Software for Digital microscopy

Learn More About Us
our-business

Our business

For pre- and clinical research and diagnostic purposes, it’s important that medical images are not just images, diagnosis reports are not just unstructured information, molecular data are not just separated from results of image analysis. Researchers would be supported best, if suspicious or malicious structures of tissue samples are identified at the early stages. In that case the researchers and doctors be able to apply their expert knowledge more efficiently and to focus on the illness itself.
This is exactly, what the HS Analysis GmbH from Karlsruhe (GER) does science 2015, when it comes to automated image analysis, structure reports and bring them with knowledge from molecular data. The integration of these approaches in the hardware and its infrastructure is one of the main skills of HS Analysis.
The company is working worldwide closely with different partners together, when digitization, digitalization and automation of medical data is going to be a part of daily routine in diagnosis and research with AI techniques.
With the help of modern AI, Deep Learning and Active Learning supported methods HS Analysis, believes to seriously inspire quantification and prediction in diagnosis, CDx and research areas such as the evaluation or early detection of e.g. cancer.


Technology

Autonomous AI Systems as stand alone or only for the LAN in a lab or in the cloud


Scope

Combine deep learning for image and molecular analysis with natural language processing as a perfect assistant for you



Projects & Services

get knowledge from your fluorescence, bright field images and EM data – NGS, PCR, NMR data – DB and textual data



technology

Technology

The most up-to-date artificial intelligence is one key technology of automatic interpretation of tissue samples in the software of HS Analysis. Technically speaking, virtual slides are analyzed with deep learning. Virtual slides are digital images, that are produced by microscope roboters called slide scanner. In this way you quickly get quantitative results about effects of drug candidates on tissue.

Specialists of HS Analysis use the appropriate technology depending on requirements of customers.
Two examples of most used technologies are listed below.

Digital microscopy is a promising method for quantitative automatization in pharmaceutical research. The key technology was the introduction of Whole-Slide Imaging (WSI) or well plates with digital slide Scanners or HCS. They allow virtualization of the whole slide or well plate with large tissue sections or cell cultures in high microscopic resolution. Combination of WSI or well plates with automatic detection and quantification software enables numerous new possibilities and methods to search for drugs against diseases. HS Analysis develops automatic WSI and well plates quantification solutions for efficient workflows in pharmaceutical research. Our solutions are compatible with all slide scanner or HCS file formats.
Recent evolution in artificial neural networks allows reliable recognition of structures with complex morphology traits, similar like the human brain does. HS Analysis specialized in neural network applications and Deep Learning to detect your Region of Interest in automatic manner. Avoid manual annotations for ROI definition to achieve high throughput with short analysis times.

details

Details

Cell tracking and analysis in real time

HSA in-house developed deep learning


Tracking of cellular morphology changes and behavior in concert with surrounding cells. Cells with increasing area are marked blue, decreasing area red. Length of individual cell borders, circumference, area and percentile change, compared to previous frame (Δ area, Δ circum.), are tracked in real time.

Self developed AI networks:
  • + Cell border gaps in the original image are closed intelligently
  • + Automatic tracking of length of individual cell borders, cellular morphology changes and behavior in concert with surrounding cells
  • + Visual representation of selected morphology changes (Area, circularity, neighbor cell behavior etc.)
  • + Simple export of all statistical data into an Exel file

 

Standard Deep Learning U-Net


Simple intensity dependent annotation of cell borders vs original image





Old U-Net based AI network:
  • – Gaps in the original image are not closed, therefore subsequent statistical analysis is impaired
  • – Tracking of cellular parameters difficult or only possible by additional programs
  • – No visual representation of selected changes in cellular morphology



Counting and segmentation of
cells in chamber slides

Detection and segmentation of nuclei and cytoplasm in cultured cells in chamber slides.




Segmentation of Tumor
and Stroma cells

Segmentation of nuclei and cytoplasm in tumour cells and cytoplasmic regions in stroma cells. If nuclear staining is not present or unspecific, nuclei can be detected by nuclear “holes” in cytoplasmic staining.



Kidney Tool

Segmentation of glomeruli and tubuli by using a Deep Learning model for brightfield data.



Kidney Tool

Segmentation of glomeruli and their compartments by using a tensorflow model for brightfield data.



Cell Counting

Segmentation and counting of cells.  






Endoplasmic Reticulum

The endoplasmic reticulum (ER) is a dynamic structure consisting of branched domains and tubular sections. Automatic segmentation opens up new possibilities for quantitative analysis and efficient representation of this membrane structure at the ultra-structural level.