Bean, C.A.: Representation of medical knowledge for automated semantic interpretation of clinical reports (2004)
0.02
0.018961856 = product of:
0.13273299 = sum of:
0.13273299 = weight(_text_:interpretation in 2660) [ClassicSimilarity], result of:
0.13273299 = score(doc=2660,freq=12.0), product of:
0.21405315 = queryWeight, product of:
5.7281795 = idf(docFreq=390, maxDocs=44218)
0.037368443 = queryNorm
0.6200936 = fieldWeight in 2660, product of:
3.4641016 = tf(freq=12.0), with freq of:
12.0 = termFreq=12.0
5.7281795 = idf(docFreq=390, maxDocs=44218)
0.03125 = fieldNorm(doc=2660)
0.14285715 = coord(1/7)
- Abstract
- A set of cardiac catheterisation case reports was analysed to identify and encode for automated interpretation of the semantic indicators of location and severity of disease in coronary arteries. Presence of disease was indicated by the use of specific or general disease terms, typically with a modifier, while absence of disease was indicated by negation of similar phrases. Disease modifiers indicating severity could be qualitative or quantitative, and a 7-point severity scale was devised to normalise these modifiers based an relative clinical significance. Location of disease was indicated in three basic ways: By situation in arbitrary topographic divisions, by situation relative to a named structure, or by using named structures as boundary delimiters to describe disease extent. In addition, semantic indicators were identified for such topological relationships as proximity, contiguity, overlap, and enclosure. Spatial reasoning was often necessary to understand the specific localisation of disease, demonstrating the need for a general Spatial extension to the underlying knowledge base.
- Content
- 1. Introduction In automated semantic interpretation, the expressions in natural language text are mapped to a knowledge model, thus providing a means of normalising the relevant concepts and relationships encountered. However, the ultimate goal of comprehensive and consistent semantic interpretation of unrestrained text, even within a single domain such as medicine, is still beyond the current state of the art of natural language processing. In order to scale back the complexity of the task of automated semantic interpretation, we have restricted our domain of interest to coronary artery anatomy and our text to cardiac catheterisation reports. Using a multi-phased approach, a staged series of projects is enhancing the development of a semantic interpretation system for free clinical text in the specific subdomain of coronary arteriography.
