The Cell Tracking Challenge: 10 years of objective benchmarking

Area of Science:

• * Biology
• * Computer Science
• * Image Analysis

Background:

• * The Cell Tracking Challenge is a key initiative for evaluating cell segmentation and tracking algorithms.
• * Previous reports established its significance in algorithm development.

Purpose of the Study:

• * To report significant improvements and updates to the Cell Tracking Challenge since 2017.
• * To provide an enriched dataset repository and new benchmarking tools.
• * To analyze algorithm performance, generalizability, and reusability.

Main Methods:

• * Introduction of a segmentation-only benchmark.
• * Expansion of the dataset repository with diverse and complex datasets.
• * Creation of a silver standard reference corpus for deep learning models.
• * Up-to-date leaderboards for cell segmentation and tracking.
• * Analysis of method performance against dataset properties.
• * Studies on the generalizability and reusability of top methods.

Main Results:

• * Enhanced dataset diversity and complexity.
• * A new silver standard corpus beneficial for deep learning.
• * Comprehensive leaderboards reflecting current state-of-the-art.
• * In-depth analysis linking method performance to dataset characteristics.
• * Novel insights into the generalizability and reusability of algorithms.

Conclusions:

• * The updated challenge offers improved resources for cell tracking algorithm development.
• * Performance analysis provides practical guidance for algorithm selection and application.
• * Findings are relevant for both traditional and machine learning-based approaches.