ELEPHANT

ELEPHANT performs 4D cell tracking (3D space over time) by combining incremental deep learning with annotation, prediction, and proofreading to generate validated cell lineages from time-lapse microscopy.

Key Features:

• Incremental Deep Learning Approach: Starts from a small set of annotated nuclei and iteratively improves models through successive prediction–validation cycles, enriching training data and increasing tracking accuracy.
• Integrated Annotation, Prediction, and Proofreading: Combines annotation, deep learning–based prediction, and proofreading workflows to produce validated cell tracks and lineages.
• Low-annotation Requirement: Addresses limited annotated data by enabling model training and refinement from minimal initial nuclei annotations.
• Performance Validation and Benchmarking: Validated against state-of-the-art methods and demonstrated by tracking cell lineages during leg regeneration in a crustacean across 504 timepoints.

Scientific Applications:

• Developmental Biology: Reconstruction of cell lineages and cellular dynamics in developmental studies using 3D time-lapse microscopy.
• Regenerative Biology: Analysis of tissue regeneration dynamics, exemplified by leg regeneration in a crustacean tracked over 504 timepoints.
• Morphogenesis: Investigation of morphogenetic processes by tracking cellular movements and divisions over time.
• Disease Progression: Examination of cellular behaviors in disease models through validated 4D cell tracking.

Methodology:

Begin with a minimal set of annotated nuclei, apply deep learning algorithms that are iteratively improved, and perform cycles of prediction followed by validation to refine the model and expand the training dataset.