Jordão Bragantini - ultrack: large-scale versatile cell tracking in Python | SciPy 2024

ULTrack is a new Python-based software for joint cell segmentation and tracking, capable of handling large-scale datasets (up to 4TB)

Combines traditional image processing with optimization techniques instead of relying solely on deep learning approaches

Can process both 2D and 3D microscopy data, supporting multiple cell types and fluorescence imaging modalities

Uses a two-step process: pre-processing for feature extraction followed by clustering and optimization

Handles challenging scenarios like cell division, cells entering/leaving field of view, and cells with varying fluorescence intensities

Achieves state-of-the-art results on the Cell Tracking Challenge benchmark, particularly excelling with traditional image processing approaches

Scales efficiently with memory usage through windowed processing and scheduling schemes

Integrates with popular scientific Python tools (Napari, Dask, SciPy) and supports multiple file formats

Performs well even on laptop hardware for moderate datasets, while supporting HPC deployment for larger datasets

Emphasizes user control and interpretability over black-box ML approaches, allowing parameter tweaking and multiple segmentation hypotheses