Colorectal cancer is one of the leading causes of cancer-related deaths worldwide. It is estimated to be the third most common cancer and the second leading cause of cancer death in both men and women. A development known as a polyp, which may initially be noncancerous (benign) or later turn cancerous, is how colorectal cancer typically manifests.

The everwidening uses of AI in healthcare industry are astonishing. AI can play a significant role in detecting colorectal cancer by assisting in various aspects of the screening and diagnosis process. A few very important examples are:

  • In medical imaging, AI algorithms are capable of finding polyps, lesions, and cancers.
  • Radiologists and pathologists can assess x-rays with the use of AI-powered CAD systems.
  • Using patient data, AI models can evaluate the risk of colorectal cancer.
  • Natural Language Processing (NLP) can be used by AI systems to extract information from medical records and recognize colorectal cancer.

A very effective HyperCRC-NET module from HS Analysis GmbH can quickly identify both malignant and tumor cells that are on the verge of developing into cancer. Without appropriate data for analyses, detection alone is not sufficient. The HSA KIT software offers the chance to extract all the data into an excel sheet and present the results as reports, graphs, and figures.

HyperCRC-NET module in HSA KIT

The preview of HSA KIT above shows the following features:

  • Module name on the top left and the image information
  • A preview window at bottom left for navigation
  • Tool panel on the right side consisting of annotation tools
  • View, ROIs and AI tab on the right:
    • View tab shows all the images selected for the project
    • ROIs tab shows Base ROI and structures of interest
    • AI tab includes custom model imports and option to train a new model


The HSA KIT software has an intuitive user interface and high-end professional annotation features to assure accuracy down to the last pixel. By standardizing processes, maintaining consistency, and promoting reproducibility, this can enhance the overall analysis process.

HyperCRC-Net Module in HSA KIT very efficiently categorises between different types of glands within the Base ROI (Region of Interest):

  • Healthy glands- green
  • Adinoma glands- yellow
  • Carcinoma glands- red
  • Other tissues (muscle, adipose or stroma)- blue

An example of various glands within Base Roi.

An example of detecting Carcinoma and Adinoma.

Zoom out view: Numerous healthy glands detected.

Zoom out view: Numerous carcinoma glands detected.

Healthy glands detected in both longitudnal section and cross-section.

Adipose/Muscle tissue detected under ‚Other tissues‘.

Precised recognition of carcinoma glands.

Exclusion of stroma with unbelievable accuracy.

The HSA KIT includes professional, top-notch annotation capabilities and offers zooming in to the very last pixel in the image. Existing DL models can be trained on by the clients and optimized for their purposes. Additional annotations can be readily added to the fundamental structures that are already present, in order to boost the model’s detection precision. A high-quality deep learning model will be generated as a result of improved training, thanks to the ground truth data (GTDs) that are all recorded in the database from the beginning.

This is a demonstration of zoom-in capabilities in HSA KIT, making precise annotations possible for GTDs

With a huge set of GTDs, a deep learning AI model is established which has the capability to objectively identify structures and ensure consistency and repeatability; while avoiding subjective viewpoints or preconceptions that effect the analysis procedures.

A demonstration of How to Annotate for HyperCRC-Net in HSA KIT

Ground Truth Data (GTD) creation

Mean Average Precision (mAP) is a metric used to evaluate object detection models. mAP calculates the average precision across different levels of object detection thresholds, providing a comprehensive measure of the model’s performance. The four sub-metrics that make up the core of the mAP accuracy formula are Confusion Matrix, Intersection over Union (IoU), Recall, and Precision. Finding the Average Precision (AP) for each class and then averaging it across several courses is how the mAP is determined.

The confusion matrix allows to assess the model’s performance by comparing its predictions with the actual outcomes. It helps in understanding the types of errors made by the model and can be used to calculate various evaluation metrics such as accuracy, precision, recall, and F1 score.

Confusion matrix

Cost and Loss Function: A cost function calculates the overall error of a model over the full training set. By lowering the aggregated error, which is frequently the sum or average of individual loss values, it guides the optimization process.. A loss function in a machine learning model measures the discrepancy between expected and actual values. It gauges how well the predictions made by the model correspond to reality.

Intersection over Union: IOU offers a measurement of how closely a predicted object matches the annotation of the real object, allowing for the evaluation of model accuracy and the fine-tuning of algorithms for better outcomes. An improved alignment between the anticipated and real regions is indicated by a greater IOU value, which reflects a more accurate model.

The samples used for the DL model were obtained from the diseased patients and tissues were stained with HE staining. To create the GTD needed for the DL model, the digitalised slides were loaded into HSA KIT software and the target area i.e. the Base ROI was annotated and classified into healthy, adenoma or carcinoma classes. The first model was trained with 1647 GTD, but was not good enough, so more GTD were created and the model trained again with 4631 GTD. The class distribution and percentages in the created data are given in the table below:

Two different types of models trained to check influence of the amount of GTD used

Depending on the type of GTD in each file, different HyperParameters were set. Both models were trained using Segmentation and Instance segmentation with:

Epochs= 200

Pyramid size = 1.1-1.4

Tile size= 256-1024

The HyperCRC-Net module has two versions:

  1. Type 1 based on the Vision Transformer (ViT) architecture, focuses on instance segmentation.
  2. Type 2 using the U-Net architecture, for segmentation.

After the model has been trained and the project has been completed, the HSA KIT software provides the information required to understand the results in a sheet format, including the name of the files, the number of objects (structures), the dimension of the analyzed areas in square meters within specific Base ROI (Region of Interest) area.

Interpretation of Results

Using metrics to compare the performance of different models and identify any patterns or trends can provide insights on the accuracy of the model. By analyzing the graphs, one can make informed decisions about its suitability for predicting Colorectal Cancer.

Accuracy calculations for each model

This indicates that increasing the amount of ground truth data (GTD) used for training the models leads to improved accuracy in their predictions. Model trained for healthy glands (Gesund) has an overall higher accuracy because their shape is more defined than Adenoma or Carcinoma structures, which are more distorted and irregular in nature.

Loss and mAP value of GTD used for the “Gesund-Adenoma-Carcinoma” structures trained models (Type 1)
Loss and mAP value of GTD used for the “Adenoma-Carcinoma” structures trained models (Type 1)

For HyperCRC-Net Type 1 using Instance Segmentation, it is evident that Loss function declines while mAP rises with increasing number of GTDs. This indicates that the model is effectively learning to accurately identify and segment instances in the given dataset.

Loss and mAP value of GTD used for the “Gesund-Adenoma-Carcinoma” structures trained models (Type 2)

For the HyperCRC-Net Type 2 that uses segmentation, it is evident that the Loss function also falls and the IoU rises as the number of GTDs increases.

Visualization of HyperCrcNet (Type 1) Training Results for 1.6K and 4.6K GTD, including “Gesund, Adenoma and carcinoma” Structures.
Visualization of HyperCrcNet (Type 2) Training Results for 1.6K and 4.6K GTD, including “Adenoma and Carcinoma” Structures.

The variations in adenoma and carzinoma sizes and shapes are most likely to blame for the discordance in results, which emphasizes the need for more information and model adjustments. because Gesund maintains its original size and shape whereas adenoma and karzinom change.  More data is required for AI to produce an effective model. To train the algorithm on a more varied dataset, this would entail gathering photos of adenoma and Karzinom in a variety of sizes and forms. Improvements to the model design and the use of methods like data augmentation may also help it perform better.

HE and Ki67 staining

In histopathology, H&E staining is regarded as a fundamental and standard procedure that provides useful information about tissue samples and aids in the diagnosis and comprehension of a wide range of illnesses and ailments.

Hematoxylin is a blue-purple dye that only stains the nucleus of cells. It turns the cell nucleus a dark blue or purple color after attaching to DNA and other acidic components there.

Eosin staining: Eosin is a pink or red dye that stains the cytoplasm and extracellular elements of tissues. It produces a pink or red tint when it attaches to fundamental components like collagen fibers and cytoplasmic proteins.

The separation between various cell types within the tissue sample as well as the viewing of cellular features and tissue architecture are all made possible by the use of hematoxylin and eosin staining.

Whereas, an immunohistochemical staining method called „Ki-67 staining“ is used to identify and quantify the expression of the Ki-67 protein in cells. During different stages of the cell cycle, actively dividing cells express the nuclear protein ki-67, which is linked to cell proliferation. In many different scientific and clinical settings, Ki-67 staining is frequently employed. This is especially true in cancer research, where it is used to evaluate tumor cell proliferation, ascertain the growth fraction of tumors, and help predict prognosis or treatment response.

Ki-67 antigen, which is expressed during the active phases of the cell cycle (G1, S, G2, and mitosis), is detected and mapped out by staining. A faster proliferation rate and enhanced cell division in the tumor are both indicated by elevated Ki-67 expression.

Under a microscope, the Ki-67 staining reveals the presence and localization of Ki-67 protein expression within the tissue sample, allowing for the evaluation of cell proliferation rates and identification of actively dividing cells.

When analyzing histopathological slides, the HSA KIT offers a one-stop solution for detecting colorectal cancer using several stains. Large datasets of annotated histopathology images are used to train the HyperCRC-Net module on patterns and feature specific to colorectal cancer.

HSA KIT offers the possibility to digitally analyse colorectal carcinoma in biopsies with different staining

Pertinent data from the slides, such as cellular morphology, architectural patterns, and Ki-67 expression patterns is collected, to produce precise predictions or classifications. The detection, segmentation, classification, and grading of colorectal cancer are based on histological characteristics, Ki-67 expression, and proliferation indices. Additionally, they offer prognostic data on patient outcomes and therapy options.

HSA Manipulation Tool

Manipulation Tool combines sophisticated AI technology with in-depth expertise in microscopy to enable precise and efficient image analysis.

Our software uses advanced AI models and offers a wide range of options for intelligent data use and creation.

Handling Large Amounts of Data: A particular advantage of our software is its ability to work with Whole Slide Images (WSIs) of any size. This allows researchers and laboratory personnel to perform comprehensive image analysis without limitations.

User-Friendly and Efficient Labeling Tools: Our intuitive interface offers users an efficient way to work:

• Easy selection: click on the polygon of interest.

• Intuitive usability: Change the polygon using common rotation handlers.

• Customizable editing: The configurable options can be used to determine whether the selected polygon should be changed or a new one should be created. It is also possible to integrate substructures into the transformation process or even change all polygons based on the model you have chosen.

• Information transfer: In conjunction with the manipulation tool, the HSA KIT also offers the option of transferring polygons to other files in the split screen view.

Selection of annotation with left click
Scaling of the annotation over corners of the rotation handler
Rotation and displacement of the annotation via external and central control points

When examining tissue and cells, it is often necessary to use different stains to visualize specific cell types or structures. The dyes used only penetrate certain structures of the cell and thereby mark their positions. In this way, these structures can be distinguished from other, unstained cell components.

HE – Ki67 Before transfer
HE – Ki67 After transfer
HE – Ki67 Close-up after transmission

By transferring annotations one can identify and compare the same cells and structures in different colorings. This can help increase the accuracy of diagnosis and facilitate research. Transferring annotations from one stain to another can also aid in the interpretation of histological sections. For example, one can use one coloring to mark specific elements and then transfer that annotation to another coloring to mark other structures.

With this new function it is possible to rotate and mirror raw data in 90 degree increments for better viewing and analysis.

Rotating images 90 degrees can help reveal details that may not have been visible due to the orientation of the image, thereby gaining a better understanding of the object’s structure and composition. Flipping images can also help you see certain features better by flipping the image horizontally or vertically.

HSA Case Viewer

The HSA Case Viewer provides a comprehensive and efficient way for medical professionals to analyze and evaluate cases, saving them valuable time and effort. By seamlessly accessing all relevant information in one platform, healthcare providers can make more accurate diagnoses and provide better patient care. Additionally, the ability to closely examine tissue sections allows for a more detailed understanding of the case, aiding in the development of personalized treatment plans. 

The HSA Caseviewer displays the cases that are currently in the system along with some preliminary metadata in the form of tables, such as name, case number, gender, etc

. This allows users to easily access and navigate through the cases, making it convenient for them to retrieve specific information quickly. Additionally, the Caseviewer also provides filters and search functionalities to further streamline the process of finding relevant cases based on specific criteria or keywords. 

Each case can have detailed information shown, including a sample of the tissue sections related to the case.

This feature allows users to visually inspect the tissue sections and gain a better understanding of the case before diving into the details. Furthermore, the Caseviewer offers a secure and centralized platform for authorized personnel to collaborate and share their insights on specific cases, facilitating efficient communication and decision-making within the medical community.

A summary of the patient, a clinical evaluation, a diagnosis, and comments can be seen in addition to the tissue sections and other more in-depth information.

This comprehensive platform not only enhances the overall efficiency of case analysis but also promotes interdisciplinary collaboration among medical professionals. By providing a holistic view of the patient’s condition, the Caseviewer enables clinicians to make well-informed decisions and develop personalized treatment plans.

Additionally, the ability to access and review previous cases stored in the platform’s database allows for continuous learning and improvement in medical practices.

Because the case viewer and file viewer are integrated directly, you can view all pertinent information straight from the case viewer while also more closely examining the tissue sections of the case.

Advanced Heatmap tool

HyperCRC-Net was created with the intention of not only detecting colorectal carcinoma but also measuring the size, extent, and perineural invasion of tumors or malignant glands, all of which are clinically significant traits. The use of this approach in a pathology lab for routine clinical use would be extremely beneficial to pathologists. HSA KIT also provides a heatmap tool which gives slide level scores of specifically chosen endpoints. These heatmaps show the likelihoods or probabilities of the endpoints being present at each pixel on the slide.

The use of this technology by professionals to diagnose and treat cancer patients will completely changed the area of oncology. Future approaches for identifying and treating cancer should become progressively more precise and effective as technology develops.

The advanced Heatmap tool in HSA KIT can be used to understand the decisions of already trained Deep Learning Models. Once a DL model is trained and let run on the files with Colorectal biopsies, carcinoma areas can easily be visualised. By activating xAI and heatmap icon, red region shows the pixels that are mostly important for the DL model to build the class Carcinoma. Blue region shows that DL model use these pixels for building the same class, but does not priories them. In that way it becomes easy to understand the decision of deep learning models during the training process as well as after training.

Heatmap tool used to detect healthy glands

The underlying training data and training procedure, as well as the results from AI models, provide the finest understanding of an AI system. This knowledge needs the capacity to map a trained AI model to the actual dataset used to train it, as well as the ability to closely analyze that data. Paying close attention to the data used to train a model is one of the simplest methods to improve its explainability. Teams must establish where the data to train an algorithm will come from, whether the data contains bias, and what can be done to reduce that bias during the design process. xAI Tool in HSA KIT finds the target area containing signal in a certain location and it is a discriminative localization technique that can give visual explanations for any project without needing architectural modifications or training.

By activating the “Heatmap” icon in HSA KIT and selecting a structure, the results appears for that structure in visual way.

The images below shows the heatmap of Colorectal carcinoma. The warm colors (yellow and red) represent areas with higher intensity detected by the model and blue regions show low priority pixels for the respective class.

Custom Modules in HSA KIT

In a lab setting, the HSA KIT is the ultimate tool for developing and training custom deep learning models. With hundreds of scalable modules available, this kit is designed to meet long-term requirements and build flexible solutions that can adapt to changing needs over time. Whether you’re working on a complex project or a simple one, the HSA KIT has everything you need to get started.

From data preprocessing to model building and evaluation, all the necessary tools for successful deep learning projects are included. With user-friendly interface and comprehensive documentation, even beginners can quickly learn how to use this powerful tool. So if you’re looking for an all-in-one solution for your deep learning needs, look no further than the HSA KIT. 

Out of hundreds of custom modules, only 15 can be seen in the image above

Moreover, through a client-centric approach, we leverage the information from iteraction to design, and deliver modules that perfectly align with clients‘ specific goals and objectives; ensuring maximum value and satisfaction.

Better workflow with HSA KIT

At HS Analysis, we take a comprehensive approach to slide analysis. We don’t just analyze the slides themselves, but we also integrate our solutions within the existing infrastructure. This includes connecting to LIS (Laboratory Information System) and naming the slides in a way that makes them easy to save, search, and use for medical purposes. Our goal is to provide a seamless experience for our clients, allowing them to access and utilize the data they need quickly and efficiently.

Traditional workflow of medical image analysis is rather cumbersome where the medical images obtained from CT Scans, MRIs or X-Rays is pre-processed and analysed by a radiologist who evaluates and interprets the images to prepare a very subjective report based on his own knowledge. This might be followed up with a quality assurance by another radiologist for second opinion.

On the other hand, AI based analyses using HSA KIT would provide:

  • Standardised process with subjective/objective analysis
  • Extraction of relevant features from raw data and create meaningful representations for training AI models
  • Module selection and configuration without excessive coding
  • Easy to learn software: Annotate , Train and Automate
  • Quick and efficient analysis of multiple medical images which cuts down time for diagnosis or treament
  • Automated report generation to boost productivity and help physicians or radiologists in evaluation process
Development of the microscopy infrastructure HSA KIT

The HSA KIT software is designed to seamlessly integrate into existing laboratory information systems (LIS), enabling seamless data exchange and interoperability. This integration streamlines the entire workflow, from sample collection to report generation, minimizing manual data entry and reducing the chances of transcription errors. Real-time data synchronization between the HSA KIT software and the LIS ensures that all relevant information is readily accessible, enabling a smooth and efficient diagnostic process.

Furthermore, with HSA Case Veiwer it becomes incredibly easy to store client files in a much more organized way. We provide a smart data management system with HSA KIT that makes structured data available for case-based workflows. Our system is a useful tool to streamline operations because it enables users to quickly organize and analyze data. Furthermore, the sophisticated features guarantee data security and privacy, providing users with comfort when handling sensitive information.

Data storage with HSA Case Viewer

We understand the importance of having accurate and efficient slide analysis software. That’s why we offer a range of features to help you achieve your goals.

  • Our software is compatible with both Linux and Windows operating systems, and it runs smoothly in Docker. With offline functionality, you can work on your slides without an internet connection, but if you need to extend to online use, that’s possible too.
  • We also offer professional integration into your network infrastructure, so you can seamlessly incorporate our software into your existing systems.
  • Our software is designed to work with other programs as well, ensuring that you have all the tools you need at your disposal.
  • Even on weaker computers, our software delivers full performance, so you can analyze slides quickly and accurately.
  • And with the ability to generate reports in CSV and Excel formats, you’ll have all the data you need at your fingertips.
  • Finally, user profile management ensures that each user has their own history and preferences stored for easy access.

Over a short period of time, HS Analysis GmbH has partnered with numerous companies, hospitals and laboratories as clients which provide microscopic slides or data that needs to be analysed. The physical slides can be digitalised to whole slide images (WSI) with the help of HSA KIT software that can be integrated to commonly used microscopes, and data can be retrieved in the form of reports and atomated graphical representaions. This helps to eliminate the extensive process that is usually practiced by medical professionals.

AI based analyses can be way more efficient, however it can only act as a supportive tool to human expertise and never be a complete alternative. AI algorithms have the tendency to produce false positives and false negatives, and may have trouble adapting to a data that deviates from training dataset.

With our advanced technology and expert team of professionals, we are confident in our ability to deliver top-notch slide analysis services that meet the unique needs of each individual client. Whether you’re looking for a one-time analysis or ongoing support, we have the tools and expertise necessary to help you achieve your goals.

Digitalisation of slides

The HSA KIT is extremely versatile and pairs well with some of the best automatic slide scanners like the Leica GT 450. It was designed specifically to seamlessly integrate with such devices and enable quick and precise slide scanning. The HSA KIT also offers cutting-edge features like automated slide handling and adjustable scanning parameters, further enhancing its incorporation within digital pathology infrastructure. Users can use these features to personalize their scans and streamline the scanning process to suit their needs. Professionals in disciplines like pathology and research will find the HSA KIT to be an excellent choice because of its seamless integration with their pre-existing technology.


At the same time, HS Analysis provides an affordable alternative to other automatic scanners- HSA Scan M software. The slides can be manually digitalized directly on current microscope to create manual WSI (whole slide images). This allows for greater flexibility and control over the scanning process. Additionally, the HSA KIT can be used with a variety of microscope models, making it a versatile option for labs with multiple microscopes. The software is user-friendly and easy to navigate, allowing for efficient scanning and image processing.

Digitalization of microscopic slides to medical images with HSA SCAN

With the ability to create manual WSI, labs can save money on expensive scanners while still producing high-quality digital images. Overall, the HSA KIT and HSA SCAN M software provide a cost-effective solution for labs looking to digitize their slides without sacrificing quality or control. Manual scanning converts a slide to a file with HSA SCAN M software so that it can be trained and then automatically detected by a trained AI model using HSA KIT software. Labs can easily manage their slide inventory and access digital versions of their slides from anywhere.

Manual slide scanning process with HSA SCAN M, integrated with HSA KIT

The software’s user-friendly interface allows for easy navigation and customization of scanning settings to meet specific needs. Additionally, the integration with HSA KIT software allows for advanced image analysis and interpretation, providing valuable insights for research and diagnosis. Overall, HSA SCAN M software offers a comprehensive solution for labs seeking to streamline their slide digitization process while maintaining quality control. 

A glimple of how to manually scan slides with HSA SCAN M

To automatically scan the slides, the existing microscope can be upgraded to an automated microscope station (HSA SCAN A) for low-cost, high-quality performance in a short period of time.

HSA only requires the dimensions of the microscope and any specifications to include, and a custom-fit stand and motor for the microscope will be delivered. This solution eliminates the need for manual adjustments and ensures consistent results.

This will help in:

  • Scanning accurately and quickly
  • Keeping budget in check: Less expensive than automatic scanners
  • Saving time: Automates tasks that require manual intervention
  • Better workflow and quality results: reduced risk of errors or inconsistencies
  • Intuitive, user friendly controls: Anyone with limited experience with microscopes can operate

A glimpse of intcorporating HSA KIT into routine analysis procedures, bringing more accurate and precise results; while also saving time and resources.

For more information or placing an order, contact :

Note: This website will be updated in future.