System and method for assigning mammographic view and laterality to individual images in groups of digitized mammograms
A method for identifying a type of a mammographic view for a digital mammography image. The method comprises the steps of: identifying two or more candidate view types; identifying at least one feature capable of distinguishing between the two or more candidate view types; determining the feature for the digital mammography image; and corresponding the determined feature of the digital mammography image with one of the two or more candidate view types to identify the type of a mammographic view of the digital mammography image in accordance with the correspondence.
1. A method for automatically assigning a type of a mammographic view for a digital mammography image, said method comprising the steps of:
identifying two or more candidate view types;
determining at least one feature for the digital image, the at least one feature being distinguishable between the two of more candidate view types;
correlating the determined feature of the digital image with one of the two or more candidate view types by determining, for the digital mammography image, a classification coefficient corresponding to each of the two or more candidate view types; and
assigning the type of a mammographic view to the digital mammography image in accordance with the correlation,
wherein the step of determining the classification coefficient comprises the steps of:
assigning the determined feature to a plurality of predetermined nodes of a neural network;
calculating for the digital mammography image, using the neural network, the classification coefficient corresponding to each of the two or more candidate view types; and
employing the calculated classification coefficients to assign the type of a mammographic view to the digital mammography image.
2. The method of claim 1 , wherein the step of determining the classification coefficient is accomplished by a vectorial multiplication of an average of the determined feature of preselected images of a known type with the determined feature of the digital mammography image to generate a classification coefficient corresponding for each of the two or more candidate view types.
3. The method of claim 1 , further comprising a step of optimization.
4. The method of claim 3 , wherein the step of optimization comprises the steps of:
a) comparing the classification coefficients corresponding to each of the two of more candidate views types for a first digital image to select the type with the highest classification coefficient;
b) assigning said selected type to said first image; and
c) repeating steps a) and b) for each image with the proviso that a given type, once assigned, is removed from said possible types.
5. The method of claim 3 , wherein the step of optimization comprises the steps of:
a) maximizing a sum of said classification coefficients of each type of all images; and
b) assigning to each image the type associated with that image in the maximized sum.
6. The method of claim 3 , wherein the step of optimization is accomplished by linear programing.
7. The method of claim 1 , wherein said features are calculated from a plurality of predetermined regions in said digitized image.
8. The method of claim 7 , wherein the feature is brightness, contrast, or a combination of brightness and contrast.
9. The method of claim 8 , wherein said features are calculated from at least about 20 grid elements in said digitized image.
10. The method of claim 7 , wherein the feature is brightness, contrast, or a combination of brightness and contrast.
11. The method of claim 10 , wherein the feature is normalized relative to the plurality of predetermined regions.
12. The method of claim 1 , wherein the digital mammography image is provided by digitizing X-ray film in any orientation.
13. The method of claim 1 , wherein the digital mammography image is provided by digitizing X-ray film in a pre-determined orientation.
14. The method of claim 13 , wherein the predetermined orientation is selected from an orientation of emulsion side up and emulsion side down.
15. The method of claim 14 , wherein the predetermined orientation is further selected from an orientation of portrait and landscape.
16. The method of claim 1 , wherein the step of determining the classification coefficient is accomplished by calculating a RMS of a vector difference between an average of the determined feature of preselected images of a known type with the determined feature of the digital mammography image to generate a classification coefficient corresponding for each of the two or more candidate view types.
17. The method of claim 16 , further comprising a step of optimization, wherein the step of optimization comprises the steps of:
a) minimizing a sum of said classification coefficients of each type of all images; and
b) assigning to each image the type associated with that image in the minimized sum.