A coronary calcification diagnosis system based on helical CT images

Abstract

Describes a computer assisted diagnostic algorithm for coronary calcifications based on helical X-ray CT images which is used in mass screening process for lung cancer diagnosis. The authors' diagnostic algorithm consists of four processes: First, they choose the heart slices from the CT images which were taken at the mass screening. They classify the heart slices into three sections which have different coronary geometries, using the information of the heart shape, trachea, CT values in the heart region, and the bone. Second, the authors extract the heart region in each slice, using the information of the lung shape and the body of vertebra. Third, they detect the candidate regions of the coronary calcifications using an edge filter and thresholding pixel values. Finally, to increase the effectiveness of the diagnosis, the authors exclude the artifact regions included in the candidate regions by using the diagnostic rule based on a neural network. They applied this algorithm to helical CT images of 462 patients screened for lung cancer. The results generated by this system were compared with a physician's diagnosis. This system could detect 213 of 214 regions which were diagnosed as coronary calcifications or probably coronary calcifications by a physician. There was only one false negative case. The false positive ratio was 0.30 per patient.