Short Trajectory Segmentation With 1D Unet Framework: Application To Secretory Vesicle Dynamics

Area of Science:

• Cellular biology
• Biophysics
• Microscopy imaging

Background:

• Automated techniques are crucial for studying protein transport dynamics within secretory vesicles in living cells.
• Quantitative analysis of vesicle dynamics necessitates accurate trajectory segmentation due to inconsistent movement patterns.

Purpose of the Study:

• To introduce a novel 1D U-Net based framework for automated trajectory segmentation of secretory vesicles.
• To overcome limitations of existing methods, particularly the requirement for long trajectories.

Main Methods:

• Development and application of a 1D U-Net architecture for trajectory segmentation.
• Utilizing data from spinning disk microscopy imaging of protein trafficking in *Drosophila* epithelial cells.
• Segmentation is performed within sliding windows to capture short trajectory segments.

Main Results:

• The proposed 1D U-Net framework achieves 77.7% accuracy in trajectory segmentation.
• The method effectively segments both short (5 points) and long (135 points) trajectories.
• Unlike traditional methods, it does not require long trajectories for effective segmentation.

Conclusions:

• The novel 1D U-Net based approach provides an effective and automated method for quantifying vesicle dynamics through trajectory segmentation.
• This framework enhances the study of protein transport by accurately capturing dynamics even from short cellular tracks.
• The method offers a significant advancement for analyzing vesicle movement in cell biology research.