Structured Literature Image Finder: Parsing Text and Figures in Biomedical Literature.

Area of Science:

• Biomedical Informatics
• Computational Biology
• Digital Pathology

Background:

• Biomedical literature contains vast amounts of unstructured text and image data.
• Extracting structured information from this literature is crucial for advancing research.
• Current methods often lack integrated approaches for text and image analysis.

Purpose of the Study:

• To develop and present the SLIF system for extracting structured information from biomedical literature.
• To integrate text-mining and image processing techniques for comprehensive data extraction.
• To provide accessible tools for querying and browsing extracted biomedical data.

Main Methods:

• Utilizing text-mining to parse biological entities (proteins, cell types) from captions.
• Employing image processing to classify image types (e.g., micrographs, gels).
• Applying deep learning for classifying subcellular localization in fluorescence microscopy images.
• Developing a web interface and web service for data querying.
• Implementing latent topic models for browsing annotated text and image data.

Main Results:

• Successful extraction of biological entities and image classifications from published papers.
• Development of a functional web application and publicly available datasets.
• Demonstration of integrated text and image analysis for structured information retrieval.
• Support for both targeted querying and topic-based browsing of biomedical literature.

Conclusions:

• The SLIF project offers a novel, integrated approach to extract and access structured information from biomedical literature.
• The system enhances discoverability and analysis of complex biological data.
• SLIF provides valuable resources for the research community, facilitating new scientific insights.