SAMJ: fast image annotation on ImageJ/Fiji via segment anything model

Area of Science:

• Biomedical image analysis
• Artificial intelligence in medicine
• Bioimage informatics

Background:

• Accurate AI training for biomedical image analysis requires time-consuming pixel-level annotations.
• Existing foundation models like Segment Anything Model (SAM) have limited integration into bioimage analysis platforms.
• Technical expertise and specialized hardware are often needed for advanced AI annotation.

Purpose of the Study:

• To develop a user-friendly plugin for ImageJ/Fiji that simplifies AI-assisted image annotation.
• To enable fast, interactive, and accurate annotation of biomedical images without programming skills.
• To broaden access to advanced AI annotation tools for the research community.

Main Methods:

• Integration of efficient Segment Anything Model (SAM) variants into ImageJ/Fiji via the SAMJ plugin.
• Development of a graphical user interface for interactive object segmentation using clicks or bounding boxes.
• Implementation on standard computers without requiring specialized hardware.

Main Results:

• SAMJ enables real-time delineation of objects in large scientific images.
• The plugin significantly reduces the effort and time required for image annotation.
• SAMJ facilitates faster creation of datasets for AI model training.

Conclusions:

• SAMJ provides a fast, interactive, and accurate solution for biomedical image annotation.
• The plugin lowers the barrier to entry for using AI-assisted annotation tools in research.
• SAMJ accelerates AI development in biomedical image analysis by simplifying dataset creation.