Development of an automated detection algorithm for mediastinal lymph nodes on chest x-ray CT images

The purpose of this study is to develop an automated method to detect mediastinal lymph nodes on chest CT scans for patients with lung cancer. First, the mediastinal region in which lymph nodes can exist was extracted. Second, detection of lymph node candidates by a CT-value thresholding technique and procedure of false-positive reduction by features were executed. Third, in order to utilize a priori knowledge of the lymph node position, the relationship between the slice location and anatomic section in human body was determined using an automatic classification of slice positions in terms of trachea and also using a genetic algorithm. Lastly, to remove the false-positive candidates, the knowledge information processing based on the anatomical information of the chest was executed. The sensitivity was 76%, and the correct detection rate was 72%. Our results show that the algorithm may be useful for the automated detection of lymph nodes.