Detailed extraction method for the lung region using the active cylinder model in chest X-ray CT images

We are developing an automatic system for recognizing lung cancer shadows in chest X-ray CT images to support lung cancer mass screening. The development of an extraction method for the lung region in chest X-ray CT images is an important preprocessing method for the above computer-aided diagnosis system. We have already proposed a model-based rough extraction method for the lung region using a typical patient sample as a model. This method involves the affine transformation of the model to fit it to the input images, and the transformed model image is defined as the rough lung region for the input image. In this paper, we propose a detailed extraction method using the Active Cylinder Model. The results of the rough extraction method are used as initial control points in the Active Cylinder Model. The control points are moved based on the internal energy and the image energy. The internal energy is position information of the adjoining control points, and the image energy is image information such as edge information calculated from the input image. This method was applied to 68 sample chest CT images, and was found to be very effective and practical. (author)