METHODS FOR SEGMENTING IMAGES AND DETECTING SPECIFIC STRUCTURES

• US 20120070052A1
• Filed: 02/10/2010
• Published: 03/22/2012
• Est. Priority Date: 02/27/2009
• Status: Active Grant

• Current Assignee: Commissariat a L'Energie Atomique (AES Corporation)
• Original Assignee: Commissariat a L'Energie Atomique (AES Corporation)
• Inventors: Maroy, Renaud, Trebossen, Regine, Zanotti Fregonara, Paolo
• Granted Patent: US 9,092,867 B2
• Time in Patent Office: Days
• US Class Current: 382/131
• CPC Class Codes: G06T 2207/10076 4D tomography; Time-sequential 3D tomography
• CPC Class Codes: G06T 2207/10104 Positron emission tomography [PET]
• CPC Class Codes: G06T 2207/20016 Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
• CPC Class Codes: G06T 2207/30096 Tumor; Lesion
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/149 involving deformable models, e.g. active contour models

Abstract

It relates to a method for segmenting images, to a method for detecting specific structures, and to a related computer device. The method for segmenting a three-dimensional image of a subject includes dividing the image into a plurality of regions, and then hierarchically merging the regions resulting from the division in order to obtain a three-dimensional image partitioned into regions of interest. The hierarchical merging includes a merging step using shape and size criteria of the regions so as to avoid merging small regions. The method for detecting specific structures in an image, for example tumors, includes a step of segmenting an image into regions of interest. The method further includes calculating, for each region of interest, a plurality of criteria, including a shape criterion, of the regions of interest in order to discriminate the specific structures to be detected from the regions of interest of the segmented image.

Claims

1. A method for segmenting a three-dimensional image or sequence of three-dimensional images of an object so as to obtain a three-dimensional image or sequence of images partitioned into regions of interest comprising, in the performance order of the method, the following steps:

dividing the image or sequence of images into a plurality of regions;

then

hierarchically linking the regions resulting from the division so as to obtain the image or sequence of three-dimensional images partitioned into regions of interest;

wherein the hierarchical linkage step comprises a linkage step using shape and size criteria for regions so as not to link small regions.

2. The segmenting method according to claim 1, wherein the hierarchical linkage step further comprises a linkage step of the regions without shape and size criteria for the regions configured to link the regions into regions of interest, without distinguishing by shape or size.

3. The segmenting method according to claim 2, wherein the step for merging the regions without shape and size of the regions is done after the linkage step using shape and size criteria of the regions and from the image or sequence of images resulting from the performance of this linkage step.

4. The segmenting method according to claim 3, wherein the hierarchical linkage step is a double linkage also comprising a linkage step without shape and size criteria for the regions from the image or sequence of images divided into regions resulting from the performance of the division step.

5. The segmenting method according to 4claim 2, wherein the or each linkage step for the regions without shape and size criteria of the regions comprises a step for calculating a linkage cost, between a region R₁ and a region R₂, defined by:

Cost(R₁,R₂)=H(R₁,R₂)

H being a function depending on variations of the signal in regions R1 and R2;

then a linkage step for the images according to a minimum linkage cost rule.

6. The segmenting method according to claim 1, wherein the linkage step using shape and size criteria for the regions so as not to link the small regions comprises a step for calculating a linkage cost, between a region R₁ and a region R₂, defined by:

7. The segmenting method according claim 1, wherein the image or sequence of images comprises a plurality of voxels and measurements for each of the voxels of a variable of the image or sequence of images, during n time intervals (n≧

1) and the segmentation method comprises, before the step for division into regions, the following steps configured to define the number of regions separated during the division step;

calculating a local spatial variance map of said measured variable of the image or sequence of images, over time and for each voxel of the image or sequence of images;

then

automatically extracting at least one core point of the image or sequence of images, the core points being the pixels of the minima of the local spatial variance map of said variable;

the number of core points extracted during this step defining the number of regions of the image or sequence of images divided at the end of the division step.

8. The segmentation method according to claim 7, wherein it further comprises, after the step for automatic extraction of the core points and before the division step, a step for extracting a temporal evolution of said variable of the or each core point and an estimation of the global parameters of the noise, configured to calculate the linkage cost during the hierarchical linkage step.

9. The segmenting method according to claim 7, wherein each region resulting from the step for division of the image or sequence of images corresponds to a core point resulting from the step for automatic extraction of the core points of the image or sequence of images and each region groups together the pixels having a similar temporal evaluation of said variable.

10. The segmenting method according to claim 1 wherein it further comprises a step for excluding a background of the image or sequence of images comprising a step for defining a threshold for the image or sequence of images configured to extract a mask of the object from an image acquired in transmission of the object so as to apply the segmentation method only to the object.

11. The segmenting method according to claim 10, wherein the exclusion step of the background comprises a smoothing step to maximize a contrast of the image in the small regions of interest before the step for defining a threshold.

12. The segmenting method according to claim 1, wherein said image or sequence of images is acquired beforehand by an imaging device according to the positron emission tomography technique.

13. The segmenting method according to claim 1, wherein the image or sequence of images of an object is an image or sequence of images of an entire body, said segmentation method segmenting the body according to a partition into pharmaco-organs.

14. The segmenting method according to claim 13, wherein the body is animated by physiological movements either of the periodic type, the period of which is reduced in comparison with the acquisition duration of the image or each of the images of the sequence, or the non-periodic type.

15. The segmenting method according to claim 13, wherein said variable represents the radioactive concentration at a given moment of at least one active ingredient marked and injected into the body, the set of voxels inside each pharmaco-organ having pharmacological kinetics for distribution of said active ingredient that are similar.

16. A method for detecting at least one particular structure over an image or sequence of images, wherein it comprises the following steps:

segmenting the image or sequence of images into a plurality of regions of interest;

the segmentation of the image or sequence of images being the result of the performance of a method for segmenting a three-dimensional image or sequence of three-dimensional images of an object according to claim 1;

calculating a plurality of criteria to discriminate the particular structures to be detected for each region of interest;

calculating an identification function depending on the plurality of criteria for each region of interest, the identification function being a parametric function depending on a set of parameters;

and in that the plurality of criteria comprises a shape criterion for the regions of interest so as to discriminate the particular structures among the regions of interest.

17. The method for detecting a particular structure according to claim 16, wherein segmenting the image or sequence of images into a plurality of regions of interest is the result of the performance of a method for segmenting an image or sequence of three-dimensional images of an object according to claim 3.

18. The method for detecting a particular structure according to claim 16, wherein it comprises a step for classifying regions of interest according to a probability of being a particular structure to be detected from the calculation of the identification function.

19. A computer device for segmenting a three-dimensional image or sequence of images of an object acquired by an imaging device comprising:

a first device for dividing the image or sequence of images into a plurality of regions;

a second device for hierarchical linkage of the regions resulting from the division so as to obtain the three-dimensional image or sequence of images partitioned into regions of interest;

wherein the second device comprises a linkage device using shape and size criteria for the regions so as not to link the small regions.

20. A computer device for detecting at least one particular structure on an image or sequence of images, wherein it comprises:

a device for segmenting the image or sequence of images according to claim 19 configured to segment it into a plurality of regions of interest;

a detector comprising;

a first calculator to calculate a plurality of criteria to discriminate the particular structures to be detected for each region of interest;

the plurality of criteria comprising a shape criteria for the regions of interest so as to discriminate the particular structures among the regions of interest;

a second calculator to calculate an identification function depending on the plurality of criteria for each region of interest, the identification function being a parametric function depending on a set of parameters.

21. The computer detection device according to claim 20, wherein it includes:

a first device for automatic adaptation of a color palette for a region of interest of the image or sequence of images designated by an operator;

the device for automatic adaptation of a color palette being configured to help the operator establish a diagnosis; and

a man-machine interface comprising a display configured to display information generated by the automatic adaptation device of the color palette.

22. The computer detection device according to claim 21, wherein the man-machine interface comprises a display for displaying structures detected and classified by the detector.

23. The detection device according to claim 20, wherein it comprises a quantification device for diagnostic parameters in a structure designated by the operator, and a comparison device for comparing the diagnostic parameters between at least two images;

and the man-machine interface comprises a display for displaying diagnostic parameters and a result of the comparison done by the comparison device.