Robust rib cage segmentation in CT image series using active contour models.

We report on initial approaches to automatically segment healthy and atelectatic lung areas in CT images for di- agnostic purposes such as quantification of aerated and atelectatic pulmonary subvolumes. We address mainly the problem of automatic segmentation of the chest wall, including the lung parenchyma in the presence of atelectasis as in this case the differentiation between atelectatic lung parenchyma and surrounding thoracic soft tissue is of- ten complicated. We propose an adequately parameterized active contour model (snake) for the segmentation. The snake evolves under the influence of the image based external forces to the border between the normal lung and the chest wall tissue and rests there. At the same time the internal forces will halt the snake on the approxi- mate border between the chest wall and the atelectatic lung thus segmenting the lung compartment including the atelectatic area.