Granted Patent B2

US 7,221,787 · App. 10/315,884 · Granted May 22, 2007

Abstract

Claims (40)

1. A method for automatically segmenting lung regions in a chest radiographic image comprising:

providing an input digital chest radiographic image;

preprocessing said input digital radiographic image;

extracting a chest body midline and lung centerlines from said preprocessed image;

locating one-by-one, the chest body model, the spine model and the two lung models in said image based on said extracted chest body midline and two lung centerlines; and

detecting the lung regions by deforming the lung shape models to converge to the true lung boundaries using both region and edge information.

2. The method of claim 1 wherein said pre-processing includes:

estimating the two lung regions and a mediastinum region; and

normalizing said radiographic image with said estimated two lung regions and said mediastinum region.

3. The method of claim 2 wherein the background, lung and mediastinum regions are detected by threshold analysis.

4. The method of claim 2 wherein in said normalizing said radiographic image, there is used the minimal gray-level of said selected lung regions and the maximal gray-level of said detected mediastinum region.

5. The method of claim 1 , wherein the step of preprocessing includes segmenting the image based on first and second thresholds by:

thresholding said image;

labeling all regions with gray-level below the first threshold;

checking the labeled regions to ensure if they belong to background; and

removing the background regions.

6. The method of claim 1 wherein said chest body model is located by aligning its centerline with the chest body midline and said chest body model size is derived from the distance between said two lung centerlines.

7. The method of claim 6 wherein said spine model is placed in the middle of said chest body model according to their anatomic spatial relationship.

8. The method of claim 1 wherein said deforming the lung shape models includes:

selecting proper target points for each landmark based on the initial location of the lung means shape model, and the related region and edge information;

adjusting the pose parameters to best fit said target points;

updating said current shape;

repeating said selecting, adjusting, deforming and updating until the change of shape becomes negligible and the shape contours of the lung regions are established.

9. The method of claim 8 wherein said selecting proper target points includes:

constructing a small local search region for each landmark point;

checking the region properties of said local search region;

setting a target point if said local search region is inside or outside of lung regions, otherwise;

determining a target point based on edge information.

10. The method of claim 9 wherein determining a target point based on edge information includes:

computing four derivative images from said image;

checking the position of a landmark;

searching the closest edge point in the related derivative image.

11. A method for automatically segmenting lung regions in a chest radiographic image, comprising:

providing an input digital chest radiographic image;

preprocessing the input digital radiographic image;

extracting a chest body midline and lung centerlines from the preprocessed image;

locating one-by-one, the chest body model, the spine model and the two lung models in the image based on the extracted chest body midline and two lung centerlines; and

detecting the lung regions by deforming the lung shape models to converge to the true lung boundaries using both region and edge information,

wherein said extracting makes use of the 0th order-X-direction derivative image of said radiographic image and said estimated lung regions and estimated mediastinum region to detect said chest body midline and said two lung centerlines.

12. The method of claim 11 wherein said chest body midline is detected by searching the maximal pixels in the 0 th -order X-direction derivative image in the estimated mediastinum region, and said lung centerlines are detected by searching the minimal pixels in the estimated lung regions.

Assignments (15)