A new “image analysis pipeline” gives scientists a quick new insight into how disease or injury has changed the body, down to the individual cell.
It’s called TDAExplore, which uses the detailed imaging provided by microscopy, pairs it with a hot area of mathematics called topology, which gives insight into how things are organized and the analytical power of intelligence. artificial to give, for example, a new perspective on changes in a cell resulting from ALS and where in the cell they occur, explains Dr. Eric Vitriol, cell biologist and neuroscientist at the Medical College of Georgia.
It is an “accessible and powerful option” for using a personal computer to generate -; measurable and therefore objective -; information from microscopic images that could possibly also be applied to other standard imaging techniques such as x-rays and CT scans, they report in the journal Reasons.
We think this is an exciting advancement in the use of computers to give us new insight into how sets of images are different from each other. What are the real biological changes that are happening, including those that I might not be able to see, because they are too small, or because I have some kind of prejudice about where I should be looking. “
Dr Eric Vitriol, Cell Biologist and Neuroscientist, Medical College of Georgia
At least in the data analysis department, computers make our brains beat, says the neuroscientist, not only in their objectivity but in the amount of data they can assess. Computer vision, which enables computers to extract information from digital images, is a type of machine learning that has been around for decades, which is why he and his colleague and corresponding author colleague, Dr Peter Bubenik, mathematician at the University of Florida and expert on topological data analysis, decided to combine the detail of microscopy with the science of topology and the analytical power of AI. Topology and Bubenik were essential, says Vitriol.
Topology is “perfect” for image analysis because images are made up of patterns, objects arranged in space, he says, and topological data analysis (the TDA in TDAExplore) helps it. computer to also recognize the configuration of the land, in this case the actin -; a protein and an essential building block of fibers, or filaments, which help give shape and movement to cells -; has moved or changed density. It’s an efficient system that, instead of literally taking hundreds of images to train the computer to recognize and classify them, can learn about 20 to 25 images.
Part of the magic is that the computer now learns the images in chunks which they call patches. Breaking down the microscopy images into these chunks allows for more precise classification, less computer training on what “normal” looks like, and ultimately the extraction of meaningful data, they write.
There is no doubt that microscopy, which allows close examination of things invisible to the human eye, produces beautiful detailed images and dynamic videos that are a mainstay for many scientists. “You can’t have a medical school without sophisticated microscopy facilities,” he says.
But to understand first what is normal and what happens in disease states, Vitriol needs a detailed analysis of images, such as the number of filaments; where are the filaments in the alveoli -; near the edge, the center, scattered everywhere -; and if some cell regions have more.
The diagrams that emerge in this case tell him where the actin is and how it is organized -; a major factor in its function -; and where, how and if it changed with the disease or damage.
When he looks at the actin clustering around the edges of a central nervous system cell, for example, the assemblage tells him that the cell is expanding, moving, and sending out projections that become its leading edge. In this case, the cell, which has mostly slept in a dish, can spread out and stretch its legs.
Some of the problems with scientists directly analyzing images and calculating what they see include that it takes time and even scientists have biases.
As an example, and especially with so much action going on, their eyes may fall on the familiar, in the case of Vitriol, that actin on the leading edge of a cell. As he looks again at the dark frame around the periphery of the cell clearly indicating the actin concentrating there, this could imply that this is the main point of action.
“How do I know that when I decide what’s different it’s the most different thing or is it just what I wanted to see?” ” he says. “We want to bring computational objectivity to it and we want to bring a higher degree of pattern recognition in image analysis.”
AI has been known to be able to “classify” things, such as recognizing a dog or cat every time, even if the image is blurry, by first learning several million variables associated with each animal up to what he knows about a dog when he sees one, but he can’t say why it’s a dog. This approach, which requires so many frames for training purposes and still doesn’t provide a lot of frame statistics, doesn’t really work for his purposes, which is why he and his colleagues created a new classifier limited to topological data analysis.
The bottom line is that the unique coupling used in TDAExplore effectively and objectively tells scientists where and to what extent the disturbed cell image differs from the training image, or normal image, information that also provides new ideas and directions of research, he says.
Returning to the image of the cell which shows the clustering of actin along its perimeter, where the ‘leading edge’ was clearly different with the disturbances, TDAExplore showed that some of the biggest changes were in done inside the cell.
“A lot of my job is trying to find patterns in hard-to-see images,” says Vitriol, “because I have to identify those patterns so that I can find a way to get numbers from those images. ” His results consist of determining how the actin cytoskeleton works, for which the filaments provide the scaffolding and which in turn provides support to neurons, and what is wrong with conditions like ALS.
Some of those machine learning models that require hundreds of images to train and classify images do not describe which part of the image contributed to classification, the investigators write. These huge amounts of data that must be analyzed, and which can include around 20 million variables, require a supercomputer. Rather, the new system requires relatively few high-resolution images and characterizes the “patches” that led to the selected classification. In minutes, the scientist’s standard personal computer can complete the new image analysis pipeline.
The unique approach used in TDAExplore objectively tells scientists where and to what extent the disturbed image differs from the training image, information that also provides new ideas and directions for research, he says.
The ability to get more and better information from the images ultimately means that the information generated by fundamental scientists like Vitriol, who often ultimately changes what is considered the facts of a disease and how it is treated are more specific. This could include the ability to recognize changes, like those the new system signaled inside the cell, that were previously overlooked.
Currently, scientists apply stains to allow for better contrast, then use software to extract information about what they see in the images, such as how actin is organized into a larger structure, he says.
“We had to find a new way to get relevant data from images and that is the subject of this article.”
The published study provides all the evidence for other scientists to use TDAExplore.
The research was supported by the National Institutes of Health.
Georgia Medical College at Augusta University
Edwards, P., et al. (2021) TDAExplore: quantitative analysis of fluorescence microscopy images via topology-based machine learning. Reasons. doi.org/10.1016/j.patter.2021.100367.