Image processing device and method for detecting developing lesions

• US 7,822,240 B2
• Filed: 10/04/2002
• Issued: 10/26/2010
• Est. Priority Date: 10/12/2001
• Status: Expired due to Fees

• Current Assignee: Inria Institut National de Recherche EN Informatique et EN Automatique
• Original Assignee: Inria Institut National de Recherche EN Informatique et EN Automatique
• Inventors: Ayache, Nicholas, Rey, David, Malandain, Gregoire, Stoeckel, Jonathan
• Primary Examiner(s): Mehta, Bhavesh M
• Assistant Examiner(s): ZARKA, DAVID PETER
• Application Number:
  US10/492,015
• Publication Number:
  US 20050141757A1
• Time in Patent Office: 2,944 Days
• Field of Search: 382/128
• US Class Current: 382/128
• CPC Class Codes: G06T 2207/10072 Tomographic images
• CPC Class Codes: G06T 2207/30016 Brain
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/30 Determination of transform parameters for the alignment of images, i.e. image registration

Abstract

The invention concerns an image processing device, including an input for receiving a time series of data sets representing comparable volume digital images, each data having a position component and an intensity component, pre-processing means for modifying the data sets so as to obtain images updated in position and intensity, and comparative processing means for examining sets of time series of image elements and to detect therein signs of variations. The processing means include a modeling function for adjusting a parametric model separately on some of the sets of time series of image elements, to obtain pairs of data (image element, time) and a function for statistical analysis of said pairs of data to isolate the pairs of data representing a significant variation.

Claims

1. An image processing device, comprising:

an input for receiving a time series of data sets representing comparable volume-based digital images, each data set comprising a position component and an intensity component,

preprocessing means for modifying the data sets so as to obtain images registered geometrically and intensity-wise, and

comparative processing means for examining sets of time series of image elements, so as to detect signs of variation therein, characterized in that the comparative processing means comprise;

a modeling function, fitting a predetermined parametric model separately over at least one of the sets of time series of image elements, thereby providing information pairs of the kind;

image element, time, wherein the predetermined parametric model is of the asymmetric Gaussian type, comprising a maximum amplitude parameter, a minimum amplitude parameter, a reference width parameter and a dissymmetry parameter; and

a function for statistical analysis of the information pairs, which isolates those of the information pairs that are representative of a significant variation; and

outputting means for displaying the obtained images.

2. The device as claimed in claim 1, wherein the predetermined parametric model further comprises a time at the peak parameter.

3. The device as claimed in claim 2, wherein the predetermined parametric model comprises a mean model whose parameters are established on a basis of sets of time series of image elements measured over developing pathological zones.

4. The device as claimed in claim 3, wherein said parameters of the mean model are obtained on the basis of sets of time-normalized and amplitude-normalized time series.

5. The device as claimed in claim 3, wherein the modeling function comprises an optimization function applied to the parameters of the parametric model.

6. The device as claimed in claim 1, wherein the modeling function furthermore comprises a general linear model applied to each set of time series of image elements so as to obtain, for each of said sets of series, a linear combination of at least a mean model (X₁) and a constant (X₂) according to a vector comprising at least a first and a second coefficient.

7. The device as claimed in claim 6, wherein the statistical analysis function comprises an operation of a T-test type revealing connected pathologically developing zone where said first coefficient is significantly different from zero.

8. The device as claimed in claim 1, wherein the modeling function furthermore comprises fitting at least one parametric model for at least one time series of image elements.

9. The device as claimed in claim 8, wherein at least one parameter of the model is fixed on the basis of sets of time-normalized and amplitude-normalized time series of image elements.

10. The device as claimed in claim 8, wherein at least one parameter of the predetermined parametric model is obtained on a basis of the set of the time series of said image element.

11. The device as claimed in claim 1, wherein the statistical analysis function comprises repeating, for at least one time series in a different time order from said image element, the modeling function comprising fitting a parametric model for a time series of an image element.

12. The device as claimed in claim 11, wherein the statistical analysis function comprises the determination of groups of neighboring image elements corresponding to series of image elements having at least one of the parameters of their model greater than a given threshold, the groups of neighboring image elements forming developing pathological zones of a size determined by the statistical analysis function.

13. The device as claimed in claim 12, wherein the statistical analysis function furthermore comprises a distribution of said developing pathological zones according to their size and a determination of significant developing pathological zones.

14. The device as claimed in claim 1, wherein the preprocessing comprises a rigid registration, followed by a correction of the temporal bias, characterized in that the correction of a temporal bias comprises the construction of a 3D joint histogram of the intensities of images taken pairwise.

15. The device as claimed in claim 14, wherein the correction of the temporal bias comprises a construction of a regression line on said 3D joint histogram by a robust orthogonal least squares technique.

16. The device as claimed in claim 14, wherein the correction of the temporal bias comprises a consequential correction of one of the images with respect to the other.

17. A method of image processing, comprising the following steps:

a—

receiving, with one or more computers, a time series of data sets representing comparable volume-based digital images, each data set comprising a position component and an intensity component,

b—

modifying, with one or more computers, the data sets so as to obtain images registered geometrically and intensity-wise, wherein the modifying step comprises complementary steps,

c—

fitting, with one or more computers, a predetermined parametric model separately over at least one of the sets of time series of image elements, thereby providing information pairs of the kind;

image element, time, wherein the predetermined parametric model is of the asymmetric Gaussian type, comprising a maximum amplitude parameter, a minimum amplitude parameter, a reference width parameter and a dissymmetry parameter and

d—

isolating those of the information pairs that are representative of a significant variation by statistical analysis.

18. The method as claimed in claim 17, wherein the predetermined parametric model of step c—

further comprises a time at the peak parameter.

19. The method as claimed in claim 18, wherein the parametric model of step c—

comprises a mean model whose parameters are established on the basis of sets of time series of image elements measured over developing pathological zones.

20. The method as claimed in claim 19, wherein said parameters of the mean model of step c—

are obtained on the basis of sets of time-normalized and amplitude-normalized time series.

21. The method as claimed in claim 19, wherein step c—

comprises optimizing the parameters of the parametric model.

22. The method as claimed in claim 17, wherein step c—

comprises c1—

applying a general linear model to each set of time series of image elements so as to obtain, for each of said sets of series, a linear combination of at least a mean model and a constant according to a vector comprising at least a first and second coefficient.

23. The method as claimed in claim 22, wherein step d—

comprises d1—

applying to the information pairs an operation of a T-test type revealing connected pathologically developing zones where said first coefficient is significantly different from zero.

24. The method as claimed in claim 17, wherein step c—

comprises

c1—

fitting at least one parametric model for at least one time series of image elements.

25. The method as claimed in claim 24, wherein step c1 —

furthermore comprises fixing at least one parameter of models on the basis of sets of time-normalized and amplitude-normalized time series.

26. The method as claimed in claim 24, wherein step c1—

comprises obtaining at least one parameter of the model on a basis of the set of the time series of said image element.

27. The method as claimed in claim 24, wherein step d—

comprises

d4—

distributing developing pathological zones according to their size,

d5—

determining one or more significant developing pathological zones.

28. The method as claimed in claim 17, wherein step d-comprises

d1—

repeating step c—

for at least one series in a different time order from said image element.

29. The method as claimed in claim 28, wherein step d—

comprises

d2—

determining groups of neighboring image elements corresponding to series of image elements having at least one of the parameters of their model greater than a given threshold,

d3—

determining the size of developing pathological zones formed by these groups of neighboring image elements.

30. The method as claimed in claim 17, in which step b—

comprises performing a rigid registration, followed by a correction of the temporal bias, characterized in that step b—

comprises performing the correction of a temporal bias by the construction of a 3D joint histogram of the intensities of image taken pairwise.

31. The method as claimed in claim 30, wherein step b—

comprises performing the correction of the temporal bias by a construction of a regression line on said histogram by a robust orthogonal least squares technique.

32. The method as claimed in claim 30, wherein step b—

comprises performing the correction of the temporal bias by the consequential correction of one of the images with respect to the other.

33. The method as claimed in claim 17, further comprising an initial step of imaging a human body during a predetermined period of time to produce a time series of data sets.