Cased-Based Reasoning for medical knowledge-based systems

https://doi.org/10.1016/S1386-5056(01)00221-0Get rights and content

Abstract

In this paper we present the results of the MIE/GMDS-2000 Workshop ‘Case-Based Reasoning for Medical Knowledge-based Systems’. While in many domains Cased-Based Reasoning (CBR) has become a successful technique for knowledge-based systems, in the medical field attempts to apply the complete CBR cycle are rather exceptional. Some systems have recently been developed, which on the one hand use only parts of the CBR method, mainly the retrieval, and on the other hand enrich the method by a generalisation step to fill the knowledge gap between the specificity of single cases and general rules. And some systems rely on integrating CBR and other problem solving methodologies. In this paper we discuss the appropriateness of CBR for medical knowledge-based systems, point out problems, limitations and possible ways to cope with them.

Introduction

Cased-Based Reasoning (CBR) has become a successful technique for knowledge-based systems in many domains, while in the medical field some more problems still arise. In this paper, we are going to discuss the appropriateness of CBR for medical knowledge-based systems and to point out its problems, limitations and possible ways how they can partly be overcome.

Case-Based Reasoning means to retrieve former, already solved problems similar to the current one and to attempt to modify their solutions to fit for the current problem (Fig. 1 shows the Cased-Based Reasoning cycle developed by Aamodt [1]). The underlying idea is the assumption that similar problems have similar solutions. Though this assumption is not always true, it holds for many practical domains.

CBR fulfils two main tasks [1], [2]: the first is the retrieval, which means to search for or to calculate the most similar cases. If the case base is rather small, a sequential calculation is possible, otherwise faster non-sequential indexing [2], [3] or classification algorithms (e.g. ID3 [4] or Nearest Neighbour match [5]) should be applied. For this task much research has been undertaken in the recent years and for nearly every sort of application problem it has actually become correspondingly easy to find a suitable, sophisticated CBR retrieval algorithm. The second task, the adaptation (reuse and revision), means a modification of solutions of former similar cases to fit for a current one. If there are no important differences between a current and a similar case, a simple solution transfer is sufficient. Sometimes only few substitutions are required, but in other situations the adaptation is a very complicated process. So far, no general adaptation methods or algorithms have been developed. The adaptation is still absolutely domain dependent.

Especially in medicine, the knowledge of experts does not only consist of rules, but of a mixture of textbook knowledge and experience. The latter consists of cases, typical and exceptional ones, and the reasoning of physicians takes them into account [6]. Medical knowledge based systems therefore contain two knowledge types: objective knowledge, which can be found in textbooks, and subjective knowledge, which is limited in space and time and changes frequently.

Both sorts of knowledge can clearly be separated: objective textbook knowledge can be represented in forms of rules or functions, while subjective knowledge is contained in cases. The problem of updating the changeable subjective knowledge can partly be solved by incrementally incorporating new up-to-date cases [6].

So, the arguments for the use of case-oriented methods can be summarised as follows:

  • 1

    Reasoning with cases corresponds with the typical decision making process of physicians.

  • 2

    Incorporating new cases means automatically updating parts of the changeable knowledge.

  • 3

    Objective and subjective knowledge can be clearly separated (of course they can be used together in one system).

  • 4

    As cases are routinely stored, integration into a Hospital Information System (HIS) is easy.

Section snippets

Medical Cased-Based Reasoning systems

In medicine, CBR has mainly been applied for diagnostic and partly for therapeutic tasks. Related methods have been used in further fields: case-oriented methods for tutoring (e.g. D3 [7]) and retrieval methods to search for similar images (e.g. MACRAD [8]). In this paragraph just three medical case-based decision support systems are mentioned. For further systems we refer to Ref. [9].

One of the earliest medical decision support systems that applies CBR is CASEY [10]. It deals with heart

Problems of Cased-Based Reasoning for medical applications

To use Cased-Based Reasoning a few problems have to be solved: a representation form for cases has to be determined, and an appropriate retrieval algorithm has to be selected. Moreover, an infinite growth of the case base should be avoided e.g. by clustering cases into prototypes and removing redundant ones, or by restricting the case base to a fixed number of cases and updating it during expert consultation sessions [8]. However, the main problem of the CBR method is the adaptation task.

Retrieval-only: time course prognoses of the kidney function

As intensive care patients are often no longer able to maintain adequate fluid and electrolyte balances due to impaired organ functions or because they are ventilated, physicians need objective criteria for the monitoring of the kidney function and to diagnose therapeutic interventions as necessary. At our intensive care unit the renal function monitoring system NIMON [26] was developed that daily prints a renal report that consists of 13 measured and 33 calculated parameter values. However,

Conclusion

Cased-Based Reasoning seems to be a suitable technique for medical knowledge based systems. However, the adaptation task is the bottleneck that has to be solved. Though adaptation is sometimes a rather easy task (as in FLORENCE), in many medical applications it may become an insurmountable difficulty. In this paper we have presented three possible solutions, all of them are justified for specific applications and none of them is an ultimate solution. Retrieval-only systems are especially useful

References (39)

  • J. Quinlan

    C4.5, Programs for Machine Learning

    (1993)
  • L. Gierl

    Klassifikationsverfahren in Expertensystemen für die Medizin

    (1992)
  • F. Puppe

    Generated critic in the knowledge based neurology trainer

  • R. Macura et al.

    MacRad: radiology image resource with a Cased-Based retrieval system

  • L. Gierl

    CBR in medicine

  • P. Koton

    Reasoning about evidence in causal explanations

  • C. Bradburn et al.

    The application of Cased-Based reasoning to the tasks of health care planning

  • R. Turner

    Organizing and using schematic knowledge for medical diagnosis

  • R. Bergmann et al.

    Towards a new formal model of transformational adaptation in Cased-Based reasoning

  • Cited by (151)

    • Artificial Intelligence: The Future for Diabetes Care

      2020, American Journal of Medicine
      Citation Excerpt :

      Case-based reasoning (CBR). CBR, an AI technique to solve new problems based on learning from similar past encounters, is being extensively used in diabetes management.17,18 The 4 Diabetes Support System is an example of CBR that has been used in diabetes care.

    • Neural Insights for Digital Marketing Content Design

      2023, Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
    View all citing articles on Scopus
    View full text