The term colorectal refers to the combination of the colon (large intestine) and the rectum, both of which are digestive system components. The colon absorbs water and electrolytes from digested food, whilst the rectum functions as the lowest portion of the large intestine, leading to the anus. Colorectal health is critical for digestion and waste removal. Colorectal cancer is a form of cancer that begins in the colon or rectum, which are referred to collectively as the colorectal region.

Colorectal cancer is the third most prevalent cancer worldwide and the fourth leading cause of death. It begins with benign polyps in the colon or rectum lining, which can develop into malignant tumors over time.

Risk factors include

  • obesity
  • ow-fruit and vegetable diet
  • physical inactivity
  • smoking

Treatment options include surgery, chemotherapy, radiation therapy, targeted therapy, or a combination. Preventive measures include a healthy lifestyle, exercise, weight management, and early medical intervention.

AI significantly aids in various aspects of colorectal cancer care, including early detection, personalized treatment planning, prognostic predictions, and drug discovery. It enhances medical imaging analysis, risk assessment, and patient data evaluation. By streamlining processes and providing insights, AI contributes to improved outcomes in colorectal cancer management.

HS Analysis GmbH specializes in recognizing cells on the verge of developing cancer and malignancies. However, detection is insufficient without adequate data analysis. The HSA KIT program can extract data into Excel spreadsheets, allowing for complete result display via reports, graphs, and figures.


The HSA KIT software provides an intuitive interface and advanced annotation features for precise detailing down to pixels. It enhances analysis by standardizing processes, maintaining consistency, and enabling reproducibility.

The Base ROI (Region of Interest) is extremely effectively classified by the HyperCRC-Net Module in HSA KIT into many sorts of glands:

  • Healthy tissue
  • Adinoma
  • Carcinoma
  • Other tissues (muscle, adipose or stroma)

In the context of colorectal cancer:

  1. Gesund – Healthy Tissue: Refers to normal, non-cancerous tissue in the colon or rectum.
  2. Adenoma: A precancerous growth or tumor that can develop in the colon or rectum lining. It has the potential to transform into cancer over time.
  3. Carcinoma: Cancer originating from epithelial cells in the colon or rectum. It can invade nearby tissues and potentially spread to other parts of the body.

These terms help differentiate between healthy tissue, precancerous growths, and cancerous tumors in colorectal cancer cases.

A picture of the detection of Healthy tissue and Adenoma.

An exapmle of detecting Carcinoma and Adinoma.

Zoom out view: Many healthy glands were found.

Zoom in view: Many healthy glands were found.

Zoom out view: Numerous Adenoma detected.

Zoom in view: Numerous Adenoma detected.

Zoom out view: Numerous Carcinoma detected.

Zoom in view: Numerous Carcinoma detected.

Staining and Markers

HE (Hematoxylin and Eosin) staining is a crucial histological technique in pathology. By highlighting cellular elements, it helps pathologists assess colorectal cancer using biopsy samples. The process involves staining tissue slices with Hematoxylin for nuclei and Eosin for cytoplasm and matrix, enabling the analysis of tissue properties and architecture. This aids in detecting, grading, and planning treatment for colorectal cancer patients.

Hematoxylin and eosin staining enables the differentiation of cell types within tissue samples and visualization of cellular features and tissue structure.

„Ki-67 staining“ is an immunohistochemistry method that identifies Ki-67 protein expression in cells, indicating cell proliferation. It’s crucial in cancer research, predicting prognosis, and treatment evaluation. Under a microscope, it reveals Ki-67 protein presence, aiding in assessing cell proliferation rates and identifying dividing cells. It’s especially useful in distinguishing tumors from other growths and assessing disease severity risk.

Example Ki67 immunohistochemistry-stained images of colorectal cancer

The HSA KIT offers a holistic approach to detect colorectal cancer from histopathological slides. It utilizes diverse stains and annotated datasets to train the HyperCRC-Net module, emphasizing colorectal cancer features. Details like cell structure, patterns, and Ki-67 expression are gathered for precise predictions. Detection, segmentation, classification, and grading of colorectal cancer rely on histological characteristics, Ki-67 expression, and proliferation rates. Additionally, it provides prognostic information for patient outcomes and treatment decisions.

HSA KIT includes splitscreen function. For the diagnosis you can just open two or more digital slides in the splitscreen function and check the environment of Ki-67 slide and the same position of HE slide.

you can zoom in only one slide and the second slide will change the same zoom level and position. In that way you can easy move and compare for the diagnosis two stanning’s.

In the image you see the Ki-67 stained nuclei and corresponding nuclei in HE stain and can make better decision for the correct diagnosis.

When analyzing histopathological slides, the HSA KIT offers a one-stop solution for detecting colorectal cancer using several stains.

An example of HE and ki67 stains

Deep Learning (DL)

the connection between deep learning, machine learning, and AI. Artificial intelligence (AI) refers to techniques that enable computers to mimic human behavior. Using algorithms that have been trained on data, machine learning enables computers to classify or predict things. Deep learning is a type of machine learning that analyzes data and detects patterns using multi-layered neural networks, much like the way the human brain does. The structure of the human brain served as the inspiration for the neural network’s architecture.

AI vs. machine learning vs. DL.


Whole Slide Imaging (WSI)

Whole Slide Imaging (WSI) began in the late 20th century with slide scanning systems capturing high-resolution images of entire tissue slides. Advances in hardware, software, validation studies, and regulatory approvals facilitated its integration into pathology workflows. WSI allows remote access, collaboration, and digital archiving of histopathological images, improving diagnostic accuracy and efficiency.

Whole Slide Imaging (WSI) is a digital pathology technique that involves scanning entire glass slides containing tissue samples to create high-resolution digital images. This technology has transformed the field of pathology by enabling remote access, sharing, and analysis of histopathological images. WSI offers benefits such as improved collaboration among pathologists, efficient archiving of slides, and the potential for advanced image analysis using computational tools. It has found applications in research, education, and clinical diagnostics, enhancing the way tissue samples are visualized and interpreted.

Feature Extraction

Feature extraction simplifies machine learning by converting input data into meaningful features, reducing redundancy and complexity. It’s connected to dimensionality reduction for handling large datasets effectively. This involves selecting relevant attributes to enhance analysis and classification accuracy, often used in tasks like pattern detection.

Deep Learning, a Machine Learning subset, learns task-specific representations from input data using layers of non-linear units. This hierarchical structure creates abstraction layers. It covers supervised, unsupervised, and semi-supervised learning, benefiting from labeled data and powerful resources like GPUs. This leads to faster training and efficient data processing.

Only ML algorithms require feature extraction, DL algorithms do not

Ground Truth Data (GTD)

„Ground truth“ is a vital concept in statistics and machine learning, representing the correct solution for evaluating models. In machine learning, it signifies accurate labeled data used for training. This process, „ground truthing,“ involves collecting objective data for validation.

In machine learning, ground truth reflects the true essence of a problem through relevant datasets. Supervised models learn from labeled data to predict outcomes for new data. Larger datasets enable learning across diverse scenarios and edge cases.

Creating ground truth involves diverse data collection, precise labeling, and validation through an iterative process. Validation data tests model performance, and balanced augmented datasets contribute to training. Iterative training and fine-tuning refine both data and models. Accurate ground truth data is pivotal for effective model training, necessitating careful planning and continuous improvement.

The GTD used in this project is presented in table below: