 200 to 250 (available range 0 to 255, black being 0). The contrast analysis segments the hematoxylin-stained cells (foreground) and analyzes their mean gray intensity, which is then divided by the background mean gray intensity. The contrast is considered to be optimal if the foreground mean gray intensity is 50 to 80% of the background mean gray intensity.
From a pool of 100 immunohistochemically stained slides, 50 were selected to be included in the training set (25 stained for Ki-67, 13 for PR, and 12 for ER). The labeling indexes (percentage of positively stained nuclei by visual assessment) were evenly distributed, ranging from 0 to 100%. From each training set slide, one image representative for invasive carcinoma was acquired. Each acquired image was analyzed visually by counting positively and negatively stained carcinoma cells on a computer screen (a minimum of 500 cells total per image). The percent of DAB-stained nuclei out of the total nuclei (DAB- and hematoxylin-stained) was calculated as the labeling index. This result was used as the gold standard for ImmunoRatio calibration. The images were then analyzed using non-calibrated ImmunoRatio, and the results were compared with visual counting in a scatter plot. Owing to the non-linear relation, a third degree polynomial was fitted to the data. ImmunoRatio was then calibrated by embedding the fitted polynomial as a correction function into the analysis algorithm. To validate the calibration and demonstrate the accuracy of the analysis, the remaining 50 samples (25 stained for Ki-67, 13 for PR, and 12 for ER) were used as a test set, which was analyzed using calibrated ImmunoRatio. In the final step of the validation, the minimum number of images needed to be averaged from a typical tumor sample (diameter 1 to 2 cm) was defined. From 10 samples, 12 images per sample representing central and peripheral tumor areas were acquired using 20× objective.
ImmunoRatio was initially developed and calibrated using ER-, PR-, and Ki-67-stained slides, which were considered optimal by an external quality assurance program . To simulate interlaboratory variability in staining results, the effect of suboptimal primary antibody (Ki-67 MIB-1) dilution and hematoxylin counterstaining intensity was studied. The robustness of ImmunoRatio to variations in image acquisition settings was examined by comparing the optical resolutions provided by 10×, 20×, and 40× microscope objectives, and by comparing the analysis results obtained with six microscope cameras: Scion CFW-1612C (Scion Corporation, Frederick, MD, USA), Altra 20 (Olympus Corporation, Tokyo, Japan), ColorView II (Olympus Corporation, Tokyo, Japan), Leica DFC290 HD (Leica Microsystems, Wetzlar, Germany), Mightex 3MP Color CMOS (Mightex Systems, Pleasanton, CA, USA), and Nikon DS-Fi1 (Nikon Corporation, Tokyo, Japan). The primary output file format used in the acquisition was uncompressed (i.e., lossless). Images were also acquired using JPEG file format (quality factors 10, 20, 40, 60, 80, and 100) to study the suitability of lossy compression for ImmunoRatio analysis. In addition, for each camera, the average diameter (pixels per μm) of a hematoxylin-stained nucleus was measured. Linear regression was used to fit a first degree polynomial to the data and the polynomial was then embedded into the Scale Finder function of ImmunoRatio. The Scale Finder assists the user in determining a rough scale estimate for the microscope setup, if not known prior to analysis.
Main features of ImmunoRatio web application
Users can analyze either the whole image or a region of interest.
Hardware and software requirements
Runs within the web browser, requiring no additional program or plugin installations. Is compatible with all modern web browsers and operating systems.
Compatibility with different microscope setups
Adapts to various combinations of microscope objective lenses, phototubes, and camera resolutions.
Supports most existing camera models and image formats (JPEG, JPEG2000, TIFF, BMP, PNG).
Users can define optimal camera brightness and contrast settings with the Camera Adjustment Wizard.
Users can calibrate the application to match with their own visual cell counting data.
Includes a basic mode for introductory analyses and a full-featured mode.
Calibration of ImmunoRatio
The effect of variability in staining and image acquisition settings
The compatibility of ImmunoRatio with variable staining and image acquisition settings
Immunostaining/image acquisition feature
Compatibility with ImmunoRatio
Primary antibody dilution
(defined for MIB-1 Ki-67)
too dilute (1:400)
too low labeling index (Figure 6a)
+ + +
best match with visual counting (Figure 6b)
too strong (1:25)
cytoplasmic background causing overly high labeling indexes (Figure 6c)
insufficient nuclear segmentation (Figure 6d)
+ + +
best match with visual counting (Figure 6e)
false segmentation of cytoplasmic structures (Figure 6f)
Microscope objective magnification
(using 1× phototube)
non-carcinomatous cells often included*
+ + +
for accurate result, an average of three images per sample is recommended
for accurate result, averaging several images per sample is recommended
underexposed (too dim)
mean gray intensity of the blankfield image <200
in optimal range
+ + +
as guided by the Camera Adjustment Wizard of ImmunoRatio
overexposed (too bright)
mean gray intensity of the blankfield image >250
foreground mean gray intensity over 85% of the background mean gray intensity
in optimal range
+ + +
as guided by the Camera Adjustment Wizard of ImmunoRatio
foreground mean gray intensity under 50% of the background mean gray intensity
slow network transmission (slower overall analysis time)
JPEG, quality factor 50 to 100 (lossy)
+ + +
optimal for ImmunoRatio
JPEG, quality factor <50 (lossy)
visible image artifacts
Owing to the Scale Finder function, the results of images acquired using 10×, 20×, and 40× objective lenses (with a 1× phototube) were highly similar (data not shown). The 20× objective was deemed to be optimal, requiring at least three images per sample to be averaged. The same results could be achieved by using the 40× objective, but more images per sample needed to be averaged. When using a 10× objective, considerably more non-carcinomatous cells were often included in the analysis. However, the ROI functionality of ImmunoRatio can be used to circumvent this problem. The differences in ImmunoRatio analysis results between the tested camera models and repeated staining batches were found to be small (data not shown).
Variation in image brightness and uneven illumination can be accurately corrected by using the blankfield image, captured using the same microscope and camera settings. However, greatly underexposed images (blankfield image mean gray intensity <200) as well as overly overexposed images (blankfield image mean gray intensity >250) may cause false labeling indexes. For accurate nuclei segmentation, the image contrast must be relatively high; the foreground mean gray intensity should be 50 to 80% of the background mean gray intensity. Users can validate their image acquisition settings by using the Camera Adjustment Wizard function of ImmunoRatio.
For the ImmunoRatio web application, it is advantageous to use lossy image file formats (e.g., JPEG) to minimize the data uploaded to the server for analysis. We found that using lossy JPEG compression with quality factors 50 to 100 had no significant effect on the accuracy of ImmunoRatio analysis results (data not shown). This compression level allows a typical 5 megabyte uncompressed image to be compressed into 250 kilobytes (about 20:1 compression ratio), enabling rapid image transfer with almost any network bandwidth. Using very low JPEG quality factors (<50) can cause image distortion and artifacts, making the analysis unreliable.
In this study, we described an image analysis application, ImmunoRatio, which is an easy-to-use tool for assessing ER, PR, and Ki-67 labeling indexes in hematoxylin-counterstained tissue sections. ImmunoRatio analysis is based on defining positively stained pixel counts, which, according to our calibration data, correlates very well with cell nuclei enumerated visually. The calibration was performed using a training set of 50 samples and validation using a separate test set of 50 samples representing ER-, PR-, and Ki-67-stained routine breast cancer specimens. The correlation between manual and automated analysis was very high and matched, or exceeded, corresponding results of other similar image analysis software [30, 31]. Due to the significant inter-observer variability in visually defined labeling indexes, we recommend that the users calibrate ImmunoRatio with their own labeling index data, as demonstrated in Figure 4 for the calibration training set. Once calibrated, ImmunoRatio can be easily integrated with routine diagnostic work.
Another important aspect of calibration is to determine the optimal Ki-67 cutoff used for prognostic assessment. We tested this with a retrospective analysis of data from 123 primary breast cancer patients followed up for 20 years. The Ki-67 labeling index 20% (the median value in this material) gave a strong prognostic discrimination (HR = 2.2). Although cut-off values 15% and 25% yielded similar prognostication in this patient material, we recommend each laboratory to define their own cut-off value. We recommend using the median value of the Ki-67 labeling index as cut-off. This allows comparisons of different patient materials and provides a reproducible classification of patients according to Ki-67 labeling index.
In addition to accurate calibration, it is clear that for routine use, an image analysis system must accept variation in staining intensity, in microscope setup, and in image acquisition settings. We found up to eight-fold range in primary antibody (Ki-67) dilution to be acceptable for ImmunoRatio. However, when setting up an optimal staining protocol, the users should pay close attention to the hematoxylin counterstaining, which must be bright and clearly separate the nuclei from the background (see example Figure 6). In terms of optical resolution, we recommend using a microscope setup that roughly corresponds to 20× objective lens magnification, 1× phototube, and a 1.5 megapixel camera. Using this setup, a representative result from a typical breast cancer tumor (diameter 1 to 2 cm) can be obtained by averaging at least three images. Variation in image brightness is well-tolerated owing to the blankfield image correction. The Camera Adjustment Wizard function is designed to help the user find the optimal image brightness and contrast settings. A collection of reference images with optimal staining and imaging settings are presented on our website .
ImmunoRatio analysis is based on the color deconvolution algorithm , which is one of the several existing alternatives for separating the staining components. In addition to color deconvolution, stain separation and nuclei segmentation have been performed using texture analysis , cyan-magenta-yellow-black (CMYK) color model , hue-saturation-intensity color model , CIE 1976 L*u*v (CIELUV) color model , pattern recognition , cluster analysis , and immunofluorescence with Automated QUantitative Analysis (AQUA) . However, the software applications described in the above mentioned studies are mainly for research purposes and they have not been released for public use. Many of the methods may require considerable work if employed in a routine clinical process. The color deconvolution-based approach for separating two stains is straightforward and fast, and is readily usable for images captured with conventional microscope color cameras. If more than two staining components are used or the analysis requires accurate intensity-based quantification, the AQUA method or multispectral imaging would most likely be better alternatives .
ImmunoRatio was developed using ImageJ, which is a public domain (i.e., completely free and open source) image analysis software. However, a major obstacle in adopting ImageJ, or any other image analysis software, in clinical laboratories is usually the strict computer security policy. The local system and network rules usually prohibit users to download, install, and/or run external applications. To address these constraints, we released ImmunoRatio as a web application, which provides an easy-to-use web interface, requires no software downloads or installations, and can be used in highly restricted environments.
To the best of our knowledge, ImmunoRatio is the first ready-to-use web application for analyzing nuclear immunostains (e.g., ER, PR, and Ki-67). We want to point out that ImmunoRatio is meant to be used as a diagnostic aid by personnel trained to score immunostained breast cancer slides. Furthermore, the analysis results should always be interpreted together with the pseudo-colored images and the original sample slides. ImmunoRatio has already been used in the authors' laboratory for more than 1,000 cases and tested by several collaborators. The application is open to free public access on our research group website . Complementary software for analyzing cell membrane staining (e.g., HER-2) is currently being developed.
automated quantitative analysis
CIE 1976 L*u*v
region of interest.
This study was financially supported by the Finnish Cancer Foundation, the Sigrid Juselius Foundation, the Tampere University Hospital, and the Seinäjoki Central Hospital (governmental EVO grant). The authors wish to thank laboratory technicians Sari Toivola, Kristiina Ryömä, Ritva Kujala, Helvi Salmela, and Pirjo Pekkala for assisting in sample preparation and immunostaining.
- Allred DC, Carlson RW, Berry DA, Burstein HJ, Edge SB, Goldstein LJ, Gown A, Hammond ME, Iglehart JD, Moench S, Pierce LJ, Ravdin P, Schnitt SJ, Wolff AC: NCCN Task Force Report: Estrogen Receptor and Progesterone Receptor Testing in Breast Cancer by Immunohistochemistry. J Natl Compr Canc Netw. 2009, 7: S1-S21. quiz S22-23PubMedGoogle Scholar
- United Kingdom National External Quality Assessment Service (UK NEQAS). [http://www.ukneqas.org.uk/]
- Nordic immunohistochemical Quality Control (NordiQC). [http://www.nordiqc.org/]
- de Azambuja E, Cardoso F, de Castro G, Colozza M, Mano MS, Durbecq V, Sotiriou C, Larsimont D, Piccart-Gebhart MJ, Paesmans M: Ki-67 as prognostic marker in early breast cancer: a meta-analysis of published studies involving 12,155 patients. Br J Cancer. 2007, 96: 1504-1513. 10.1038/sj.bjc.6603756.PubMedPubMed CentralView ArticleGoogle Scholar
- Stuart-Harris R, Caldas C, Pinder SE, Pharoah P: Proliferation markers and survival in early breast cancer: a systematic review and meta-analysis of 85 studies in 32,825 patients. Breast. 2008, 17: 323-334. 10.1016/j.breast.2008.02.002.PubMedView ArticleGoogle Scholar
- Goldhirsch A, Ingle JN, Gelber RD, Coates AS, Thürlimann B, Senn HJ, Panel members: Thresholds for therapies: highlights of the St Gallen International Expert Consensus on the primary therapy of early breast cancer 2009. Ann Oncol. 2009, 20: 1319-1329. 10.1093/annonc/mdp322.PubMedPubMed CentralView ArticleGoogle Scholar
- Harvey JM, Clark GM, Osborne CK, Allred DC: Estrogen receptor status by immunohistochemistry is superior to the ligand-binding assay for predicting response to adjuvant endocrine therapy in breast cancer. J Clin Oncol. 1999, 17: 1474-1481.PubMedGoogle Scholar
- Shousha S: Oestrogen receptor status of breast carcinoma: Allred/H score conversion table. Histopathology. 2008, 53: 346-347. 10.1111/j.1365-2559.2008.03075.x.PubMedView ArticleGoogle Scholar
- Rüdiger T, Höfler H, Kreipe HH, Nizze H, Pfeifer U, Stein H, Dallenbach FE, Fischer HP, Mengel M, von Wasielewski R, Müller-Hermelink HK: Quality assurance in immunohistochemistry: results of an interlaboratory trial involving 172 pathologists. Am J Surg Pathol. 2002, 26: 873-882. 10.1097/00000478-200207000-00005.PubMedView ArticleGoogle Scholar
- Walker RA: Quantification of immunohistochemistry - issues concerning methods, utility and semiquantitative assessment I. Histopathology. 2006, 49: 406-410. 10.1111/j.1365-2559.2006.02514.x.PubMedView ArticleGoogle Scholar
- Taylor C, Levenson R: Quantification of immunohistochemistry - issues concerning methods, utility and semiquantitative assessment II. Histopathology. 2006, 49: 411-424. 10.1111/j.1365-2559.2006.02513.x.PubMedView ArticleGoogle Scholar
- Ruifrok AC, Johnston DA: Quantification of histochemical staining by color deconvolution. Anal Quant Cytol Histol. 2001, 23: 291-299.PubMedGoogle Scholar
- van der Loos CM: Multiple immunoenzyme staining: methods and visualizations for the observation with spectral imaging. J Histochem Cytochem. 2008, 56: 313-328. 10.1369/jhc.2007.950170.PubMedPubMed CentralView ArticleGoogle Scholar
- Rojo MG, Bueno G, Slodkowska J: Review of imaging solutions for integrated quantitative immunohistochemistry in the Pathology daily practice. Folia Histochem Cytobiol. 2009, 47: 349-354. 10.2478/v10042-008-0114-4.PubMedGoogle Scholar
- ImmunoRatio web application. [http://jvsmicroscope.uta.fi/immunoratio/]
- ImageJ image analysis software. [http://rsb.info.nih.gov/ij/]
- Java programming language. [http://java.sun.com/]
- Calculator Plus ImageJ plugin. [http://rsbweb.nih.gov/ij/plugins/calculator-plus.html]
- Sternberg SR: Biomedical Image Processing. IEEE Computer. 1983, 16: 22-34.View ArticleGoogle Scholar
- Colour Deconvolution ImageJ plugin. [http://www.dentistry.bham.ac.uk/landinig/software/cdeconv/cdeconv.html]
- Ridler TW, Calvard S: Picture thresholding using an iterative selection method. IEEE Trans Systems Man and Cybernetics. 1978, 8: 630-632. 10.1109/TSMC.1978.4310039.View ArticleGoogle Scholar
- Beucher S, Meyer F: The morphological approach to segmentation: The watershed transformation. Mathematical Morphology in Image Processing. Edited by: Dougherty ER. 1993, New York: Marcel Dekker, 433-481. 1Google Scholar
- Google Web Toolkit. [http://code.google.com/webtoolkit/]
- Apache Commons FileUpload library. [http://commons.apache.org/fileupload/]
- Apache Commons IO library. [http://commons.apache.org/io/]
- The Laboratory for Optical and Computational Instrumentation (LOCI) Bio-Formats library. [http://www.loci.wisc.edu/software/bio-formats/]
- Apache Tomcat Java servlet container. [http://tomcat.apache.org/]
- The TWAIN standard specification. [http://www.twain.org/]
- ImmunoRatio ImageJ plugin. [http://jvsmicroscope.uta.fi/immunoratio-plugin/]
- Mofidi R, Walsh R, Ridgway PF: Objective measurement of breast cancer oestrogen receptor status through digital image analysis. Eur J Surg Oncol. 2003, 29: 20-24. 10.1053/ejso.2002.1373.PubMedView ArticleGoogle Scholar
- Faratian D, Kay C, Robson T, Campbell FM, Grant M, Rea D, Bartlett JM: Automated image analysis for high-throughput quantitative detection of ER and PR expression levels in large-scale clinical studies: the TEAM Trial Experience. Histopathology. 2009, 55: 587-593. 10.1111/j.1365-2559.2009.03419.x.PubMedView ArticleGoogle Scholar
- Kostopoulos S, Cavouras D, Daskalakis A, Bougioukos P, Georgiadis P, Kagadis GC, Kalatzis I, Ravazoula P, Nikiforidis G: Colour-texture based image analysis method for assessing the hormone receptors status in breast tissue sections. Conf Proc IEEE Eng Med Biol Soc. 2007, 2007: 4985-4988.PubMedGoogle Scholar
- Pham NA, Morrison A, Schwock J, Aviel-Ronen S, Iakovlev V, Tsao MS, Ho J, Hedley DW: Quantitative image analysis of immunohistochemical stains using a CMYK color model. Diagn Pathol. 2007, 2: 8-10.1186/1746-1596-2-8.PubMedPubMed CentralView ArticleGoogle Scholar
- Ruifrok AC, Katz RL, Johnston DA: Comparison of quantification of histochemical staining by hue-saturation-intensity (HSI) transformation and color-deconvolution. Appl Immunohistochem Mol Morphol. 2003, 11: 85-91. 10.1097/00022744-200303000-00014.PubMedGoogle Scholar
- Rexhepaj E, Brennan DJ, Holloway P, Kay EW, McCann AH, Landberg G, Duffy MJ, Jirstrom K, Gallagher WM: Novel image analysis approach for quantifying expression of nuclear proteins assessed by immunohistochemistry: application to measurement of oestrogen and progesterone receptor levels in breast cancer. Breast Cancer Res. 2008, 10: R89-10.1186/bcr2187.PubMedPubMed CentralView ArticleGoogle Scholar
- Kostopoulos S, Cavouras D, Daskalakis A, Kagadis GC, Kalatzis I, Georgiadis P, Ravazoula P, Nikiforidis G: Cascade pattern recognition structure for improving quantitative assessment of estrogen receptor status in breast tissue carcinomas. Anal Quant Cytol Histol. 2008, 30: 218-225.PubMedGoogle Scholar
- Gustavson MD, Bourke-Martin B, Reilly DM, Cregger M, Williams C, Tedeschi G, Pinard R, Christiansen J: Development of an unsupervised pixel-based clustering algorithm for compartmentalization of immunohistochemical expression using Automated QUantitative Analysis. Appl Immunohistochem Mol Morphol. 2009, 17: 329-337. 10.1097/PAI.0b013e318195ecaa.PubMedView ArticleGoogle Scholar
- Camp RL, Chung GG, Rimm DL: Automated subcellular localization and quantification of protein expression in tissue microarrays. Nat Med. 2002, 8: 1323-1327. 10.1038/nm791.PubMedView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.