A joint research team led by Hong Kong City University (CityU) has developed a new computing tool that can reconstruct and visualize three-dimensional (3D) shapes and time changes of cells, speeding up the analysis process from hundreds of hours to hand over to a computer for several hours.
By revolutionizing the way biologists analyze imaging data, this tool can advance further research in developmental and cell biology, such as cancer cell growth.
The interdisciplinary study was led by Professor Yan Hong, Chair of Computer Engineering and Wong Chung Hong, Professor of Data Engineering at the Department of Electrical Engineering (EE) at CityU, along with biologists from Hong Kong Baptist University (HKBU) and Beijing University.
Their findings were published in a scientific journal Nature Communications, by name “Establishment of morphological atlas of Caenorhabditis elegans embryo using 4D segmentation based on deep learning“.
The tool the team developed is called “CShaper”. “It is a powerful computational tool that can systematically segment and analyze cell images at the single-cell level, which is essential for studying cell division and cell and gene functions,” Professor Yan described.
A bottleneck in analyzing a huge amount of data on cell division
Biologists are investigating how animals grow from a single cell, a fertilized egg, into organs and the whole body through countless cell divisions. They especially want to know genetic functions, such as specific genes involved in cell division to form different organs or what causes abnormal cell divisions that lead to tumor growth.
The way to find the answer is to use the gene knockout technique. With all the genes present, the researchers first get pictures of the cells and the vine tree.
They then “knock out” (remove) the gene from the DNA sequence and compare the two vine trees to analyze changes in the cells and infer gene functions. They then repeat the experiment with other genes that have been dropped.
A collaborating team of biologists was used in the study Caenorhabditis elegans (C. elegans) embryos to produce terabytes of data for Professor Jan’s team to perform computational analysis. C. elegans is a type of worm that shares many essential biological characteristics with humans and provides a valuable model for studying the process of tumor growth in humans.
“With an estimated 20,000 genes C. elegans, this means that almost 20,000 experiments would be required if one gene was ejected simultaneously. And the data would be huge. Therefore, it is necessary to use an automated image analysis system. And this leads us to develop more efficiently, ”he said.
Breakthrough in automatic segmentation of cell images
Cell images are usually obtained by laser scanning. Existing image analysis systems can only detect cell nuclei well with poor cell membrane image quality, which interferes with the reconstruction of cell forms.
Also, a reliable algorithm for segmenting temporal 3D images (i.e. 4D images) of cell division is lacking. Image segmentation is a critical process in computer form that involves dividing visual input into segments to simplify image analysis. But researchers have to spend hundreds of hours manually labeling many cell images.
The breakthrough in CShaper is that it can detect cell membranes, build cell shapes in 3D, and more importantly, automatically segment cell images at the cell level. “Using CShaper, biologists can decipher the contents of these images within a few hours.
It can characterize cell shapes and surface structures and provide a 3D representation of cells at different time points, ”said Cao Jianfeng, a doctoral student in Professor Yan’s group and co-author of this paper.
To achieve this, the deep learning DMapNet model plays a key role in the CShaper system.
“By learning to capture more discrete distances between image pixels, DMapNet draws the contour of the membrane by taking into account shape information, not just intensity characteristics. Therefore, CShaper achieved 95.95% accuracy in cell identification, which far surpassed other methods,” he explained.
With CShaper, the team generated a 3D atlas of cell morphology time intervals for C. elegans embryo from stage 4- to 350 cells, including shape, volume, surface area, migration, nucleus position, and cell contact with confirmed cell identity.
Progress in further tumor growth studies
“As far as we know, CShaper is the first computational system for image segmentation and analysis C. elegans embryo systematically at the single-cell level, ”Mr. Cao said.“ Through close collaboration with biologists, we have proudly developed a useful computational tool for automated analysis of large amounts of cell imaging data.
We believe it can promote further studies in developmental and cell biology, especially in understanding the origin and growth of cancer cells, ”added Professor Yan.
They also tested CShaper on plant tissue cells, showing promising results. They believe that computer tools can be adapted to other biological studies.
Hong Kong City University
Cao, J., and others. (2020) Establishment of the morphological atlas of the Caenorhabditis elegans embryo using 4D segmentation based on deep learning. Nature Communications. doi.org/10.1038/s41467-020-19863-x.