Collaborative filtering to improve navigation of large radiology knowledge resources

Area of Science:

• Radiology Informatics
• Information Science
• Human-Computer Interaction

Background:

• Web-based radiology resources offer vast information but can be challenging to navigate.
• Knowledge-discovery techniques are needed to enhance user experience and information retrieval.
• Collaborative filtering leverages user behavior to recommend relevant content.

Purpose of the Study:

• To evaluate the impact of collaborative filtering on user navigation within a large, web-based radiology knowledge resource.
• To assess if collaborative filtering improves the utilization and ease of use of online radiology information.

Main Methods:

• An item-based collaborative filtering algorithm was implemented on 1,168 radiology hypertext documents.
• The algorithm identified the six most related documents for each item based on 248,304 page views over 18 days.
• Links to related documents were integrated into the resource, and usage was analyzed over a subsequent 5-day period.

Main Results:

• The average number of documents viewed per user visit significantly increased from 1.57 to 1.74 (P < 0.0001).
• This indicates a measurable improvement in user engagement and exploration of the resource.

Conclusions:

• Collaborative filtering effectively enhances the utilization and perceived usefulness of radiology information resources.
• The technique shows significant promise for improving navigation and user experience in large, internet-based radiology knowledge bases.