AnnotatorJ

AnnotatorJ facilitates semi-automatic annotation and segmentation of cellular compartments in 2D microscopy images to generate high-quality training datasets for deep learning and enable downstream quantitative analyses such as expression measurements.

Key Features:

• ImageJ plugin: Operates as a plugin for ImageJ to integrate with microscopy image processing workflows.
• U-Net-based pre-segmentation: Applies U-Net-based pre-segmentation algorithms to propose initial object masks.
• Deep learning integration: Combines deep learning outputs with manual corrections to refine annotations.
• Semi-automatic workflow: Implements a semi-automatic annotation approach that accelerates hand-annotating cells and other objects on 2D microscopy images.
• Training dataset generation: Enables production of large volumes of accurate annotated datasets suitable for training deep-learning models.
• Cell identification and segmentation: Focuses on identification and segmentation of cellular compartments for downstream analyses.
• 2D microscopy image support: Processes 2D microscopy images as the primary input modality.

Scientific Applications:

• Training data creation for deep learning: Produces annotated datasets used to train segmentation models such as U-Net.
• Quantitative cellular analysis: Provides segmentations used for downstream quantitative analyses, including expression measurements.
• Cellular imaging research: Supplies annotations for method development and validation in cellular imaging studies.

Methodology:

Uses U-Net-based pre-segmentation combined with manual annotation in a semi-automatic workflow to produce final segmentations.