Automated segmentation of cell organelles in volume electron microscopy using deep learning.

Area of Science:

• Cell biology
• Biophysics
• Computational biology

Background:

• Artificial intelligence (AI) and deep learning are increasingly used for image analysis in life sciences.
• Training AI algorithms requires large, certified labeled datasets, a time-consuming process.
• Accurate instance segmentation of cellular structures is crucial for quantitative biological insights.

Purpose of the Study:

• To develop a rapid, automated method for segmenting and quantifying 3D cell organelles from electron microscopy data.
• To overcome the limitations of manual segmentation in terms of time and expertise required.
• To enable high-throughput quantitative analysis of cellular structures.

Main Methods:

• A deep-learning based approach named FAMOUS (fast automatic outline segmentation) was developed.
• FAMOUS integrates organelle detection, image morphology, and 3D meshing for automated segmentation.
• The workflow was applied to volume electron microscopy datasets, including HeLa cells and yeast cells.

Main Results:

• FAMOUS provides complete segmentation results for unseen datasets within one week.
• The method was successfully applied across different electron microscopy modalities and cell lines.
• FAMOUS demonstrates superior performance compared to manual segmentation in both speed and accuracy.

Conclusions:

• FAMOUS offers a significant acceleration of the 3D cell organelle segmentation and quantification process.
• This AI-driven workflow facilitates high-throughput quantitative cell biology research.
• The method's efficiency and accuracy make it a valuable tool for analyzing complex biological datasets.