An automatic system for extracting figure-caption pair from medical documents: a six-fold approach

Area of Science:

• Medical Informatics
• Image Analysis
• Document Processing

Background:

• Medical figures and captions are vital information sources in documentation.
• Researchers increasingly seek to extract these elements from published medical papers for knowledge acquisition.

Purpose of the Study:

• To introduce a novel six-stage methodology for the automated extraction of figure-caption pairs from medical documents.

Main Methods:

• Utilized A-torus wavelet transform for edge detection.
• Employed maximally stable extremal regions and multi-layer perceptron for content isolation and retrieval.
• Applied bounding box approach for figure-caption pair extraction.

Main Results:

• The proposed method successfully extracts figure-caption pairs.
• Evaluated on a custom database from five open-access medical books.
• Demonstrated significant efficiency gains and robustness compared to previous systems.

Conclusions:

• The developed methodology offers a robust and efficient solution for extracting valuable figure-caption information from medical literature.
• Facilitates enhanced knowledge discovery and data utilization for researchers.