Integration of datamining techniques into the CBR cycle to predict the result of immunotherapy treatment

摘要

The functioning of the medical diagnostic process is very comparable to the pattern of the CBR cycle. The doctor often starts to analyze the whole situation and takes advantage of previous situations resolved successfully in order to efficiently diagnose a new situation. More generally, this mode of operation, based on experience and analogy, can be found practically in all diagnostic fields (medical, industrial, etc.). In this work, we propose the integration of datamining techniques in the CBR cycle for medical decision support. This integration aims to improve the performance of the retrieval phase by reducing the number of attributes considered in the similarity calculation through the selection of the most relevant attributes. To evaluate our approach, we have customized the jCOLIBRI 2.1 framework with a real case base to predict the response to immunotherapy treatment for patients who have plantar and common warts disease.