Application of image-recognition techniques to automated micronucleus detection in the in vitro micronucleus assay

Hiromi Yoda,Kazuya Abe,Hideya Takeo,Takeji Takamura-Enya,Ayumi Koike-Takeshita

DOI: https://doi.org/10.1186/s41021-024-00305-9

2024-04-25

Genes and Environment

Abstract:An in vitro micronucleus assay is a standard genotoxicity test. Although the technique and interpretation of the results are simple, manual counting of the total and micronucleus-containing cells in a microscopic field is tedious. To address this issue, several systems have been developed for quick and efficient micronucleus counting, including flow cytometry and automated detection based on specialized software and detection systems that analyze images.

genetics & heredity,toxicology

What problem does this paper attempt to address?

The paper attempts to address the time-consuming and error-prone issue of manually counting micronucleated cells in in vitro micronucleus assays. Specifically, the authors propose an automated micronucleus detection method based on image recognition technology to improve the efficiency and accuracy of micronucleated cell counting.

### Background
The in vitro micronucleus assay is a standard genotoxicity test used to detect chromosomal abnormalities caused by clastogenic, aneugenic, and non-aneugenic chemicals. Although the operation and interpretation of results for this technique are relatively simple, manually counting the total cells and micronucleated cells in microscope fields is a tedious task. To address this issue, various rapid and efficient micronucleus counting systems have been developed, including flow cytometry and image analysis methods based on dedicated software and detection systems.

### Methods
The authors propose a simple and effective automated micronucleus counting method, which involves the following steps:
1. **Image Separation**: Separate the RGB channels of color micrographs of cells stained with acridine orange.
2. **Nucleus and Micronucleus Detection**: Detect nuclei and micronuclei by scaling the green image, while the cytoplasm is identified from a composite image of the red and green images.
3. **Cell Recognition**: Identify overlapping cytoplasm and micronuclei as micronucleated cells and display the number of micronucleated cells and the total number of cells.

### Results
The results of this method are comparable to those of manual counting, accurately detecting the total number of cells and the number of micronucleated cells in microscopic images. The accuracy of the micronucleus count depends on the cell staining conditions, but users can achieve optimal results by manually optimizing parameters such as thresholding, noise reduction, and binarization. The optimization process is simple and straightforward, reducing workload and increasing efficiency.

### Conclusion
The micronucleus detection (MN/cell detection software) system developed by the authors can accurately detect the total number of cells and the number of micronucleated cells in microscopic images at the same level of accuracy as manual counting. The software provides easy-to-operate parameter optimization options, allowing users to easily obtain accurate results.