How a pathologist would analyse this H&E image?

How a pathologist would analyse this H&E image?

We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

I am working on a project which involves writing computer software to analyse histological images. A typical image looks like this:

It is a Hematoxylin and Eosin stained biopsy of breast cancer tissue. Being a programmer without biology background, I would like to get insight into how a pathologist analyses such images.

More specifically:

  • Which cells are cancer cells and which are regular cells?
  • Can an invasive margin (a curve marking the boundary of cancer invasion into the tissue) be seen in this image?
  • Can a pathologist provide a TNM classification looking at the image? If yes, what would be the T, N, M, G values for this image?

Prof. Peter Hamilton

Peter is Head of the Pathology Bioimaging and Informatics Laboratory within the Centre for Cancer Research and Cell Biology at Queen’s University Belfast. He is also Founder of and VP for Research and Development with PathXL Ltd, a global company specializing in digital pathology software for tumor analysis and biomarker discovery.

For the past 25 years, he has been leading research on digital pathology, computer vision and tissue bioimaging in diagnostic and molecular cancer pathology for the high throughput quantitative analysis identification of novel tissue and cell biomarkers markers. Being a pioneer of early techniques in tissue measurement, image analysis, pathology informatics and tissue biomarker discovery, he has been published in over 150 peer reviewed publications in some of the worlds leading journals.

Peter established the digital pathology laboratory at Queens University Belfast with a full range of whole slide scanning technology, specialized imaging techniques and associated image analysis software, which is now embedded within the wider Molecular Pathology Laboratory program. He also heads up a team of software developers in image analysis and pathology informatics, focused on developing new tools for automated tissue microarray analysis and clinogenomic data integration for biomarker discovery including high performance tissue imaging for biomarker discovery in prostate cancer.

As informatics lead for the Northern Ireland Biobank (NIB), a major initiative aimed at the prospective collection of high quality cancer tissue samples for translational research in cancer, he oversees the informatics to support tissue collection, tracking, storage and retrieval of tissue samples and integration of clinical, pathological and epidemiological data. He is also external informatics adviser to two national biobanks in Italy and South Africa. He has sat on MRC Panel of Experts, the Pathological Society Committee, Journal of Pathology Editorial Board and the International Society for Cellular Oncology.

As founder of PathXL Ltd (, Professor Hamilton has led the company through venture capital investment to a point where it now has an experienced senior management team, with sales across the world and an advanced portfolio of sophisticated digital pathology solutions for education, research and clinical practice. Recently, PathXL Ltd has been developing novel approaches to tumor imaging, cancer cell recognition, biomarker analysis and data integration aimed at supporting modern pathology in translational personalized medicine. PathXL has won the Frost and Sullivan Enabling Technology Award and the European New Product Innovation Award with its TissueMark™ platform for the automated identification and measurement of tumor tissue for molecular evaluation of solid tumors. PathXL is rapidly expanding its software products through Xplore to support the integration of digital pathology with big data integration. In addition to PathXL, he also sits on the board of a number of other Queen’s University spin-out biotechnology companies. Professor Hamilton has successfully married an outstanding academic career with a thriving digital pathology business, with the goal of translating innovative academic research into industry.

Volume 2

Matthew G. Hanna , Liron Pantanowitz , in Encyclopedia of Biomedical Engineering , 2019

Image Analysis

Image analysis is the extraction of meaningful information from images. Precision medicine demands precision diagnostics. Image analysis (computer vision) is an ideal application of digital pathology to address this need, helping pathologists to transition from providing just traditional qualitative (descriptive) reports to more quantitative results. Specialized image analysis software platforms are commercially available (e.g., Visiopharm, Definiens, Indica Labs, Virtuoso from Roche, and Genie from Leica). Analyses have shifted from FOV (static) images to WSI and from one to multiple markers (multiplexing), often even using many colors (multispectral imaging). Image algorithms include several steps such as image preprocessing (e.g., color normalization), detection, segmentation, feature extraction, classification, and quantification. Software algorithms have been developed to identify rare events (e.g., screening for microorganisms, counting mitoses, and detecting micrometastases in lymph nodes) and quantify stains (e.g., most commonly breast biomarkers) and various features (e.g., extent of tissue fibrosis and degree of liver steatosis). They can also analyze spatial patterns and feature distribution better than humans.

There are several benefits of image analysis. Algorithms offer better accuracy, because they provide more precise and even continuous quantitative measurements compared with humans. Using image algorithms also permits standardization due to more reproducible results, especially for intermediate scoring categories and complex scoring systems. Automated image analysis can reduce time consumption for pathologists, especially for performing mundane tasks like counting. Image analysis promises to introduce CAD, helping pathologists find, diagnose, and grade cancer. With deep learning and convoluted neural networks, some algorithms have been shown to even assist with predicting cancer diagnoses.

Types of Pathology

There are three main subtypes of pathology: anatomical pathology, clinical pathology, and molecular pathology. These subtypes can be broken down into even more specific categories pathology is a diverse field because so many different diseases and ways of studying diseases exist.

Anatomical Pathology

Anatomical pathology is the study of anatomical features, such as tissue removed from the body, or even an entire body in the case of an autopsy, to diagnose and increase knowledge of disease. Anatomical pathology can include looking at cells under a microscope, but it also involves looking at organs in general (e.g. a ruptured spleen). It also includes investigation of the chemical properties of cells, and their immunological markers. There are several broad subcategories of anatomical pathology:

  • Surgical pathology is the examination of tissues removed during surgery. A common example is the examination of a small piece of tumor tissue to determine whether the tumor is malignant (cancerous) or benign and make a diagnosis. This procedure is called a biopsy.
  • Histopathology is the examination of cells under a microscope that have been stained with dye to make them visible or easier to see. Often, antibodies are used to label different parts of the cells with different colors of dye or fluorescence. After the microscope became widespread in pathology, many different methods of preserving and dyeing tissue were developed.
  • Cytopathology is the study of small groups of cells shed in bodily fluids or obtained through scraping, such as those taken during a cervical Pap smear. A Pap smear detects cervical cancer and some types of infections. The cells are taken by swabbing the cervix, and are then processed and examined under a microscope to check for abnormalities.

Clinical Pathology

  • Chemical pathology, or clinical chemistry, involves the chemical analysis of bodily fluids, through testing and microscopy. Commonly, chemical pathology involves the study of blood and its immune components like white blood cells.
  • Hematology is also related to the study of blood, but it has more to do with identifying blood diseases specifically than chemical pathology does. Hematologists also study the lymph system and bone marrow, which are other parts of the hematopoietic system.
  • Immunology, or immunopathology, is the study of immune system disorders. It deals with immune responses to foreign molecules, allergies, immunodeficiencies, and organ transplant rejection.

Molecular Pathology

Molecular pathology is the study of abnormalities of tissues and cells at the molecular level. It is a broad category that is used to refer to the study of disease of any organ or tissue in the body by examining what molecules are present in cells. It can combine aspects of both anatomical and clinical pathology. Some techniques that can be used in molecular pathology include polymerase chain reaction (PCR) to amplify DNA, fluorescence labeling, karyotype imaging of chromosomes, and DNA microarrays (small samples of DNA placed onto biochips).

Researchers Develop a Mice Model to Identify and Study the Vascular Dynamics of CCM Lesions Leading to Cerebral Hemorrhage, Strokes, and Seizures

Yale researcher Huanjiao Jenny Zhou (Pathology) and additional School of Medicine researchers have established a Cerebral Cavernous Malformations (“CCM”) mice model which “identifies and uncovers the mechanism in which CCM3 mutation induced caveolae-Tie2 signaling contributes to CCM pathogenesis.”

Cerebral cavernous malformations (CCMs) are “vascular abnormalities which primarily occur in adulthood and lead to cerebral hemorrhage, stroke, and seizures.” CCMs can be initiated by an “endothelial cell (EC) loss of any one of the three Ccm genes: CCM1 (KRIT1), CCM2 (OSM), or CCM3 (PDCD10).”

Dr. Zhou and her research colleagues utilize mice “with a brain EC-specific deletion of Pdcd10 (Pdcd10BECKO)” which typically survive between 6-12 months. During this period, the mice develop bona fide CCM lesions in all regions of their brain which allowed the research team to study the “vascular dynamics of CCM lesions by using transcranial two-photon microscopy.”

These researchers have not only created a CCM mice model which studies vascular abnormalities, but they have also identified “caveolae-mediated Tie2 receptor signaling in the presence of its ligand Angpt2 which can offer a therapeutic approach to treat CCM disease.”

Other members of the Yale research team include: Lingfeng Qin, Quan Jiang, Haifeng Zhang, Busu Li, Qun Lin, her collaborators Katie N. Murray, Jaime Grutzendler and Wang Min in the Program of Vascular Biology and Therapeutics, Departments of Pathology, Neurobiology and Cell Biology You can read more about this exciting research in the January 25th online edition of Nature Communications.

Drs. Adam Bagg, Zubair Baloch, David Elder, Franz Fogt, Virginia LiVolsi, and Xiaowei (George) Xu are among the the eight top pathologists in the region.

While we are still facing a long road to normalcy, it is exciting to hear how successful our progress has been so far. And yet I also believe that a shared experience of struggle and the acknowledgment of suffering among us can be an opportunity for us to learn as a Department.

Want to keep in touch with what’s happening in the Department? Sign up for the PLM Newsletter:

Thank you for subscribing!

How a pathologist would analyse this H&E image? - Biology

By Charlotte Plestant (Scientific Content Marketing Manager), Eunice Stennett (Former CMO), David Guet (Digital pathology Specialist) - 26 May 2021

Over the last years, the world of Pathology has considerably evolved and brought out new challenges. Development of whole-slide scanners alongside Artificial Intelligence (AI) has empowered pathological analysis by digitizing immunohistochemistry, immunocytochemistry and H&E slides, providing better possibilities in patient selection and treatment. These changes have come with new hurdles, related to a sharp increase of the complexity of the images, higher expectations for image analysis and a shift in the pathologist's daily practices.

In this post we caught up with David Guet, Digital Pathology Application Specialist at Keen Eye. We asked him what his vision was with regards to Computational Pathology and why Artificial Intelligence driven solutions could support every expert in the field of Pathology.

The slide digitization opened the path to Computational Pathology

Two decades ago, hospitals, pharmaceutical industries and CROs (Contract Research Organizations) started their own journey on the path of innovation. Thanks to slide scanners, they moved from the slide review with a microscope to whole-slide imaging, introducing Digital Pathology. “We can break down Digital Pathology into two major steps”, explains David, “the first one is the process of digitization of histopathology, IHC or cytology slides using whole slide scanners. It is then followed by the management, the analysis and the interpretation of these whole-slide images (WSIs) using computational approaches”.

Computational Pathology incorporates multiple sources of data through AI developed tools (or mathematical models), to generate inferences. It requires managing biological and clinical information from large and high-throughput data sets. Computational Pathology relying on Deep Learning tools gives access to better patient selection and drug treatment therapies.

AI-driven solutions have become mandatory to face the increasing complexity of image analysis

Originally in research, pre-clinical, clinical and diagnostic areas, analysis of images to support scientific experts was not systematically required. As the data to process became more and more complex, AI-automated image analysis has turned out to be vital for robust interpretations and to extract relevant information, ranging from biomarkers to anatomo-pathological features within each image.

This inevitably has had an impact on the workflow of the different experts. David emphasizes that “there were more slides to screen, more biomarkers to analyze, and within shorter time frames”. This increase in the number of digital slides requiring review from the pathologists could also lead to an increase in the variability of the results, inter and intra-operator.

Alongside these parameters, “we have to take into account a shortage in experienced pathologists, while the network between companies and labs working together all over the world is getting more and more complex”, adds David. WSIs are heavy data to handle: it is crucial to be able to access and to share these images through a responsive and scalable viewer. Being able to answer such needs in all the collaborative research approaches is a cornerstone in Computational Pathology.

Positive impact of AI-driven image analysis on the work of Digital Pathology experts

Pathologists and researchers are facing an increase in their workload. It is crucial to create the right analysis solutions that will provide them with the relevant information on a study, allowing them to cope with the high volumes of data: around 75% of pathologists throughout 59 countries worldwide have declared to be interested in using AI as a diagnostic tool¹.

AI-driven algorithms have to be designed under pathologist guidance and validation, in order to accompany and help them all along their work. Our Application Specialist highlights that “these solutions are made in collaboration with pathologists, for pathologists”. Indeed, with proper training and development, an AI image analysis can bring outstanding outcomes, thereby supporting the work of these experts. Moreover, the use of batch processing with a validated algorithm will allow faster and continuous analysis, with a quality assurance control and a reduced number of errors.

Having a reliable and productive AI-driven tool frees more time for researchers and pathologists to focus on the essential and make breakthrough discoveries.

The Phenoptics Instrument Portfolio

Our Phenoptics digital pathology slide scanners enable you to visualize, analyze, quantify, and phenotype cells revealing complex biology in whole tissue sections. Our multispectral imaging technique enables you to capture the multiple interactions occurring between cells because we’ve carefully unmixed each color from one another while also isolating autofluorescence into its own color channel so you can easily exclude it from your digital slide analysis. That means you can have confidence in accurately quantifying the interactions that are really occurring in the biology.

Vectra Polaris Automated Quantitative Pathology Imaging System
The Vectra ® Polaris™ is a world class digital pathology slide scanner featuring MOTiF™ 6-plex, 7-color whole-slide multispectral scanning that can accurately detect and measure weakly expressed and overlapping biomarkers within a single H&E, IHC or mIF intact FFPE tissue section or TMA, to accelerate your research.

Vectra 3 Automated Quantitative Pathology Imaging System
The Vectra ® 3 automated quantitative pathology imaging system is a six-slide pathology scanner that can help you better visualize, analyze, quantify, and phenotype immune cells in situ in FFPE tissue sections and TMAs.

Mantra 2 Quantitative Pathology Workstation
Mantra™ 2 quantitative pathology workstation is a manual microscope equipped with inForm® image analysis software enabling easy visualization, quantification and phenotyping of multiple types of immune cells simultaneously in intact FFPE tissue sections for cancer immunology research.

Precision Histology: Steatosis – Quantitative Analysis of Abnormal Lipid Accumulation

Abnormal lipid accumulation in tissue, or steatosis, is associated with numerous pathological conditions. Hepatic steatosis is a hallmark of alcoholic liver disease (ALD) and is also an early marker of Nonalcoholic Fatty Liver Disease (NAFLD), which is now the most prevalent cause of liver dysfunction in the United States. Histological scoring of hepatic steatosis is therefore an essential clinical tool for the evaluation of liver disease.

In addition to the liver, abnormal lipid deposits can be observed in numerous other tissues in the context of disease. For example, Gaucher disease and Nieman-Pick disease are inherited lipid storage metabolic disorders characterized by the harmful accumulation of lipids in the kidneys, spleen, lungs, bone marrow and brain. Moreover, identification of liposarcomas, tumors derived from fat cells, requires evaluation of lipid accumulation in tissue samples. The clinical relevance of pathological lipid deposits in tissue mean there is a critical need for sensitive histological approaches to evaluate and quantify steatosis in both patients and animal models of human disease.

Lipid accumulation is traditionally assessed by a Board Certified Pathologist however, recent advances in image analysis and machine learning have enabled more precise quantitative analysis to be performed independently or to assist the pathologist at scale. Abnormal lipid accumulation is commonly assessed in preclinical research and clinical trials using any of all of the following approaches.

Method One: Hematoxylin and Eosin (H&E)

Figure 1. A) Liver tissue stained with H&E. B) Mask applied to original image to isolate lipid droplets (Blue) for quantification. The scale bar represents 50 μm.

H&E staining on formalin-fixed, paraffin-embedded (FFPE) samples is an easy, low cost approach to identify lipid accumulation in many tissues. Morphologically, lipid droplets appear as circular voids in the tissue as the fat content is dissolved by solvent treatment during tissue processing (Figure 1A). H&E staining is therefore an indirect method for the quantification of lipid accumulation. Whole slide images of H&E staining, obtained using bright field microscopy, can be analyzed to determine the number and size of lipid droplets as well as the degree of steatosis as a percentage of the total area of the tissue (Figure 1B). While H&E is a clinically approved method to score steatosis, additional histological methods can directly detect lipid deposits in tissues.

Method Two: Oil Red O

Lipid accumulation can be evaluated directly in fresh frozen tissue sections by staining with Oil Red O, a fat-soluble dye that specifically stains triglycerides and neutral lipids a deep red color (Figure 2A). The use of fresh frozen tissues avoids the loss of lipid content that occurs during processing of formalin-fixed tissue.

Figure 2. A) Frozen liver tissue section stained with Oil Red O (Red) and hematoxylin nuclear counterstain (Blue). B) Mask applied to original image to isolate lipid droplets (Red) for quantification. The scale bar represents 10 μm.

Whole slide images of tissue sections stained with Oil Red O can be generated using bright field microscopy. ImageDx™ analysis software precisely measures a range of parameters to quantify Oil Red O staining, including percentage of positivity, average number of droplets per unit area and mean droplet size (Figure 2B).

Method Three: Perilipin Immunostaining

Perilipin, also called lipid droplet-associated protein, is a protein that specifically coats the periphery of lipid droplets. Perilipin can be detected in both FFPE and frozen tissue sections by immunohistochemistry (IHC), a technique that uses primary antibodies raised against their specific target antigens. These primary antibodies are then detected using a range of secondary reagents, and visualized with 3,3-Diaminobenzidine (DAB – brown). Primary antibodies can also be detected using a range of fluorescent markers in a process known as immunofluorescence (IF). Staining tissue sections with perilipin specifically labels the outline of fat vacuoles, as shown in Figure 3.

Figure 3: Perilipin staining visualized with DAB (brown) on FFPE liver section. Nuclei are conterstained with hematoxylin (blue). The scale bar represents 50 μm.

Whole slide imaging of perilipin immunostaining can be performed using either bright field or immunofluorescence microscopy, depending on the secondary detection strategy. A wide range of quantification parameters can be reported including the mean number of lipid droplets per unit area, droplet size, and lipid area. Perilipin staining is extremely sensitive as it detects multiple small lipid droplets that may be missed by H&E. In addition, use of perilipin as a marker of steatosis offers the considerable advantage that it can be multiplexed with other biomarkers of interest on the same section.

In conclusion, the precise measurement and analysis of lipid accumulation in tissue sections provides valuable data when evaluating steatosis in both humans and animal models of human disease. Selection of the most appropriate method for your research depends on several parameters including the format of available tissue and the data end points required. Contact us to discuss how precision histology can benefit your next study.

Deep learning trained on H&E tumor ROIs predicts HER2 status and Trastuzumab treatment response in HER2+ breast cancer

The current standard of care for many patients with HER2-positive breast cancer is neoadjuvant chemotherapy in combination with anti-HER2 agents, based on HER2 amplification as detected by in situ hybridization (ISH) or protein immunohistochemistry (IHC). However, hematoxylin & eosin (H&E) tumor stains are more commonly available, and accurate prediction of HER2 status and anti-HER2 treatment response from H&E would reduce costs and increase the speed of treatment selection. Computational algorithms for H&E have been effective in predicting a variety of cancer features and clinical outcomes, including moderate success in predicting HER2 status. In this work, we present a novel convolutional neural network (CNN) approach able to predict HER2 status with increased accuracy over prior methods. We trained a CNN classifier on 188 H&E whole slide images (WSIs) manually annotated for tumor regions of interest (ROIs) by our pathology team. Our classifier achieved an area under the curve (AUC) of 0.90 in cross-validation of slide-level HER2 status and 0.81 on an independent TCGA test set. Within slides, we observed strong agreement between pathologist annotated ROIs and blinded computational predictions of tumor regions / HER2 status. Moreover, we trained our classifier on pre-treatment samples from 187 HER2+ patients that subsequently received trastuzumab therapy. Our classifier achieved an AUC of 0.80 in a five-fold cross validation. Our work provides an H&E-based algorithm that can predict HER2 status and trastuzumab response in breast cancer at an accuracy that is better than IHC and may benefit clinical evaluations.



Cell-based image analysis, assays, and partnership to advance clinical trials and regulatory acceptance.

Clinical Diagnostics

Dependable clinical diagnostic testing to connect patients to individualized, immuno-oncology therapies.

Related Resources

Posters & Publications

About Flagship Biosciences

We are a multi-disciplinary team of pathologists, biologists, analysts, software engineers, and technicians who are committed to creating successful approaches which match patients with life-saving drugs.
Learn More >

Connect With Us

11800 Ridge Parkway, Suite 450
Broomfield, CO 80021

  • Copyright 2021 Flagship Biosciences Inc.
  • Privacy Policy
  • 11800 Ridge Parkway, Suite 450, Broomfield, CO 80021

Privacy Overview

Necessary cookies are absolutely essential for the website to function properly. These cookies ensure basic functionalities and security features of the website, anonymously.

cookielawinfo-checbox-analytics11 monthsThis cookie is set by GDPR Cookie Consent plugin. The cookie is used to store the user consent for the cookies in the category "Analytics".
cookielawinfo-checbox-functional11 monthsThe cookie is set by GDPR cookie consent to record the user consent for the cookies in the category "Functional".
cookielawinfo-checbox-others11 monthsThis cookie is set by GDPR Cookie Consent plugin. The cookie is used to store the user consent for the cookies in the category "Other.
cookielawinfo-checkbox-necessary11 monthsThis cookie is set by GDPR Cookie Consent plugin. The cookies is used to store the user consent for the cookies in the category "Necessary".
cookielawinfo-checkbox-performance11 monthsThis cookie is set by GDPR Cookie Consent plugin. The cookie is used to store the user consent for the cookies in the category "Performance".
viewed_cookie_policy11 monthsThe cookie is set by the GDPR Cookie Consent plugin and is used to store whether or not user has consented to the use of cookies. It does not store any personal data.

Functional cookies help to perform certain functionalities like sharing the content of the website on social media platforms, collect feedbacks, and other third-party features.

Performance cookies are used to understand and analyze the key performance indexes of the website which helps in delivering a better user experience for the visitors.

Analytical cookies are used to understand how visitors interact with the website. These cookies help provide information on metrics the number of visitors, bounce rate, traffic source, etc.

Advertisement cookies are used to provide visitors with relevant ads and marketing campaigns. These cookies track visitors across websites and collect information to provide customized ads.

Other uncategorized cookies are those that are being analyzed and have not been classified into a category as yet.

Watch the video: Α. Σωτηρόπουλος MD. PhD, Παθολόγος - Διαβητολόγος u0026 Γενικός Ιατρός, Διευθυντής ΕΣΥ (September 2022).


  1. Chatham

    And what in this case?

  2. Ashtaroth

    Gorgeous, where can I get it?

  3. Shereef

    What do you say if I say that all your posts are fiction?

  4. Shakajas

    Between us, I would try to solve this problem myself.

  5. Kikree

    Well written.

  6. Brodrig

    It is also possible on this issue, because only in a dispute can the truth be achieved.

  7. Eliazar

    So it happens. We will examine this question.

Write a message