PathoScan project: Automated digitalization in routine pathology

April 20, 2020

Erlangen: The Fraunhofer Institute for Integrated Circuits IIS is working together with partners PreciPoint GmbH, HTI Automation GmbH, the Department of Pathology at the University Hospital Regensburg and the Technical University of Munich to create an automated digital pathology system. Fraunhofer IIS intends to find a way of assessing the quality of tissue samples and weeding out any unsuitable ones before the samples are loaded for scanning. This will save valuable time by preventing unnecessary scans and means that pathologists won’t end up with substandard specimens.

Fully automated digitalization workflow
© Fraunhofer IIS
Fully automated digitalization workflow makes routine tasks in pathology labs easier.

Up to now, much of the work done in pathology labs has been performed manually. This includes growing and dying tissue samples, analog quality control of the histological sections and creating and storing image libraries. It is an area in which digital innovation has been slow to make inroads so far due to a lack of reliable analysis results and the high costs involved. This is where the PathoScan project comes in. The fully automated digitalization workflow covers a host of pathology processes – from tissue dying to definitive diagnosis. Fraunhofer IIS is developing processes based on artificial intelligence (AI) that assess how well specimens have been dyed and cut. This means specimens that have been unsuccessfully dyed or those with tears or air pockets can be discarded during sampling rather than later, at the analysis stage, once the specimens have already been captured digitally. This saves time and money.

Modular AI solution makes routine tasks easier

Fraunhofer IIS has many years of experience working on automated microscopy systems and using machine learning to analyze medical images. In addition to feature-based classification approaches, this includes deep learning, a data science method that uses convolutional neural networks (CNNs). Machine learning can help pathologists attain reliable analysis results and, because it delivers consistently high image quality, enables them to make a definitive diagnosis.

The digitalization system’s modular setup ensures that it can be easily integrated into existing pathology practice. What’s more, an innovative dye technology for applying dye reagents reduces the need for expensive reagents, especially when it comes to immunohistochemical applications.

The PathoScan digitalization project with a total volume of 3.84 million euros is sponsored by the Bavarian Ministry of Economic Affairs, Regional Development and Energy with funding totaling 1.63 million euros (FKZ ESB074/005). Fraunhofer IIS is joined in this innovative consortium of industrial companies and research institutions by the Department of Pathology at the University Hospital Regensburg, the Technical University of Munich (TUM), PreciPoint GmbH and HTI Automation GmbH.