You are here

Explaining Anomalies in the Intensive Care Unit

CISA Seminar Series
Laura Moss, Glasgow Royal Infirmary and University of Aberdeen

Anomalies identify a mismatch between a prediction (often grounded in domain theories) and observations, indicating that the theory requires refinement. Early approaches to the revision of knowledge captured in a knowledge-based system generated refinements to remove an anomaly; these refinements were generally produced after applying machine learning techniques to extensive datasets (such datasets are not always available), moreover, the refinements generated are not always acceptable to domain experts. An alternative approach uses existing domain knowledge to generate domain acceptable explanations for the anomaly which can subsequently be used to refine a knowledge base. This talk will provide an overview of an investigation into the identification and subsequent explanation of anomalous reactions to treatment by ICU patients. In the first stage of the empirical study, extensive interviews were held with domain experts; the analysis of which led to the identification of both examples of anomalies encountered in the domain, and the strategies used by the domain experts to provide (appropriate) explanations for the anomalies. In the second stage, a knowledge-based system, EIRA (Explaining, Inferencing, and Reasoning about Anomalies), was developed. EIRA is able to replicate these explanations when presented with an anomaly, potential explanations are generated by the application of expert-acquired strategies to the domain knowledge, patient data, and background clinical information. The explanations produced by EIRA have been (favorably) evaluated by ICU consultants. We believe that the EIRA system is somewhat generic as are some of the strategies reported in this study. We plan, in further work, to apply EIRA to further domains to investigate these claims.

Date and time: 
Wednesday, 17 March, 2010 - 14:00
60 minutes