Eirik Magnussen's research uses deep learning to understand and interpret infrared spectroscopic data more effectively. This could provide us with more information, cheaper and faster.
First, let's look at what deep learning is. Deep learning is a type of artificial intelligence that learns to recognize patterns and correlations in large amounts of data. In Magnussen's case, he has used deep learning to process infrared imaging data 1,000 times faster than previous methods.
"This means that we can extract information from data that we previously were unable to retrieve," says Magnussen, who is doing his doctoral degree at the Norwegian University of Life Sciences (NMBU).
He has also succeeded in performing something called quantitative microspectroscopy. This allows us to see how fatty acids are distributed in two dimensions.
More information about cells and tissue without extra cost
The main goal of Magnussen's research has been to use knowledge about electromagnetic radiation and how it interacts with biological materials. This can give us more information from infrared spectra of cells and tissues.
"Infrared spectroscopy is a fast and cheap method that gives us a lot of information about the samples we study," says Magnussen.
The method is already widely used in various scientific disciplines and in the industry. Slowly, the method is also making its way into medical clinics as a cheap and reliable alternative to traditional diagnostic methods.
Magnussen has come up with a simple and cheap way to retrieve information from spectral data that we previously were unable to get.
"This is particularly interesting because it works with data collected with conventional instruments. This means that we can get more information from previously measured spectra without any extra cost," he says.
Eirik Magnussen defends his PhD thesis "Deep learning-assisted analysis of infrared microspectroscopic data for enhanced chemical and optical characterization". Read more about the event here.