The development of computer-assisted radiology in PAC is associated with the use of new intelligent capabilities of computer-assisted radiology. In this paper we present our work on data mining to support medical image diagnosis. We use decision tree induction in order to learn the expert knowledge, presented in the form of image descriptions in the database. A methodology is developed to perform data mining in picture archiving systems. The tool for data mining is presented. It was applied in the task of early differential diagnosis of pulmonary nodules in lung tomograms and allowed to support image analysis and classification by the expert. Results of knowledge acquisition on the basis of data mining are presented and compared with the results of the semi-automated method of knowledge acquisition by a syndrome framework tool. The developed decision tree and developed syndrome-like decision rule are effective for early diagnosis of peripheral lung cancer, so that we are able to apply this method to other medical tasks, in particular for image analysis in mammography.
Data mining, decision support, image interpretation, decision trees, syndrome classifier