Constructing case bases from medical databases
A case-based reasoner that makes use of a medical domain knowledge base can be used to assist physicians in decision making. This thesis presents research on preprocessing an Intensive Care Unit (ICU) medical database to arrive at an appropriate database and case base architecture. Appropriate case bases were identified using an ICU patient model. Appropriate weights for case-based matching were after making thorough statistical analysis of the underlying data. This new approach to case-based reasoning with medical data was integrated with an existing graphical user interface IDEAS for ICUs version 2.3. A case-based reasoning shell called The Easy Reasoner was used for case-based reasoning. IDEAS for ICUs version 3.0, a user friendly clinical decision support tool, was developed using Microsoft Visual Basic 4.0. Testing was carried out using an ICU database of over 3000 patients. Testing results showed a significant improvement in the running speed compared with a previous version of case-based reasoning with the same data.