The Siena College Medical Information Retrieval System (MIRS)
Abstract
The work done by our MIRS team of three students and two faculty mentors resulted in a baseline system for content-based medical record retrieval. We also made significant progress on an alternative system based on neural computing concepts. The task for the Medical Records TREC in 2012 was to process a list of thirty-four randomly selected queries against a large medical records database to simulate searches for patients who meet the criteria for participating in various clinical trials. The task was to analyze a data set of over 100,000 reports associated with hospital visits and identify patients whose situations were relevant to the queries. Our text retrieval process was done in two separate ways: one used an index created from standard Information Retrieval (IR) software called Lucene and an alternate method based on principles of neural computing. We submitted three runs to the TREC competition, two using our standard Lucene-based approach and one that used elements of neural network analysis.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 2012
- Accession Number
- ADA581302
Entities
People
- Larry Medsker
- Sharon Small
Organizations
- Siena College