Case-Based Reasoning in the Health Sciences: Foundations and Research Directions - Computational Intelligence in Healthcare: Advanced Methodologies - page 127

- MacRad helped to interpret radiological images, supporting both diagnosis
and education through its case base of reference images [37]
- CTS segmented CT-scans of the brain to assist in determining the degree of
degenerative brain disease [38]
- SCINA interpreted myocardial perfusion scintigrams to detect coronary
artery disease [39]
- ImageCreek interpreted abdominal CT images [40]
- PhylSyst mined phylogenetic classifications [41]
- CADI was a tutoring system designed to instruct medical students and res-
idents in cardiac auscultation [42]

5 Impact of CBR in the Health Sciences

CBR in the Health Sciences is a particularly active area of research today. As
health sciences sectors expand, advanced decision-support systems play an in-
creasingly important role in the evolution of medicine towards a more stan-
dardized and computerized science. CBR decision-support systems base their
recommendations on the most similar or most reusable experiences previously
encountered. CBR is thus a methodology of choice for descriptive experimental
sciences such as the natural and life sciences, especially biology and medicine.

5.1 Impact on CBR

CBR has found in biomedicine one if its most fruitful application areas, but also
one of its most complex. The main reason for its achievements and for the inter-
est of the biomedical community is that case-based reasoning capitalizes on the
reuse of existing cases, or experiences. These abound in biology and medicine,
where knowledge stems from the study of natural phenomena, patient problem
situations, or other living beings and their sets of problems. In particular, the
important variability in the natural and life sciences fosters the development of
case-based approaches in areas where complete, causal models are not available
to fully explain naturally occurring phenomena. Biomedicine is a domain where
expertise beyond the novice level comes from learning by solving real and/or
practice cases, which is precisely what case-based reasoning accomplishes. Pro-
totypical models often describe biomedical knowledge better than other types of
models, which also argues in favor of CBR [9].
One of the complexities of biomedicine is the high-dimensionality of cases,
as found in bioinformatics [43,44,45] and also in long-term follow-up care [34].
Multimedia case representation and the development of suitable CBR methods
for handling these represent another complexity, as in medical image interpreta-
tion [40,46,47,48], sensor data interpretation [49], and time series case features
[50]. Other factors include: the co-occurrence of multiple diseases; overlapping
diagnostic categories; the need to abstract features from time series representing
temporal history [50], sensor signals [49], or other continuous input data; and
the use of data mining techniques in addition to case-based reasoning [47,51].