Case-Based Reasoning in the Health Sciences: Foundations and Research Directions - Computational Intelligence in Healthcare: Advanced Methodologies - page 134

Table 1. Sample purpose classifications

Code Purpose Code Purpose

10 Medical Purpose 20.2 Evaluation

10.1 Decision Support 20.2.1 System Level Testing

10.1.1 Diagnosis 20.2.2 Pilot Testing

10.1.2 Treatment/therapy 20.2.3 Clinical Trial

10.1.3 Prognosis 20.2.4 Routine Clinical Use

10.1.4 Follow-up 20.3 Concept

10.1.5 Classification 20.4 Method

10.2 Tutoring 20.5 Survey

10.3 Epidemiology 30 Bioinformatics Purpose

10.4 Research support 30.1 Proteomics

10.5 Image interpretation 30.2 Phylogenetics

20 Research Purpose 30.3 Genomics

20.1 Formalization 40 Research Theme

CBR systems generally support either medical clinical work, research, or bioin-
formatics. Therefore we have added these as top level purpose categories. In the
clinic, decision support systems support mostly diagnosis, treatment, prognosis,
follow-up, and/or classification, such as in image interpretation. Well known di-
agnostic systems include CARE-PARTNER [34]. Well known classification sys-
tems include PROTOS [22] and IMAGECREEK [40]. Well known treatment
planning systems include T-IDDM [65] and Auguste [35]. Several systems provide
multi-expertise, such as CARE-PARTNER [34], ensuring diagnosis and treat-
ment planning. Well known tutoring systems include ICONS [66].
More recent articles require differentiation between the purpose of the system
developed, which is generally a clinical purpose, and the purpose of the research
paper, which can be, among others, a survey paper or a classification paper.
Some papers focus on formalization such as KASIMIR [67]. Among these, the
evaluation of a system can be performed more or less thoroughly. This is an
important dimension to note about a research paper: whether the system was
tested only at the system level, which is the most frequent, at the pilot test-
ing level, at the clinical trial level, or finally whether the system is in routine
clinical use.
Finally, a paper is generally identified by a research theme by its authors.
By indexing a set of 117 papers currently in our database, we have identified
major research themes, such as CBR and electronic medical records (EMR),
knowledge morphing, CBR and clinical guidelines, or application of a generic
CBR framework.

Memory and Case Management. This is a very broad category and could
easily be subdivided. It encompasses both how the cases are represented and also