Accuracy improvement for analysis of emphysema by lung lobe segmentation

Emphysema is a kind of obstructive lung disease, and the number of patients is increasing in recent years. The increase of the number of patients cause the increase of burden of physician, so computer aided diagnosis is strongly required and many researchers have been involved. In this study, we introduce a method of lung lobe segmentation. We incorporated this segmentation method into discriminant system of Emphysem and aimed at reducing false negative rate. To segment lobes, we emphasize the rarefied regions of the vessel by projecting an image to the sagittal plane, and applied basic techniques of image processing. Lastly, we carried out discriminant analysis based on the Mahalanobis's generalized distance. As a result, there was a reduction of 18.7% false-negative rate and improvement of 11.2% determining rate of positive compared to the conventional segmentation that uses only information of height.