A new “image analysis pipeline” is providing scientists with new insights into how disease or damage has affected the body down to the cellular level.
TDAExplore combines microscopy’s detailed imaging with topology, a hot area of mathematics that provides insight into how things are arranged, and artificial intelligence’s analytical power to provide, for example, a new perspective on changes in a cell resulting from ALS and wherein the cell they occur, according to Dr. Eric Vitriol, cell biologist and neuroscientist at the Medical College of Georgia.
They report in the journal Patterns that it is an “accessible, powerful option” for using a personal computer to generate quantitative, measurable, and thus objective information from microscopic images, and that it could be applied to other standard imaging techniques like X-rays and PET scans as well.
“We think this is exciting progress into using computers to give us new information about how image sets are different from each other,” Vitriol says. “What are the actual biological changes that are happening, including ones that I might not be able to see, because they are too minute, or because I have some kind of bias about where I should be looking.”
Computers, at least when it comes to evaluating data, have our brains beat, according to the neuroscientist, not just in terms of objectivity, but also in terms of the amount of data they can examine.
Because computer vision, which allows computers to extract information from digital images, has been around for decades, he and his colleague and co-corresponding author Dr. Peter Bubenik, a mathematician at the University of Florida who specializes in topological data analysis, decided to combine the detail of microscopy with the science of topology and the analytical power of AI. According to Vitriol, topology and Bubenik were crucial.
Topology is “ideal” for image analysis because images are made up of patterns, or objects arranged in space, according to him, and topological data analysis (the TDA in TDAExplore) aids the computer in recognizing the lay of the land, in this case, where actin, a protein and essential building block of the fibers, or filaments, that help give cells shape and movement, has moved or changed density.
It’s a time-saving approach because it only needs 20 to 25 photographs to teach the computer how to detect and classify them, rather than hundreds.
The machine is now learning the visuals in patches, which adds to the magic. They explain that breaking down microscope images into these components allows for more accurate categorization, less computer training on what “normal” looks like, and eventually the extraction of relevant data.
Microscopy, which allows scientists to examine things that aren’t visible to the naked eye up close, generates gorgeous, detailed photos and dynamic videos that are a staple for many scientists. “You can’t have a college of medicine without sophisticated microscopy facilities,” he says.
Vitriol, on the other hand, requires a careful study of the images to determine what is normal and what happens in disease situations, such as the quantity of filaments; where the filaments are in the cells near the edge, the center, or spread throughout; and whether some cell sections have more.
We think this is exciting progress into using computers to give us new information about how image sets are different from each other. What are the actual biological changes that are happening, including ones that I might not be able to see, because they are too minute, or because I have some kind of bias about where I should be looking.
Dr. Eric Vitriol
The patterns that emerge in this scenario inform him where actin is and how it’s arranged, which is important for its function, as well as where, how, and whether it’s changed as a result of sickness or damage.
When he examines the actin clumping at the borders of a central nervous system cell, for example, he can see that the cell is spreading out, moving about, and sending out projections that constitute its leading edge. In this example, the cell can expand out and stretch its legs after being basically inactive in a plate.
Scientists evaluating photos directly and estimating what they see have a number of challenges, including the fact that it takes time and that even scientists have biases.
For example, with so much going on, their gaze may be drawn to the familiar, in Vitriol’s instance, actin near the cell’s leading edge. As he examines the black frame surrounding the cell’s periphery, which plainly indicates actin clumping, he may conclude that this is the main point of action.
“How do I know that when I decide what’s different that it’s the most different thing or is that just what I wanted to see?” he says. “We want to bring computer objectivity to it and we want to bring a higher degree of pattern recognition into the analysis of images.”
AI is known to be able to “classify” things, such as recognizing a dog or a cat even if the image is blurry, by learning many millions of factors connected with each animal until it recognizes a dog when it sees one, but it can’t tell why it’s a dog.
That strategy, which necessitates a large number of photos for training and still does not yield many image statistics, does not suit his needs, therefore he and his colleagues created a new classifier that is limited to topological data analysis.
The final line, he says, is that TDAExplore’s unique coupling efficiently and objectively tells scientists where and how much the perturbed cell image differs from the training, or normal, image, information that also leads to new ideas and research areas.
Returning to the cell image with actin clustering along its perimeter, TDAExplore revealed that while the “leading edge” was clearly different with perturbations, some of the most significant changes occurred inside the cell.
“A lot of my job is trying to find patterns in images that are hard to see,” Vitriol says, “Because I need to identify those patterns so I can find some way to get numbers out of those images.”
His main goals are to figure out how the actin cytoskeleton, which the filaments maintain and which provides scaffolding for neurons, works and what goes wrong in diseases like ALS.
The investigators note that some of the machine learning models that require hundreds of photos to train and categorize images don’t specify which section of the image contributed to the categorization. A supercomputer is required to analyze such large volumes of data, which might comprise up to 20 million variables.
Instead, the new technique uses a small number of high-resolution photos to characterize the “patches” that lead to the chosen classification. The new picture analysis pipeline can be completed by the scientist’s standard personal computer in a matter of minutes.
According to him, TDAExplore’s unique approach objectively tells scientists where and how much the perturbed image differs from the training image, the knowledge that leads to fresh ideas and research areas.
The capacity to extract more and better information from images implies that information generated by fundamental scientists like Vitriol is more accurate, which often leads to changes in what is considered the realities of an illness and how it is treated. That could include being able to spot changes that were previously unnoticed, such as those identified by the new mechanism inside the cell.
Scientists already utilize stains to improve contrast before using software to extract information about what they’re seeing in the photos, such as how actin is structured into larger structures, he says. “We had to come up with a new way to get relevant data from images and that is what this paper is about.”
TDAExplore is described in detail in the published article, and other scientists can utilize it as well. The National Institutes of Health (NIH) funded the study.