Context modeling for the clinical predictors of obstructive sleep apnea

Context is fundamental in medical data acquisition, data interpretation, and medical reasoning. Thus, explicit contextual information must be an integral part of primary data analysis and, also, data-mining process. This paper contends that there is a need for an explicit context modeling in medical data mining, and it presents a conceptual framework for the representation of seven contextual dimensions: pragmatic, regulatory, relational, temporal, spatial, source reliability, and absent-value semantics. The presented framework is based on two underlying approaches: (1) a semiotic approach to represent multiple interpretations of concepts and data within specific contexts and (2) a fuzzy-logic approach to represent vagueness of concepts and imprecision of data, which are characteristic in particular contextual dimensions. The authors demonstrate context modeling for data acquisition in screening, diagnosis, and for research of obstructive sleep apnea (OSA). In particular, this paper describes contextual dimensions for two important OSA predictors: large neck circumference and habitual snoring.