A new, freely available platform helps non-experts use artificial intelligence to analyse microscopy images. The platform will be of big help in research and diagnostics using modern day microscopes.

Software using artificial intelligence, AI, are revolutionizing how microscopy images are analysed. For instance, AI can be used to detect features in images (i.e., tumours in biopsy samples) or improve the quality of images by removing unwanted noise. But AI technologies remain difficult for non-experts to use.

  In the article “Democratising deep learning for microscopy with ZeroCostDL4Mic”, published in Nature Communication on 15 April 2021 (https://www.nature.com/articles/s41467-021-22518-0), researchers describe a platform called ZeroCostDL4Mic, which makes these AI technologies accessible to everyone.

“The key novelty is that ZeroCostDL4Mic runs in the cloud for free and does not require users to have any coding experience or advanced computational skills. Effectively, it runs on any computer that has a web browser,” says Guillaume Jacquemet, Senior Researcher in Cell Biology at Åbo Akademi University and Turku Bioscience Centre.

Example illustrating how AI via ZeroCostDL4Mic can be used to detect the nucleus of cancer cells from microscopy images. Left: original microscopy image, right: image where each detected cancer cell has a different colour.

Over the last 400 years, microscopes have allowed mankind to observe objects that are otherwise too small to be seen with the naked eye. Today, microscopy is a leading technology used worldwide to perform not only research but also diagnostics.

Modern microscopes are directly connected to digital cameras, leading to the acquisition of hundreds to thousands of images per sample. These images need to be processed on a computer to gain meaningful data, which is a huge undertaking.

To help with the number of images, Jacquemet and his colleagues have used AI to train a machine to do the work. In practice, ZeroCostDL4Mic is a collection of self-explanatory notebooks for Google Colab, featuring an easy-to-use graphical user interface.

Example illustrating how AI via ZeroCostDL4Mic can be used to detect the nucleus of cancer cells from microscopy images. Left: original microscopy image, right: image where each detected cancer cell has a different colour.

“We believe that ZeroCostDL4Mic will acts as ‘a gateway drug’ for AI, luring users to explore these new technologies that will transform biomedical research and diagnostics in the decades to come,” says Jacquemet.

The development of the ZeroCostDL4Mic platform was coordinated by the Jacquemet (Åbo Akademi University and Turku Bioscience, Turku, Finland) and Henriques laboratories (Instituto Gulbenkian de Ciência, Oeiras, Portugal). It involved a large international consortium encompassing 12 laboratories, spread across nine countries and two continents.

ZeroCostDL4Mic is freely available online (https://github.com/HenriquesLab/ZeroCostDL4Mic), and a video highlighting examples is available here: https://www.youtube.com/watch?v=hh2I5xJH67k.  The video illustrates how ZeroCostDL4Mic can detect and follow cancer cells in videos and improve the quality and resolution of various microscopy images.

More information:

Guillaume Jacquemet Senior Researcher in Cell Biology at Åbo Akademi University and Turku Bioscience Centre E-mail: guillaume.jacquemet@abo.fi Tel. +358 503235606