Temporal models for mitotic phase labelling.

Area of Science:

• Cell Biology
• Computational Biology
• Microscopy

Background:

• Time-lapse microscopy is crucial for understanding cellular functions and cell cycle progression.
• High-throughput analysis of microscopy data requires efficient segmentation and labeling tools.
• Identifying distinct stages of cell division (mitotic phases) is essential for cell cycle studies.

Purpose of the Study:

• To develop an automated system for segmenting and labeling mitotic phases in time-lapse microscopy data.
• To utilize temporal models that capture cell behavior across entire mitotic phases, not just single frames.
• To compare the performance of Dynamic Time Warping, Hidden Markov Models, and Semi Markov Models for this task.

Main Methods:

• Development of a novel automated system for cell segmentation and mitotic phase labeling.
• Application of temporal models including Dynamic Time Warping, Hidden Markov Models, and Semi Markov Models.
• Introduction of a new loss function for Semi Markov Models to enhance robustness to annotation inconsistencies.
• Testing model performance under varied experimental conditions to assess biological robustness.

Main Results:

• The developed system demonstrates effective automated segmentation and mitotic phase labeling.
• Temporal models, evaluated over whole phases, capture distinctive cell division behaviors.
• The proposed Semi Markov model with a new loss function shows improved robustness.
• Model performance was validated across different experimental conditions.

Conclusions:

• The presented system offers a robust solution for high-throughput analysis of cell cycle progression from time-lapse microscopy.
• Temporal modeling provides a powerful approach for accurately identifying mitotic phases.
• The novel Semi Markov model enhances the reliability of automated cell cycle analysis.