Abstract

In a cooperation project with the Palliative Care Unit at St. Olavs Hospital, we are studying the use of case-based reasoning (CBR) and Bayesian networks (BN) to support clinical personnel in the classification and treatment of cancer pain. Our activities are coupled to ongoing EU and NFR projects, in particular the NFR funded project TLCPC. In this Master Project the problem of explaining the system's suggestions, in order to increase the clinician's understanding and trust, will be studied. The underlying problem solving method will combine case-based and Bayesian reasoning, and the explanation method need to reflect that. A method shall be designed that identify one or two explanation goals, based on a pre-study made in the earlier specialization project, for example transparency and justification of the system's behaviour. An experimental implementation that demonstrate the explanation method will be made, based on the existing CBR and BN tools within the TLCPC research group, i.e. jColibri and Smile/Genie, respectively. Application of the method to case data and Bayesian models for palliative pain requires that these data and models become available in due time, as is the plan. In case of delays, and intermediate or different application will be selected, which is similar enough so that essential properties of clinical decision making is kept.

Citations