AnnotatorJ: an ImageJ plugin to ease hand annotation of cellular compartments.

Area of Science:

• Bioimage analysis
• Computational biology
• Machine learning in microscopy

Background:

• Accurate cell segmentation is crucial for reliable cellular analysis and downstream applications like expression measurements.
• Deep learning (DL) methods significantly outperform conventional techniques in cell segmentation but require extensive, high-quality annotated data.
• Creating expert-level annotations is time-consuming, expensive, and often restricted by data privacy regulations.

Purpose of the Study:

• To introduce AnnotatorJ, an ImageJ plugin designed for semiautomatic annotation of cells in 2D microscopy images.
• To accelerate the manual annotation process for creating high-quality cell datasets.
• To facilitate the development of more accurate DL models for cell segmentation and analysis.

Main Methods:

• Development of AnnotatorJ as an ImageJ plugin integrating manual annotation with DL.
• Implementation of U-Net-based presegmentation to assist in identifying cell contours.
• Semiautomatic workflow to reduce manual effort in cell annotation.

Main Results:

• AnnotatorJ significantly accelerates the manual annotation of cells compared to traditional methods.
• The plugin aids in achieving precise cell segmentation by refining contours.
• Enables the creation of larger, high-quality annotated datasets for training machine learning models.

Conclusions:

• AnnotatorJ offers an efficient solution for semiautomatic cell annotation in bioimage analysis.
• The tool empowers researchers to generate valuable training data, potentially enhancing the performance of DL-based segmentation.
• Facilitates improved accuracy and reduced bias in cellular analyses through better annotated datasets.