KnowledgeMap

KnowledgeMap was developed to enhance the delivery of medical education content to students and faculty. The volume of documentation produced during medical school makes it difficult to find information and to understand where similar topics are taught.

KnowledgeMap is a full-featured content management system. Its unique attribute is a full-text mapping of uploaded documents (slide shows, PDFs, HTML, or word processing documents) to concepts from the National Library of Medicine's Unified Medical Language System. KnowledgeMap's concept identifier uses approximate natural language processing techniques and rigorous score-based methods. Tools developed for KnowledgeMap allow quick searches of the curriculum, generation of relevant PubMed queries, and rapid assessment of where concepts are taught in the medical school. KnowledgeMap is the official vehicle for educational content delivery for the medical school. KnowledgeMap has been exported for use at other major academic centers.

KnowledgeMap concept indexer has also been applied to a number of other medical corpora, including clinical documentation, electrocardiogram reports, clinical websites, and evidence-based medicine search tools.

Links:

• “Introduction to KnowledgeMap” Video Screencast
• "Understanding" medical school curriculum content using KnowledgeMap
• Automated capture and assessment of medical school clinical experience
• "Where Do We Teach What?". Finding broad concepts in the medical school curriculum
• Automatic capture of student notes to augment mentor feedback and student performance on patient write-ups
• Tracking medical students' clinical experiences using natural language processing
• Publications
• KnowledgeMap Website
• Portfolio website

KnowledgeMap Learning Portfolio collects trainee-created documents reflecting their direct patient experiences via the Electronic Medical Records (EMR), providing a rich log of a trainee's clinical exposure without requiring additional manual effort. The interface between the EMR system and Portfolio provides historical data from which trainees can continue to build their knowledge base, tagged to core learning objectives, and supplies the mentor the opportunity to review and evaluate trainee performance.

Here is a pdf with screen shots and flow charts explaining the system. km_lp_screenshots.pdf

Links for Vanderbilt University Medical Center students and faculty:

Portfolio (VSTAR Patient Encounter Notes) website

The KnowledgeMap Concept Indexer (KMCI) is the underlying natural language processing engine used in the KnowledgeMap and Learning Portfolio website, and has been used for many clinical and genomic research studies.  It identifies biomedical concepts, mapped to Unified Medical Language System concepts, from natural language documents and clinical notes.

KMCI employs part-of-speech information to develop a shallow sentence parse, and performs variant generation and normalization using the SPECIALIST Lexicon and related tools. The KMCI system was designed particularly for poorly-formatted documents containing ad hoc abbreviations and underspecified concepts (e.g., the document phrase “ST” implying the “ST segment” of an electrocardiogram instead of abnormal finding “ST elevation”).  Using probabilistic information and concept co-occurrence data derived from PubMed, KMCI can map ambiguous strings such as “CHF” to the UMLS concept C0018802 “Congestive heart failure” in an echocardiogram report but to the concept C0009714 “Congenital hepatic fibrosis” in a document discussing infantile polycystic kidney disease (a genetically related condition to congenital hepatic fibrosis).

KMCI has performed favorably in comparison to MetaMap and has been validated in a variety of clinical and education contexts (see publications).  Later additions to KMCI include the ability to detect negated terms (e.g., "no chest pain) via a Perl implementation of NegEx