Example dataset and results for Tui, a multi-generational and expert-correctable tracker for cellular dynamics

# Tui Example Datasets and Tracking Results

This repository provides example datasets and tracking results used to demonstrate the functionality of **Tui**, a multigenerational and expert-correctable cell tracking framework designed to reconstruct complex lineage dynamics including mitosis and fusion events.

The repository contains both **synthetic benchmark datasets** and **experimental live-cell imaging datasets**, along with segmentation masks and example tracking outputs generated by the Tui framework.

These resources are intended to facilitate **method reproducibility, benchmarking, and demonstration of the Tui tracking workflow**.

—

# Repository Contents

## 1. Synthetic_200frames_41mitosis_1fusion.zip

Synthetic benchmark dataset containing simulated cell trajectories with controlled lineage events.

Dataset characteristics:

– 200 time-lapse frames
– 41 mitosis events
– 1 programmed fusion event
– full ground-truth lineage annotation in CTC format

This dataset simulates realistic cell behaviors including:

– migration
– mitosis (1→2 events)
– fusion (2→1 events)
– appearance and disappearance

The dataset allows validation of tracking algorithms under **known ground-truth lineage structures**.

—

## 2. synthetic_cell_6fusions.zip

Synthetic dataset specifically designed to test **fusion tracking capability**.

Dataset characteristics:

– multiple controlled lineage topologies
– 6 programmed fusion events
– 34 mitosis events
– 200 frames

This dataset demonstrates the ability of the Tui ILP-based tracker to correctly resolve **M→1 fusion events and 1→2 mitosis events simultaneously**.

—

## 3. DIC-C2DH_HeLa_dataset.zip

Experimental dataset from the **Cell Tracking Challenge (CTC)**:

Dataset characteristics:

– Differential interference contrast microscopy (DIC)
– 83 frames per sequence
– cell population growing from ~12 to ~21 cells per field of view
– 9 mitosis events
– lineage depth up to 2 generations
– confluency increasing from ~39.6% to ~62.2%

The dataset includes:

– microscopy images
– Detectron2-generated segmentation masks
– CTC ground-truth lineage annotations

This dataset represents **moderately dense proliferating cell populations** used to evaluate lineage reconstruction under realistic microscopy conditions.

—

## 4. T98G_electrotaxis.zip

Experimental dataset of **T98G glioblastoma cells undergoing electrotaxis** under a direct current electric field.

Dataset characteristics:

– 37 frames (~6 hours)
– ~70–72 cells per field of view
– 8 mitosis events
– lineage depth up to 2 generations
– confluency ~18–25%

The dataset includes:

– time-lapse microscopy images
– human-curated segmentation masks
– manually cur