Automatic Registration Of Lung Nodules On 3D CT Image

Lung cancer is the most common malignancy of viscera,is known the least survivable type in diagnosed cancers. The incidence of lung cancer is increasing yearly, in city it has the highest death rate in all tumour diseases.Computer-aided detection and diagnosis(CAD) techniques could help to improve the diagnostic accuracy,objectivity and reduce the radiologist's workload.The specific steps of nodule detection are as follows:data preprocessing,lung segmentation,candidate detection,set VOI,feature extraction,candidate classification.First the data need some preprocessing to make sure voxel isotropic, then using regional growth and morphology to find lung area.To get candidate nodule,multiscale selective enhancement filter is applied in lung. According to the candidates,VOI is set.After thresholding in VOI,the edge is got using 3D spiral model.Then features are calculated.Putting feature points into 2D space,we can find a line to separate real nodules and false nodules.Removing the false nodules,we get final result.There are 618 lung CT images and 10 cases used to test.In our experiment result,the sensitivity is 80% and 35 false positive exist.The results are satisfactory. Although the system is not tested by large number of cases,the result shows good performance in nodule detection to same extent.