Quantitative Analysis of Artificial Intelligence Organelles

2023-07-19

The high-brightness X-rays of BESSY II can be used to generate micrographs with spatial resolution as low as a few tens of nanometers. It is possible to examine the entire cellular volume without the need for complex sample preparation, as required by electron microscopy. Under X-ray microscopy, tiny organelles and their fine structures and boundary membranes appear clear and detailed, even in three-dimensional space. Therefore, low-temperature X-ray tomography is highly suitable for studying changes in cellular structures caused by external triggers. However, until now, the evaluation of three-dimensional tomographic images has mainly required manual and labor-intensive data analysis. To address this issue, a team led by computer scientist Dr. Frank Noé and cell biologist Dr. Helge Ewers, in collaboration with the X-ray microscopy department at HZB, has developed a new self-learning algorithm. This artificial intelligence-based analysis method is based on the automatic detection of subcellular structures, which accelerates the quantitative analysis of three-dimensional X-ray datasets. The three-dimensional images inside biological samples were obtained at the U41 beamline of BESSY II. Dr. Stephan Werner, an X-ray microscopy expert at HZB, said, "In this study, we demonstrated the effectiveness of AI-based cellular volume analysis using mammalian cells in cell cultures with protrusions." Mammalian cell structures are complex, with many different organelles, each performing different cellular functions. Protrusions are membrane extensions primarily used for cell migration. "To observe cells using low-temperature X-ray microscopy, the cell samples must first be shock-frozen, a process that is very rapid and prevents the formation of ice crystals inside the cells. This allows the cells to be in a nearly natural state, enabling us to study the impact of external factors on the internal cell structures," explained Werner. Michael Dyhr, the first author and a researcher at the Free University of Berlin, said, "Our work has generated great interest among experts. Neural networks can correctly identify about 70% of existing cell features in a very short time, allowing for rapid evaluation of the dataset. In the future, we can use this new analysis method to study how cells respond to environmental influences such as nanoparticles, viruses, or carcinogens, faster and more reliably than before." Source: https://phys.org/news/2023-07-quantitative-analysis-cell-organelles-artificial.html

RECENT AI NEWS

RECENT AI TOOLS