Granted Patent B2

US 7,352,888 · App. 10/903,050 · Granted Apr 1, 2008

Abstract

Claims (82)

1. A method for identifying an AP/PA (anterior-posterior/posterior-anterior) projection view and an orientation of a chest radiograph, the method comprising the steps of:

providing an input digital image of the chest radiograph;

preprocessing the input digital radiographic image;

searching the preprocessed image for a spine midline and lung centerlines;

classifying the view of the input digital radiographic image as an AP/PA image based on the detection of a spine midline and lung centerlines; and

determining the orientation of the input digital radiographic image using the detected spine midline and lung centerlines, if it is classified as an AP/PA image.

2. The method of claim 1 , wherein the step of preprocessing comprises the steps of:

sub-sampling the input digital radiographic image;

segmenting the sub-sampled image into foreground, background, and anatomy regions; and

detecting two lung regions and a mediastinum region from the anatomy region.

3. The method of claim 1 , wherein the step of classifying the view of said input digital radiographic image includes the steps of:

computing a X and Y direction derivative image of the input digital radiographic image; and

reviewing the X and Y direction derivative images to identify three lines, one line being representative of a spine midline and two lines being representative of lung centerlines.

4. The method of claim 3 , further comprising the steps of:

verifying the existence of the three lines;

determining the properties of the three lines; and

determining the spatial relationships of the three lines.

5. The method of claim 3 , wherein the step of identifying the three lines is accomplished by the steps of:

assigning a first value to a maximum value of the X and Y direction derivative image to identify the spine midline; and

assigning a second value to a minimum value of the X and Y direction derivative image to identify the lung centerlines.

6. The method of claim 3 , wherein the step of determining the properties of the three lines includes the steps of:

assigning a first value to a maximum value of the X and Y direction derivative image to identify the spine midline;

assigning a second value to a minimum value of the X and Y direction derivative image to identify the lung centerlines;

identifying a shape feature of the spine midline; and

determining if there is an overlap length of the spine midline with the lung centerlines.

7. The method of claim 4 , wherein the step of determining the spatial relationships of the three lines is accomplished by the steps of:

determining whether the three lines intersect; and

determining the distance between the three lines.

8. The method of claim 1 , wherein the step of determining the orientation of the input digital radiographic image includes the steps of:

examining the orientation of a spine midline;

locating a neck and abdomen position; and

determining the image orientation responsive to the orientation of the spine midline and location of the neck and abdomen positions.

9. The method of claim 8 , wherein the step of locating the neck and abdomen position in the image is accomplished by comparing an intensity means of two regions extracted along the spine midline.

10. The method of claim 8 , wherein the step of locating the neck and abdomen position in the image is accomplished by comparing a profile extracted from a region of interest in the image to at least one profile from a predetermined oriented image.

11. The method of claim 1 , wherein:

(A) the step of classifying the view of said input digital radiographic image includes the steps of:

(i) computing a X and Y direction derivative image of the input digital radiographic image;

(ii) reviewing the X and Y direction derivative images to identify three lines, one line being representative of a spine midline and two lines being representative of lung centerlines; and

(B) the step of determining the orientation of the input digital radiographic image includes the steps of:

(i) examining the orientation of the spine midline;

(ii) locating a neck and abdomen position; and

(iii) determining the image orientation responsive to the orientation of the spine midline and location of the neck and abdomen positions.

12. The method of claim 8 , further comprising the step of, after determining the image orientation, identifying a position of a heart.

13. The method of claim 12 , wherein the step of identifying the heart position includes a training step and a testing step.

14. The method of claim 13 , wherein the training step includes the steps of:

constructing N*N patches within a region of interest of the input digital radiographic image;

computing a set of features for each patch;

associating an output specifying the heart position;

computing a transformation for the feature set; and

training a classifier with the feature set and the output.

15. The method of claim 14 , wherein the testing step includes the steps of:

computing the feature set for at least one test image;

applying the transformation used in the training step to each feature set; and

classifying the feature set using the trained classifier.

16. A computer storage product having at least one computer storage medium having instructions stored therein causing one or more computers to perform the method of claim 1 .

17. A method for identifying a projection view of a chest radiograph, the method comprising the steps of:

providing an input digital image of the chest radiograph;

preprocessing the input digital radiographic image;

searching the preprocessed image for a spine midline and lung centerlines; and

classifying the view of the input digital radiographic image as an AP/PA (anterior-posterior/posterior-anterior) image based on the detection of a spine midline and lung centerlines and as a LAT (lateral) image, otherwise.

18. The method of claim 17 , wherein the step of preprocessing comprises the steps of:

sub-sampling the input digital radiographic image;

segmenting the sub-sampled image into foreground, background, and anatomy regions; and

detecting two lung regions and a mediastinum region from the anatomy region.

19. The method of claim 17 , wherein the step of classifying the view of said input digital radiographic image includes the steps of:

computing a X and Y direction derivative image of the input digital radiographic image; and

reviewing the X and Y direction derivative images to identify three lines, one line being representative of a spine midline and two lines being representative of lung centerlines.

20. The method of claim 19 , further comprising the steps of:

verifying the existence of the three lines;

determining the properties of the three lines; and

determining the spatial relationships of the three lines.

21. The method of claim 19 , wherein the step of identifying the three lines is accomplished by the steps of:

assigning a first value to a maximum value of the X and Y direction derivative image to identify the spine midline; and

assigning a second value to a minimum value of the X and Y direction derivative image to identify the lung centerlines.

22. The method of claim 19 , wherein the step of determining the properties of the three lines includes the steps of:

assigning a first value to a maximum value of the X and Y direction derivative image to identify the spine midline;

assigning a second value to a minimum value of the X and Y direction derivative image to identify the lung centerlines;

identifying a shape feature of the spine midline; and

determining if there is an overlap length of the spine midline with the lung centerlines.

23. The method of claim 20 , wherein the step of determining the spatial relationships of the three lines is accomplished by the steps of:

determining whether the three lines intersect; and

determining the distance between the three lines.

Assignments (13)