Spanish researchers develop computer vision system that automatically analyzes cells
A new computer vision system developed by researchers at Universidad Carlos III de Madrid (UC3M) can automatically analyze biomedical videos recorded using microscopy. The system characterizes and describes the behavior of cells in the images.
The newly developed method was used for measurements on living tissues made by scientists from the National Center for Cardiovascular Research. As a result of this study, the team found that neutrophils, a type of immune cell, behave differently in the blood during inflammatory processes.
In addition, they found that one of the processes is linked to cardiovascular diseases. The study was published in the journal Nature. These developments could enable new treatment methods to minimize the consequences of heart attacks.
"Transcriptional and proteomic profiling of individual cells has revolutionized the interpretation of biological phenomena, providing cellular landscapes of healthy and diseased tissues. These approaches, however, do not describe dynamic scenarios in which cells constantly change their biochemical properties and subsequent 'behavioral' outcomes. Here we used 4D live imaging to record dozens and hundreds of morphokinetic parameters describing the dynamics of individual leukocytes in foci of active inflammation," the study says.
Professor Fernando Díaz de María heads the UC3M Multimedia Processing Group and is one of the authors of the study.
“Our contribution consists of designing and developing a fully automatic system, based on computer vision methods, that allows us to characterize the studied cells by analyzing video recordings made by biologists using intravital microscopy technique”
While traditional biological studies are often confirmed by analyzing several hundred manually characterized cells, the new study included automatic measurements of the shape, size, movements and position of several thousand cells relative to a blood vessel. This enables extended biological analysis with greater statistical significance.
The researchers said the new system has various advantages in terms of both time and accuracy. Iván González Díaz is an associate professor in the Department of Signal Theory and Communications at UC3M. He is also a member of the research group.
“It is impossible for an experienced biologist to segment and track cells in a video for months. On the other hand, to give a rough idea (since it depends on the number of cells and the depth of the 3D volume), our system needs only 15 minutes to analyze a 5-minute video”
Deep neural networks, the tools these engineers rely on for cell segmentation and detection, are basically algorithms that learn from examples, so to deploy the system in a new context, enough examples must be created to train them. These networks are part of machine learning, which in turn is a discipline within artificial intelligence (AI).
In addition, the system includes other types of statistical methods and geometric models, all of which are described in another article recently published in the journal Medical Image Analysis. The software implementing the system is universal and can be adapted to other tasks in a few weeks.
“In fact, we are already applying it in other different scenarios, studying the immunological behavior of T cells and dendritic cells in cancer tissues. And the preliminary results are promising”
"Having analyzed more than 100,000 reconstructions of cell shapes and tracks over time, we obtained behavioral descriptors of individual cells and used these multidimensional datasets to build behavioral landscapes. These landscapes recognized the identity of leukocytes in inflamed skin and trachea and revealed a continuum of neutrophil states inside blood vessels, including a large sedentary state that was covered by the underlying endothelium and associated with pathogenic inflammation. Behavioral screening of 24 mouse mutants identified the kinase Fgr as a driver of this pathogenic state, and intervention on Fgr protected mice from inflammatory damage. Thus, behavioral landscapes report specific properties of the dynamic environment at high cellular resolution," the study says.
In any case, when conducting research in this area, the researchers emphasize the importance of the work of an interdisciplinary team.
“In this context, it is important to acknowledge the prior efforts of communication between biologists, mathematicians and engineers, necessary to understand the basic concepts of other disciplines before real progress can be achieved”
