Researcher Profile: Carolina Wählby

Today, biomedicine experiments commonly result in thousands of microscopy images. In order to glean and interpret information about cells, tissues and organisms, automated image analysis methods must be used. In a few hours, computers can perform analyses that would take a person several years to make manually, according to Carolina Wählby.

Using digital image analysis, I and my colleagues are able to rapidly measure how cells react to a wide range of treatments. After all, it's impossible to test 250 different drugs on a patient, but we can extract the patient's cells and grow them in tiny holes, or wells, in multi-well plates. Then, we apply different drugs to different wells and depict the cells through a fluorescence microscope, she explains.

From this, we can see that known drugs in use today have differing effects on a cellular level, and sometimes differing effects on cells from different patients. In order to measure these effects, we need advanced image analysis software.

The computer applications Carolina Wählby develops are based on mathematical algorithms able to identify objects and measure shapes, colours and patterns. Here, observable properties are translated into numbers that reveal connections the human eye can easily miss.

We try to quantify changes in cells and tissue in varying stages of disease, from cancer to pigmentation changes, and how UV light affects the cells of the cornea. In one of our projects, we've also come up with a few new image filtration algorithms and methods to quickly and efficiently find image objects of a given size.

But no matter how efficient and reproducible the methods may be, they still require compromise, Carolina Wählby says.

It's difficult to get both the image acquisition and the analysis just right, as a lot of it is subject to biological variation. That's why our methods have to be robust. Even if you have cells that are clones of one another, they aren't the same. Nature itself is not just ones and zeroes, after all.

It was the desire to measure cause and effect that led her down the path to image analysis once upon a time. After studying Molecular Biotechnology Engineering in Uppsala, she did her degree project at Karolinska Institutet in Stockholm. There, her studies had her growing cells that were photographed in a microscope.

One day, my supervisor said "we need to add some numerical data, you can print the images on a sheet of paper, and then you can have the lab assistant count the dots in the image." But I couldn?t bring myself to do it. the lab assistant was this fantastic elderly lady whom I respected greatly. Asking her to sit around counting dots was out of the question to me, Carolina Wählby says, laughing.

She decided to learn digital image analysis on her own, and enrolled in a course at the Centre for Image Analysis at Uppsala University. This was followed by a doctoral studentship in digital image analysis under Professor Ewert Bengtsson at the Department of Information Technology. Today, she heads a research team of her own at the same department.

At our laboratory, we do a lot of work based on image data produced at SciLifeLab. In addition, we also collaborate with a dozen other researchers in Sweden and the rest of the world. To name but one example, we're doing work on model organisms for large-scale clinical trials. One of our organisms is the 1-millimeter-long nematode Caenorhabditis elegans which we use to find new anti-infectious therapies and genes involved in fat metabolism.

Carolina Wählby formed several of her research connections during her six years at the Broad Institute of Harvard and MIT in Cambridge, Massachusetts, USA. In her role as project lead, she was involved in what she considers to be her greatest contribution to the research community, CellProfiler.

It's a piece of software that can be used even by biomedicine researchers with no background in IT to measure and analyse microscopic changes at a cellular level. The method for finding cells, the first step before they can be measured, is based on my doctoral thesis.

CellProfiler can be run on a regular PC or Mac, and is free to download and use. As the software is open source, anyone is welcome to improve it.

It feels great to put these methods at the disposal of so many. Every day, two new scientific articles using data produced by the software are published, on average. In this way, we advance science as a whole and can benefit several disciplinary domains at once.

In another high-profile project, her research team in Uppsala has worked with Mats Nilsson's group at the SciLifeLab and Stockholm University. Their collaboration has resulted in a new way to measure where various genes are active in a given tissue. To visualize gene expressions at different resolutions, the researchers use tools similar to Google Maps, where gene expressions can be shown or hidden in the same manner as roads and road names when looking at aerial photos, Carolina Wählby explains.

Things get even more interesting when you are trying to find differences between normal tissues and cancer and learn more about what sort of cells, for instance immune cells, are active on the surface of a tumour. The tool, which has been given the working title Tissue Maps, enables users to study the results at different resolutions and to display different sorts of information in the form of layers on top of the original image.

She concludes:

Tissue Maps can provide important answers when it comes to individual variations in cancer tumours, hopefully lead to better understanding of the onset of the disease, and ensure that patients are given the best possible treatment.

Interview by Anneli Björkman, 2015