DeepClas4Bio: Connecting bioimaging tools with deep learning frameworks for image classification

Area of Science:

• Bioimaging
• Deep Learning
• Computational Biology

Background:

• Deep learning models are powerful for bioimaging classification but lack integration with common tools like ImageJ.
• Life scientists face challenges accessing and utilizing deep learning results within their existing bioimaging workflows.

Purpose of the Study:

• To bridge the gap between deep learning frameworks and bioimaging tools.
• To enhance interoperability and accessibility of advanced computational models for life scientists.

Main Methods:

• Developed DeepClas4Bio, an extensible API providing a unified access point for deep learning classification models.
• Created a metagenerator for easy ImageJ plugin development and a Java application for model comparison.
• Demonstrated usage with ImageJ, Icy, and ImagePy, integrating various deep learning models and frameworks.

Main Results:

• Successfully created an API facilitating deep learning model integration into bioimaging software.
• Enabled easy creation of ImageJ plugins and facilitated comparison of different deep learning models.
• Showcased practical applications across multiple bioimaging tools and frameworks.

Conclusions:

• DeepClas4Bio empowers deep learning model developers to share their work through widely used life science tools.
• Facilitates bioimaging software developers in creating plugins that leverage deep learning capabilities.
• Provides bioimaging tool users with access to powerful deep learning functionalities within their familiar environments.