A single virtual microscopy platform
The problemSlides analysed one at a time, and images too large for the tools already in use.
The University Medicine Laboratory of the University of Bologna acquires images of human tissue and cells with an Olympus scanner, and keeps them on its own server.
They are large images, with detail that reaches 80 zoom levels. Students and lecturers need to look at them together, and share what they see on the same image.
The platform
It is written in Angular. The handling of the layered images goes through OpenStreetMap.
The software that recognises diseases
On top of the platform we built the second part: software that, with machine learning (in Italian), analyses the images and recognises by itself tumour cells of the liver and pancreas. It is written in TensorFlow.
When it finds something it sends an alert, and from there a lecturer looks at that finding.
What changed for the laboratory
Study work on the images sits in one place. Lecturers and students annotate the same image.
The automatic recognition of cell diseases has been released and runs on the laboratory’s images: it is the piece of dotenv’s AI that sits on real work, not on a demo.