New AI Device Can Transform Microscopy


The University of Gothenburg developed an AI device that offers new possibilities for examining images captured with microscopes. Research shows that the device, which has already received international attention, can radically transform microscopy and pave the way for inventions and fields of use within both research and industry.

The research is focusing on deep learning, a combination of machine learning and artificial intelligence (AI) that we all interact with daily, often without considering it. For example, when a new song on iTunes pops up related to songs we have earlier listened to or when our smartphone automatically gets the most desirable settings and improves tones in a photo.

According to Benjamin Midtvedt, the author of the research, Deep learning has a great impact on the world and has had an influence on numerous scientific fields, industries, and areas. We have now designed a device that makes it feasible to use the incredible capability of deep learning, with a focus on images captured by scopes.

Deep learning can be defined as a mathematical model used to solve queries that are challenging to tackle using conventional algorithmic methods. In microscopy, the problem is to recover as much data as possible from the data-packed images, and this is where deep learning has shown to be very useful.

The AI device that Midtvedt and his research associates have developed includes neural networks learning to recover precisely the data that a scientist needs from an image by scanning through a large number of images, known as training data. The device simplifies the method of composing training data compared with having to do so manually so that thousands of images can be produced in an hour rather than a hundred in a month.

AI device stages

“This makes it feasible to immediately extract more data from microscope images without requiring to create complex analysis with conventional methods. Besides, the outcomes are reproducible and customized, specific data can be recovered for a particular purpose.”

For example, the device enables the user to choose the size and material properties for very minute particles and to quickly count and classify cells. The scientists have already displayed that the device can be utilized by industries that require filtering their emissions since they can see in real-time whether all unwanted particles have been filtered out.

Scientists are confident that in the future the device can be used to track viruses in a cell and map cellular defense mechanisms, which would open up immense opportunities for new medications and treatments. Despite the microscopic difficulties, scientists can now more quickly handle studies, make discoveries, execute plans and break new ground within their fields.

Journal Reference:

Benjamin Midtvedt, Saga Helgadottir, Aykut Argun, Jesús Pineda, Daniel Midtvedt, Giovanni Volpe. Quantitative digital microscopy with deep learningApplied Physics Reviews, 2021; 8 (1): 011310 DOI: 10.1063/5.0034891


Leave a Reply

Your email address will not be published. Required fields are marked *