Automatic registration of image pairs of medical image data sets

1. A computerized device for automatic registration of image data of a medical image study, acquired with a medical imaging modality, and image data of at least one reference study, each of said medical image study and said at least one reference study comprising a plurality of series with each series comprising a plurality of image data sets, and each image data set having image position coordinates and a DICOM (Digital Imaging and Communications in Medicine) header associated therewith, and each of said series being stored in at least one data memory, said device comprising:

a cluster module configured to import image data of all series of at least said medical image study into said cluster module from said at least one data memory, and to automatically cluster said image data sets according to cluster criteria based exclusively on at least one of said image position coordinates and attributes of said DICOM header, thereby producing generated clusters, and to store said generated clusters;

a master generation unit configured to determine a master for all of the generated clusters;

a pair determination unit configured to automatically register the image data of said medical image study with said image data of said at least one reference study;

a registration unit configured to generate registered pairs among said data sets of said medical image study and said reference study by implementing a similarity measurement for each master from said medical image study by searching for at least one reference study that satisfies predetermined similarity conditions, or a search for a similar reference master of the reference study; and

a monitor in communication with said registration unit, said registration unit being configured to cause said registered pairs to be displayed at said monitor.

2. A method for automatic registration of image data of a medical image study, acquired with a medical imaging modality, and image data of at least one reference study, each of said medical image study and said at least one reference study comprising a plurality of series with each series comprising a plurality of image data sets, and each image data set having image position coordinates and a DICOM (Digital Imaging and Communications in Medicine) header associated therewith, and each of said series being stored in at least one data memory, said method comprising:

in a cluster module of a processor, importing image data of all series of at least said medical image study into said cluster module from said at least one data memory, and automatically clustering said image data sets according to cluster criteria based exclusively on at least one of said image position coordinates and attributes of said DICOM header, thereby producing generated clusters, and to store said generated clusters;

in a master generation unit of said processor, determining a master cluster for all of the generated clusters;

in a pair determination unit of said processor, automatically registering the image data of said medical image study with said image data of said at least one reference study;

in a registration unit of said processor, generating registered pairs among said data sets of said medical image study and said reference study by implementing a similarity measurement for each master cluster from said medical image study by searching for at least one reference study that satisfies predetermined similarity conditions, or a search for a similar reference master cluster of the reference study; and

at a monitor in communication with said registration unit, displaying said registered pairs.

3. A method as claimed in claim 2 comprising implementing a computerized consistency check of the registered pairs by calculating an average displacement vector for at least selected ones of the registered pairs, and designating registered pairs that deviate from said average displacement vector by more than a predetermined value as inconsistent.

4. A method as claimed in claim 3 comprising automatically implementing a computerized fallback strategy for each of said registered pairs that is identified as inconsistent, by automatically selecting a next-best candidate for a master of the reference study that best satisfies said similarity conditions, and iteratively repeating said consistency check and said fallback strategy.

5. A method as claimed in claim 2 comprising clustering said reference study according to the same cluster criteria used to cluster said medical image study, and also using said similarity measurement.

6. A method as claimed in claim 2 wherein said cluster criteria are based exclusively on said position coordinates.

7. A method as claimed in claim 2 wherein said similarity measurement comprises parsing DICOM header data in said DICOM header.

8. A method as claimed in claim 7 comprising parsing said DICOM header data by use of a name attribute that identifies an anatomical body region in the image data set associated with the respective DICOM header.

9. A method as claimed in claim 2 comprising implementing at least one of said clustering and said registering of said image data sets without access to any image data in said image data sets.

10. A method as claimed in claim 2 comprising clustering said image data sets by implementing an optimization in which each of said series is analyzed as to a distance from the cluster master associated therewith and adjacent cluster masters and, if the respective cluster is a shorter distance to one of the adjacent cluster masters, shifting the image series of the respective cluster into the adjacent cluster master having said shorter distance.

11. A method as claimed in claim 2 comprising determining said master by defining a series as said master that has a largest area coverage and a highest resolution.

12. A non-transitory, computer-readable data storage medium encoded with programming instructions, said data storage medium being loaded into a computerized device having access to at least one data memory in which image data sets of a medical image data study and image data sets of at least one reference study are stored, each of said medical image study and said at least one reference study comprising a plurality of series and each of said series comprising a plurality of image data sets, and each of said image data sets having image position coordinates and a DICOM (Digital Imaging and Communications in Medicine) header associated therewith, said programming instructions causing said computerized device to:

in a cluster module, import image data of all series of at least said medical image study into said cluster module from said at least one data memory, and automatically cluster said image data sets according to cluster criteria based exclusively on at least one of said image position coordinates and attributes of said DICOM header, thereby producing generated clusters, and store said generated clusters;

in a master generation unit, determine a master for all of the generated clusters;

in a pair determination unit, automatically register the image data of said medical image study with said image data of said at least one reference study;

in a registration unit, generate registered pairs among said data sets of said medical image study and said reference study by implementing a similarity measurement for each master from said medical image study by searching for at least one reference study that satisfies predetermined similarity conditions, or a search for a similar reference master of the reference study; and

cause said registered pairs to be displayed at a monitor in communication with said registration unit.