Semi-automated rib cage segmentation in CT images for mesothelioma detection

The segmentation of the rib cage in CT images represents a task of primary importance in medical imaging for different reasons. From the study of the segmented bone several features can be extrapolated. These features are indices of the presence of some diseases such as the Malignant Pleural Mesothelioma (MPM) which is the main focus of our research. This tumor is generally located very close to the rib cage and its structure can serve as a reference for the location of the pleural mesothelioma. An estimation of the treatment surface and the mesothelioma location must be performed through rib cage segmentation. To achieve this goal, we must extract automatically the rib cage structure from other structure in in the CT image and prevent the inclusion of some undesirable regions such as mediastinal space in the final result. This paper presents a semiautomated rib cage segmentation method based on the image features. The algorithm to segment the rib cage employs a stepwise approach and consists of five stages : First, the centroid points of the spinal canal and the sternum were identified by the user. These points allowed an adjustment of a region of interest (ROI) over the mediastinal space and an intensity enhancement to separate the mediastinal space from the thoracic cage. Second, a thresholded volume was extracted from the input volume data of the CT image by applying multiple pixel thresholding. Next, a condidate bone was extracted from the thresholded volume data using three-dimensional (3D) connected component labeling algorithm. Finally, morphological operations of dilation were applied to the candidate bone regions to compensate for artifacts caused by partial volume effects. A total of 30 patients were used in our experiments. 22 of the 30 cases were successfully segmented and 8 of the 30 cases were successfully segmented but the sternum structure has been removed.