Biomedical image representation approach using visualness and spatial information in a concept feature space for...

Area of Science:

• Biomedical image analysis
• Computer vision
• Information retrieval

Background:

• Biomedical image retrieval is crucial for accessing and analyzing medical data.
• Existing methods may lack efficiency and accuracy in handling complex image features.
• A robust system for searching and retrieving similar biomedical images is needed.

Purpose of the Study:

• To develop and validate an improved approach for biomedical image retrieval.
• To represent images using local concepts and an entropy-based feature space.
• To enable interactive region-of-interest (ROI) selection for enhanced search.

Main Methods:

• Images are mapped to local concepts, which are perceptually distinct visual patches.
• A weighted entropy-based concept feature space is utilized, measuring visual significance via Shannon entropy.
• Interactive ROI selection and spatial verification are employed to refine retrieval results.

Main Results:

• The proposed approach demonstrated improved biomedical image retrieval performance.
• Experiments were conducted on two distinct datasets from open access biomedical literature.
• The method effectively utilizes local concepts and spatial information for accurate retrieval.

Conclusions:

• The developed approach enhances biomedical image retrieval accuracy.
• Mapping image regions to local concepts and using entropy-based features is effective.
• The system shows promise for applications in biomedical image analysis and information access.