IsletNet Insights

The application is designed to assist in the process of isolation of human pancreatic islets.

Learn more


Being part of applied research is the thing Iterait is passionate about. That’s why we are really proud of cooperation with Institute for Clinical and Experimental Medicine on an AI-based service dedicated to the experimental treatment of diabetes. Did you know that almost half a billion people suffer from diabetes?

The history of the project is quite rich. As of 2019, some of our team members have been working on IsletNet for years. We started with an initial simple research-oriented proof-of-concept application which has already grown into a mature and stable on-line service.

What is IsletNet

So what is IsletNet? The application is designed to assist in the process of isolation of human pancreatic islets. It is a complex service which enables users to directly upload images of a sample captured by a microscope. These images, together with additional metadata, are automatically processed by our tailor-made neural network that distinguishes between pancreatic islets, their exocrine tissue and image background.

We use this information for the subsequent postprocessing which results in a human-readable report on the sample quality. In addition, the service also provides various machine-readable information and statistics which can be imported into e.g. Excel.

One of the key components of the whole service is its web-based interface. This is crucial for custom image-by-image browsing through the neural network outputs, various visualizations and visual comparison with human annotations.

The AI Behind IsletNet

The core of IsletNet is a deep convolutional neural network processing the input image into a pair of two black&white images. The white pixels in the results images represent positions of the actual islets and exocrine tissue, respectively.

We also apply a various old-school computer vision algorithms in order to fix the network inaccuracies. By using this approach, IsletNet is able to separate even overlapping islets that could be easily mistaken for a single islet.

The best part of IsletNet is that it’s free to use. Anyone can try to analyze their images at Feel free to use this image or any image published at the service website.

Found the article interesting?

Contact us

Fill out your details in the form and we will be in touch within 24 hours.

You can also visit us in our office located at Hybernska 1034/5, Nove Mesto, 110 00 Praha 1.

Or simply send us an email at

Your message has been received!
Oops! Something went wrong while submitting the form. Please try it again.