Deterministic and probabilistic approaches for tracking virus particles in time-lapse fluorescence microscopy image...

Area of Science:

• Microscopy
• Virology
• Computational Biology

Background:

• Time-lapse fluorescence microscopy allows observation of viral processes.
• Tracking viruses is crucial for understanding their spatial-temporal dynamics.

Purpose of the Study:

• To develop and compare deterministic and probabilistic methods for multiple virus tracking.
• To evaluate tracking performance using synthetic and real microscopy data.

Main Methods:

• Developed deterministic approaches: particle localization (spot-enhancing filter, 2D Gaussian fitting) and motion correspondence (global nearest neighbor).
• Developed probabilistic approaches: mixture of particle filters and independent particle filters with a penalization strategy to prevent filter coalescence.
• Evaluated eight tracking approaches on synthetic image sequences and real HIV-1 microscopy images.

Main Results:

• Probabilistic approaches, particularly those based on independent particle filters, demonstrated superior performance.
• The developed penalization strategy effectively addressed filter coalescence for closely located viruses.
• Compared favorably against deterministic methods and mixture-based particle filter approaches.

Conclusions:

• Independent particle filter-based probabilistic methods offer a superior solution for multi-virus tracking in fluorescence microscopy.
• These advanced tracking techniques enhance the study of viral dynamics and spatial-temporal relationships.