Desktop version

Home arrow Health

Region-Growing-Based Method

Simple Thresholding This method extracts the tracheobronchial tree by tracing dark regions on chest CT images. The seed point is set inside the trachea and the growing process commences. When growing is performed, it is necessary to define some conditions to determine whether the growing area is inside a lumen or not. A simple threshold value is all that is necessary in this region. If a voxel has CT values lower than a given threshold value t, the growing process is performed. If not, the voxel will not be included in the queue. One issue with this method is a selection of a threshold value. If t is too low, insufficient regions will be extracted, and not all branches will be segmented. In contrast, higher t will cause extraction of aerated lung outside the tracheobronchial tree. In such case, extraction explosion occurs [195]. When drawing a graph of the number of extracted voxels, the number rapidly increases at a certain threshold. It is possible to find such a threshold value t by monitoring the number of extracted voxels. Figure 3.49 shows examples of segmentation results obtained by an optimal threshold value and a threshold value causing explosion. Because CT values of the bronchus wall become low because of the partial volume effect, explosion easily happens. A bronchus wall enhancement filter is sometimes applied. In this case, small leakage of regions growing, which is much less than explosion, will happen.

Adaptive Thresholding The method shown in the previous section uses one threshold value in the extraction of the entire tracheobronchial tree. However, as stated before, the threshold values should be adjusted branch by branch because of the partial volume effect of the surrounding aerated lung. The method shown in [156] defines a volume of interest (VOI) for a bronchial branch. The growing process is performed inside this VOI. Because the method is designed to find the running direction of a branch, the VOI is alliteratively extended in the running direction of a branch. Luminal region extraction is simultaneously performed inside the VOI. If the method detects bifurcation, two VOIs are set for secondary branches of a target branch. The threshold value of region growing is computed for each VOI. This expansion process is performed using the breadth-first search method. Branch-

Extraction of bronchus regions with appropriate threshold and inappropriate thresholds

Fig. 3.49 Extraction of bronchus regions with appropriate threshold and inappropriate thresholds

Segmentation result of region growing with adaptive thresholding [156]

Fig. 3.50 Segmentation result of region growing with adaptive thresholding [156]

by-branch growing has another advantage in that it can detect not only bronchus regions but also branching structures. An example of extraction obtained by this method is shown in Fig. 3.50.

< Prev   CONTENTS   Source   Next >

Related topics