KI.SH: An open network for artificial intelligence in Schleswig-Holstein

Artificial intelligence applications and technologies are increasingly finding their way into our lives. This also opens up considerable economic opportunities for companies in the North. In order to make better use of these opportunities, KI.SH (Artificial Intelligence in Schleswig-Holstein) aims to provide an open network that promotes the spread of AI applications and technologies in Schleswig-Holstein and facilitates access to science and the latest research. Let's find out together at KI.SH how we can use the full potential of AI for specific challenges in order to find practical, tailor-made solutions together.  

Fraunhofer MEVIS will be represented with the following contribution:

Workshop
 

“Medical image analysis with image registration and foundation models”

(Johannes Lotz, Daniel Budelmann, Sven Kuckertz, Stefan Heldmann)

Workshop 2.6, Block 2 | 13:30 – 14:15 pm
 

The Fraunhofer MEVIS team from Lübeck will be presenting various applications of image registration and AI in the field of histopathology and radiology: 
 

1. The AI navigation system for diagnostics in histopathology enables pathologists to make faster findings by identifying relevant areas only once and then automatically detecting them immediately in further tissue sections. A software prototype is currently under development and can be tried out as part of a demo. 

2 HistokatFusion combines different immunohistochemical stains and automatically generates ground truth for training image-based AI. Tedious and expensive manual expert annotations are no longer necessary. The software can be used in the form of a web application (https://about.histo.app) as part of a demo. 

3. efficient 1-click tumour follow-up for radiological diagnostics enables the simple detection of changes in metastases and the automatic visualization of differences in CT images. A prototype based on the MEVIS software platform Satori can be tried out in a demo. 

4. foundation models are pre-trained models that can be easily adapted to new problems. They allow the interactive training of AI algorithms with just a few examples. An open source foundation model for biomedical images (https://github.com/FraunhoferMEVIS/MedicalMultitaskModeling) can be tested as part of a demo.