An efficient recognition method of lung nodules from X-ray CT images using 3-D object models

摘要

In this paper, we propose an efficient algorithm to detect candidates of nodule shadows from X-ray CT images using 3D geometric object models of cancers and blood vessels. By using such 3D geometric object models, the anatomical knowledge about the 3D structures of nodules and blood vessels can be reflected in recognition process. Additionally, we improve the performance of the recognition method using template matching techniques. The template images are generated from the object models before the recognition process, and all of the template images can be memorized in the computers. As a result of the improvement of the performance, the calculation time of the recognition method is decreased drastically. By applying our new method to actual CT images (38 patient images), a good result has been acquired. 1. Introduction In Japan, recently, the death rate by lung cancer is increasing rapidly. To cope with this problem, mass screening for lung cancer is widely performed by simple chest X-ray films with sputum cytology. At present, however, the detection ability of the observation by the simple X-ray film is not sufficient yet. It is known that the false negative ratio of the observation is considerably high.