Systems and methods for automatic robust anatomy detection through local voting and prediction

1. A method for performing a medical imaging study by a computer system comprising:

acquiring from a scanner, a preliminary scan of a patient;

automatically detecting, by processor, a set of local feature candidates from the preliminary scan;

assessing, by the processor, the accuracy of each local feature candidate using multiple combinations of the other local feature candidates and removing a local feature candidate that is assessed to have the lowest accuracy;

repeating, by the processor, the assessing and removing step until only a predetermined number of local feature candidates remain;

locating, by the processor, and outputting by a display a region of interest (ROI) from within the preliminary scan based on the remaining predetermined number of local feature candidates; and

performing a medical imaging study based on the location of the ROI within the preliminary scan.

2. The method of claim 1 , wherein the preliminary scan is a scout image that includes a two-dimensional representation of a subject being scanned.

3. The method of claim 2 , wherein the preliminary scan is a topogram image.

4. The method of claim 1 , wherein the medical imaging study is a CT scan and the preliminary scan is a CT topogram image.

5. The method of claim 1 , wherein the local feature candidates represent potential anatomical landmarks.

6. The method of claim 1 , wherein the local feature candidates are automatically detected from the preliminary scan by identifying regions of the preliminary scan that appear to be known anatomical landmarks.

7. The method of claim 6 , wherein the set of local feature candidates includes multiple local feature candidates that appear to be the same anatomical landmark.

8. The method of claim 1 , wherein the accuracy of each local feature candidate is assessed by using each combination of other local feature candidates as a voting group, wherein each voting group votes for the each local feature candidate by judging the degree to which the each local feature candidate represents a corresponding local feature wile assuming that the voting group accurately represents corresponding local features.

9. The method of claim 8 , wherein each voting group includes 1, 2 or 3 other local feature candidates.

10. The method of claim 8 , wherein each voting group includes 4 or more other local feature candidates.

11. The method of claim 1 , wherein locating a region of interest (ROI) from within the preliminary scan based on the remaining predetermined number of local feature candidates includes using the remaining predetermined number of local feature candidates as frame of reference to structurally register the preliminary scan and then finding the region of interest (ROI) within the preliminary scan based on the structural registration.

12. The method of claim 1 , wherein each of the multiple combinations of the other local feature candidates comprises a voting group that votes for each local feature candidate in assessing their accuracy and for each iteration of repeating the assessing and removing step, a local feature candidate is assessed to have the lowest accuracy when it has a lowest vote from among maximum votes received by each of the multiple combinations of the other local feature candidates.

13. The method of claim 1 , wherein for each iteration of repeating the assessing and removing step, a local feature candidate is assessed to have the lowest accuracy when it is has a sudden reduction in vote value, as determined by examining the mean of good votes from a most recent iteration.

14. A method for determining a scanning field for performing a medical imaging study comprising:

receiving, from a medical imaging device, a topogram image of a patient;

automatically detecting, by a processor, a set of local feature candidates from the topogram image;

determining, by the processor, which of the local feature candidates represent a worst candidate by having a plurality of groups of the local feature candidates vote on each individual local feature candidate, and removing the worst candidate from the set of local feature candidates;

repeating, by the processor, the voting and removal such that one feature candidate is removed from the set at each iteration, until there are only a predetermined number of remaining feature candidates; and

selecting, by the processor, and outputting by a display the scanning field based on the remaining feature candidates.

15. The method of claim 14 , wherein the local feature candidates represent potential anatomical landmarks.

16. The method of claim 14 , wherein voting is performed by using each combination of other local feature candidates as a voting group, wherein each voting group votes for the each local feature candidate by judging the degree to which the each local feature candidate represents a corresponding local feature wile assuming that the voting group accurately represents corresponding local features.

17. The method of claim 14 , wherein selecting the scanning field based on the remaining feature candidates includes finding a region of interest (ROI) within the topogram image based on the remaining feature candidates and selecting the scanning field to include the region of interest (ROI).

18. The method of claim 17 , wherein the region of interest is found within the topogram image by using the remaining feature candidates as frame of reference to structurally register the topogram image and then finding the region of interest (ROI) within the topogram image based on the structural registration.

19. The method of claim 17 , wherein the region of interest (ROI), to be found is manually selected by a user.

20. A computer system comprising:

a processor; and

a program storage device readable by the computer system, embodying a program of instructions executable by the processor to perform method steps for performing a medical imaging study, the method comprising:

acquiring a topogram image;

automatically detecting, by the processor, a set of local feature candidates representing potential anatomical landmarks from the topogram image;

assessing, by the processor, the accuracy of each local feature candidate using multiple combinations of the other local feature candidates and removing a local feature candidate that is assessed to have the lowest accuracy;

repeating, by the processor, the assessing and removing step until only a predetermined number of local feature candidates remain;

locating, by the processor and outputting by a display a provided region of interest (ROI) from within the topogram image based on the remaining predetermined number of local feature candidates; and

performing a medical imaging study based on the location of the ROI within the topogram image.