Machine learning models for segmentation and classification of cyanobacterial cells.

Area of Science:

• Microbiology
• Computational Biology
• Biotechnology

Background:

• Timelapse microscopy enables single-cell analysis of cyanobacteria metabolism and physiology.
• Accurate segmentation of individual cyanobacteria in dense colonies is challenging for traditional methods due to low contrast.
• Existing algorithms struggle with identifying cells in complex, high-density cyanobacterial populations.

Purpose of the Study:

• To develop and validate machine learning (ML) models for precise segmentation and classification of cyanobacteria cells.
• To introduce Cypose, a software package integrating ML models for cyanobacteria image analysis.
• To overcome limitations of conventional segmentation techniques in analyzing dense cyanobacterial cultures.

Main Methods:

• Utilized the Cellpose framework for developing ML-based segmentation models for cyanobacteria.
• Developed a convolutional neural network, Cyclass, for classifying cellular phenotypes.
• Compared the performance of ML models against traditional segmentation methods.

Main Results:

• Cypose demonstrated superior performance in segmenting individual cyanobacteria cells, including those with diverse morphologies.
• The models successfully differentiated between live and lysed cells and were robust to imaging artifacts like debris.
• The Cyclass model accurately classified cellular phenotypes directly from images.
• Achieved improved cell segmentation accuracy for dense cyanobacterial colonies and filamentous forms.

Conclusions:

• Cypose provides the first ML-based solution for robust cyanobacteria cell segmentation and classification.
• The developed models enhance the accuracy and efficiency of high-throughput analysis of cyanobacteria.
• This advancement facilitates deeper understanding of cyanobacteria physiology and metabolism at the single-cell level.