Abstract: For the problem that the thoracic CT (Computed tomography) DICOM (Digital Imaging and Communication of Medicine) images need to be segmented to show different tissues clearly in the multiwindow mode, these images were segmented according to the CT value range by connectedcomponent labeling. Firstly, the thoracic images were segmented into lung, muscle, bone and skin regions with stable CT value range, based on the CT imaging theory. The noise in the segmented regions was removed by the connected-component labeling algorithm according to the area of the objects. The thoracic CT images were created in different imaging systems from different patients of various ages and genders. Experimental results show that the lung, muscle, bone and skin regions can be segmented correctly in different data sets via the proposed method.

Key words: thoracic CT; multi-window; area; segmentation; noise removal