Brain Medical Image Mining Technology

Medical image based computer aided diagnosis is an important and challenging task, it has extracted more and more research work in recent years. Due to its interdisciplinarity and complexity, there remain many problems not solved. In this paper, a novel diagnosis method named SeCED-FS is proposed, which utilized as the core mechanism of our medical image based computer aided encephalopathy diagnosis system. The method combines the clusterer ensemble and feature selection technique to improve the diagnosis performance. At first, selective clusterer ensemble with feature selection technique is utilized to perform the classification of medical images in the two-level architecture, then, the Regions of Interest in positively identified image are outlined by using an ensemble of Fuzzy C-Means algorithm. The case studies on real data experiments show that, the SeCED-FS hold the improved generalization ability and achieved a satisfactory result not only in the accuracy of classification but also in correctly labeling the significant regions.It is noticeable that, although k-Means and fuzzy C-Means are employ as the base clusterer, the k-Medoids is utilized to select a subset of clusterer for ensemble and the prediction risk criteria is adopted for the subset feature selection. In fact, they are all replaceable and extendable, the core character of SeCED-FS is irrelevant to them and applicable to diverse kinds of clustering algorithms and feature selection technique. For the same reason, the measurement employed to distinguish individual clusterers also replaceable, instead of inter-clusterer similarity, perhaps SeCED-FS could achieve an even better result. Furthermore, medical image classification and discovery of ROI is just the first step to survey in this field, there are remain lots of works on discover patterns and associations need to be investigated.