The Research On Image Segment Technology In Computer Aided Diagnosis System

With the development of medical image devices, computed tomography(CT), magnetic resonance imaging(MRI) and other imaging modalities have been widely used in clinical diagnosis and therapy. However, owing to the limitation of medical image devices, what the doctors could observe are only 2D images, which brings obstacles to the diagnosis.According to these circumstances, in the medical computer-aided diagnosis system, we implemented 2D manipulations on DICOM, medical image retrieval, 3D reconstruction on tissues. The medical image retrieval based on emphysema will help in the emphysema diagnosis, it returns the most similar thorax CR(Computed Radiography) image with the image user submits. The reconstruction of skull, skin and intracranial structure will help in surgery and computation of brain cubage.Both the medical image retrieval and reconstruction of tissues are based on segmentation of ROI (Region of Interest).The segmentation of lung is based on horizontal and vertical signature of thorax image, which lays foundation for the image retrieval. The segmentation of skull is implemented through ICM (Iterated Conditional Modes) and max entropy algorithm, it succeeds in decreasing the complexity of choosing thresholds. In order to segment intracranial structures, an algorithm based on the prior knowledge of anatomy and morphology is used, this algorithm implements the automatic segmentation.Function and performance test showed that, the medical image retrieval based on emphysema provides efficient support in computer aided diagnosis; the skull segment algorithm decreases the complexity from O( n m) to O(mn), segment result can make the skull reconstruction more clear; the intracranial structures segment algorithm based on prior knowledge and morphology eliminates the impact of cubage effect, the result can be used in the intracranial structures reconstruction.