AI-empowered automated cell segmentation for microscopy images

MEDIAR, a cutting-edge AI solution, excels at identifying cells in microscopy images and was the top performer at NeurIPS 2022. Its ability to accurately identify cells in multiple modalities makes it a valuable tool for researchers

Article | Spring 2023

Identifying cell organisms in microscopy images is an essential task for various biomedical applications, and is often the first step in extracting meaningful biological signals. However, developing algorithms that can accurately segment cells in microscopy images can be challenging due to a variety of factors that can affect visual modalities in these images. These factors can include different microscopy types, tissue types, staining types, imaging protocols, cell types, cell shapes, and even magnification (as shown in Figure 1).

 

A team of researchers has recently proposed a solution to this problem called MEDIAR, which is a holistic pipeline for cell instance segmentation under multi-modality. MEDIAR combines data-centric and model-centric approaches as the learning and inference strategies (as shown in Figure 2), and won the NeurIPS 2022 Cell Segmentation challenge with a large margin, beating the second-place team with an F1-score above 0.9. (as shown in Figure 3, team OSILAB).

 

The success of MEDIAR is notable because it is able to perform cell instance segmentation evenly well across different modalities, making it a promising solution for large-scale simultaneous comprehensive analysis of microscopy images. The researchers have made the source code and trained the model for MEDIAR available as open-source, which will likely facilitate further research in this area.

 

Overall, the development of MEDIAR represents an important step forward in the field of computational biology, and has the potential to greatly improve the accuracy and efficiency of cell instance segmentation in microscopy images.