METHOD FOR RECONSTRUCTING A SIGNAL FROM DISTORTED EXPERIMENTAL MEASUREMENTS AND DEVICE FOR ITS IMPLEMENTATION

• US 20110022375A1
• Filed: 11/14/2008
• Published: 01/27/2011
• Est. Priority Date: 11/14/2007
• Status: Active Grant

• Current Assignee: Universite De Lorraine (Government of France)
• Original Assignee: Universite Henri Poincare Nancy -1
• Inventors: Odille, Freddy, Felblinger, Jacques, Vuissoz, Pierre-Andre
• Granted Patent: US 9,140,770 B2
• Time in Patent Office: Days
• US Class Current: 703/13
• CPC Class Codes: G01R 33/5608 Data processing and visualization specially adapted for MR, e.g. for feature analysis and pattern recognition on the basis of measured MR data, segmentation of measured MR data, edge contour detection on the basis of measured MR data, for enhancing measured MR data in terms of signal-to-noise ratio by means of noise filtering or apodization, for enhancing measured MR data in terms of resolution by means for deblurring, windowing, zero filling, or generation of gray-scaled images, colour-coded images or images displaying vectors instead of pixels image data processing or generation, in general G06T
• CPC Class Codes: G01R 33/5611 Parallel magnetic resonance imaging, e.g. sensitivity encoding [SENSE], simultaneous acquisition of spatial harmonics [SMASH], unaliasing by Fourier encoding of the overlaps using the temporal dimension [UNFOLD], k-t-broad-use linear acquisition speed-up technique [k-t-BLAST], k-t-SENSE structural details of arrays of sub-coils G01R33/3415
• CPC Class Codes: G01R 33/56509 due to motion, displacement or flow, e.g. gradient moment nulling G01R33/567 takes precedence
• CPC Class Codes: G01R 33/5676 Gating or triggering based on an MR signal, e.g. involving one or more navigator echoes for motion monitoring and correction

Abstract

A method for acquiring (1) experimental measures with interferences of a physical phenomenon, and for reconstructing (2) a point-by-point signal (3) representative of the phenomenon according to at least one dimension that can vary during the experimental measure acquisition, using at least one simulation model (4) of at least one acquisition chain of the experimental measures including at least one interference, and at least one model (8) of each interference in each acquisition chain, each interference model being determined at least from the measures themselves, wherein the simulation and interference models include adjustable parameters (6, 10) depending on experimental conditions, wherein at least one adjustable parameter of one of the models is coupled to at least one adjustable parameter of the other model, and in that the adjustable parameters are optimized in a coupled manner. A device for MRI imaging, NMR, or medical imaging using such a method is described.

Claims

1-41. -41. (canceled)

42. Method for acquiring (1) experimental measurements distorted by a physical phenomenon and for reconstructing (2) a signal (3) representing this physical phenomenon, implementing:

at least one acquisition chain of said experimental measurements, called measurements of primary interest, said acquisition chain comprising at least one distortion,

at least one model (4) simulating said acquisition chain, said simulation model (4) receiving a reconstructed signal (6) as input and providing simulated experimental data (7) as output, and

at least one distortion model (8) of said distortion of said acquisition chain, said distortion model (8) being determined at least from the experimental measurements themselves,

wherein;

said simulation (4) and distortion (8) models comprise adjustable parameters (6,10) dependent on the experimental measurements,

said simulation (4) and distortion (8) models are coupled in the sense that the simulation model (4) depends (5) on at least one adjustable parameter (10) of the distortion model (8), and

said adjustable parameters (6,10) are optimized (2) jointly, said optimization providing the signal (3) and the parameters (15) of the distortion model as output.

43. Method according to claim 42, further implementing means allowing experimental measurements to be acquired to supplement said measurements of primary interest.

44. Method according to claim 42, wherein the optimization (2) of said adjustable parameters (6,10) of the simulation (4) and distortion (8) models is carried out using one of the following cost functions (16):

a cost function (16) quantifying the level of distortion of the signal and the adjustable parameters of said models,

a cost function (16) quantifying the discrepancy between at least a part of the signal from the experimental measurements and at least a part of the corresponding signal reconstructed by at least one simulation model (4).

45. Method according to claim 42, wherein the optimization (2) of said adjustable parameters (6,10) of the simulation (4) and distortion (8) models comprises at least a step of minimizing a reconstruction residual defined as the discrepancy between at least a part of the signal from the experimental measurements and at least a part of the corresponding signal reconstructed by at least one simulation model (4).

46. Method according to claim 42, wherein said simulation model (4) is produced in the form of a linear operator.

47. Method according to claim 42, wherein the optimization of said parameters (6) of the simulation model (4) is carried out by solving the inverse problem of the simulation model (31), taking the experimental measurements (13) as inputs and solving the equivalent Hermitian symmetry problem.

48. Method according to claim 47, wherein the solution of said inverse problem of the simulation model (31) is carried out using an iterative method that does not require explicit knowledge of the complete Hermitian operator, such as the conjugate gradient, the generalized minimal residual (GMRES) or the biconjugate gradient.

49. Method according to claim 47, wherein:

the signal (6) from the solution of the inverse problem of the simulation model (31) is used to generate the result of the simulation model (7),

the difference between said experimental measurements of primary interest and said result of the simulation model (7) is used as the optimization criterion for said coupled simulation (4) and distortion (8) models, and

a reconstruction residual (32) is calculated from the difference between experimental measurements of primary interest and said result of the simulation model (7).

50. Method according to claim 49, wherein said reconstruction residual (32) is modelled by a residual operator, said residual operator

taking an estimate of the errors (35) made on the parameters of the distortion model as input and providing said reconstruction residual as output (32), and

being a function of the current parameters (9, 34) of the simulation and distortion models.

51. Method according to claim 50, wherein said residual operator comprises linear operators appearing in the simulation operator (17), and a composite of the gradient of the current signal (6) and the supplementary experimental measurement data.

52. Method according to claim 50, wherein said residual operator is linear for minor errors on the parameters of said distortion model.

53. Method according to claim 50, wherein:

the optimization of the parameters of the distortion model (8) is carried out by solving the inverse problem (33) described by said residual operator, said solution of the inverse problem being carried out by solving the equivalent Hermitian symmetry problem,

said estimate of the errors (35) made on the parameters of the distortion model is used to update the parameters of the distortion model (8).

54. Method according to claim 53, wherein said solution of the inverse problem (33) described by the residual operator is carried out using an iterative method that does not require explicit knowledge of the complete Hermitian operator, such as the conjugate gradient, the generalized minimal residual (GMRES) or the biconjugate gradient, and in that it comprises a regularization of one of the following types:

Tikhonov regularization applying to the parameters of the distortion model (34) plus the errors on the parameters of said distortion model,

constrain applying on the square of the norm of the spatial gradient of the parameters of the distortion model plus the errors on the parameters of said distortion model.

55. Method according to claim 42, wherein the optimization of said parameters of said simulation (4) and distortion (8) models is carried out using a fixed-point method,

each model being optimized separately, all of the others being regarded as fixed, and

said optimization being carried out alternately on each model.

56. Reconstruction method according to claim 55, wherein said optimization of the parameters of the simulation (4) and distortion (8) models is repeated according to a primary iteration (57),

the optimal parameters being those that result in the smallest residual (32), and

said primary iteration (57) being governed by a stop condition (58) which is established by any one of the following criteria:

the difference between said residuals (32) during two separate incrementations of the primary iteration (57), a predetermined minimum value of said residual (32), or a predetermined number of iterations.

57. Method according to claim 55, wherein said optimization of the parameters of the simulation (4) and distortion (8) models is repeated according to a secondary multi-resolution iteration (59) established on the basis of a level of resolution of the signal (3).

58. Method according to claim 43, wherein the simulation model (4) is produced in the form of a linear operator (17) comprising:

a spatial transformation operator (19), modelled by said distortion model (8),

an antenna sensitivity weighting operator (20),

an encoding operator (21), and

a Cartesian grid sampling operator (22).

59. Method according to claim 58, wherein said spatial transformation operator (19) is an elastic spatial transformation operator making it possible to take locally free deformations into account.

60. Method according to claim 58, wherein said spatial transformation operator (19) is modelled by at least one motion-distortion model,

said motion distorting the experimental measurements of primary interest,

said motion-distortion model capable of being broken down into a set of parameters describing the spatial dependence of the model, and at least one item of information correlated to the motion at different acquisition times of the experimental measurements,

said motion-distortion model being linear, modelling the temporal course of the parameters describing the motion by linear combinations of signals from said supplementary experimental measurements.

61. Method according to claim 60, wherein said supplementary experimental measurements are used to restrict the motion-distortion model.

62. Method according to claim 58, wherein said antenna sensitivity weighting operator (20) is modelled by at least one sensitivity-distortion model, said sensitivity-distortion model capable of being broken down into a set of parameters describing at least one antenna sensitivity map, and in that the residual operator (39) is written in the same manner as the simulation operator (17), replacing said antenna sensitivity weighting operator (20) with a diagonal operator, the diagonal of which is constituted by the resultant of the composition of the current signal (6) by a spatial transformation operator (19).

63. Method according to claim 58, wherein said encoding operator (21) is an operator of the Fourier transform type.

64. Method according to claim 43, further comprising a step of acquiring the data and preparing the reconstruction (25), said step comprising:

the acquisition of experimental measurements of primary interest and supplementary experimental measurements, said supplementary experimental measurements from at least one of an external sensor, a navigator echo, and experimental measurements of primary interest,

the synchronization of the events relating to the acquisition of the experimental measurements of primary interest and the acquisition of the supplementary experimental measurements, such as the acquisition times and the order of acquisition of the data.

65. Method according to claim 64, wherein the preparation of the reconstruction (25) comprises a sorting operation for the experimental measurements of primary interest to which similar constraints are applied,

said similar constraints being based on data of similar supplementary experimental measurements,

said sorting operation producing at least one sorted supplementary experimental datum and at least one set of indices of interest describing said sorting of the experimental measurement data of primary interest,

said experimental measurements of primary interest being combined into at least one set corresponding to said set of indices of interest.

66. Method according to claim 65, wherein at least one set of indices of interest is determined by operating a quantification of the variation space of the supplementary experimental measurement data.

67. Method according to claim 64, wherein said preparation of the reconstruction (25) comprises at least one pre-transform of the experimental measurements of primary interest, such as a Fourier transform in the direction(s) independent of the distortion caused by motion.

68. Method according to claim 65, wherein the operator of at least one of the simulation (4) and distortion (8) models is broken down into at least one sub-element indexed by said indices of interest,

said experimental acquisitions of primary interest, combined into indexed sets being the distorted experimental data to be reconstructed (13),

said data from the supplementary experimental measurements sorted according to said indices of interest being the supplementary experimental data (12) used to restrict said motion-distortion model.

69. Method according to claim 43, further comprising adjustable acquisition parameters (46),

said adjustable acquisition parameters being taken into account by the simulation and distortion models,

said adjustable acquisition parameters being a function of the parameters of the optimized distortion models,

said adjustable acquisition parameters being updated once the optimal parameters of said simulation and distortion models have been established (55),

said updating of said adjustable acquisition parameters being repeated according to an iteration governed by a stop condition.

70. Method according to claim 69, wherein:

said adjustable acquisition parameters are affine transformation parameters of an MRI acquisition volume,

said updating of said adjustable acquisition parameters is an affine transformation model of an MRI acquisition volume dependent in real time on the supplementary experimental measurements.

71. Method according to claim 42, wherein that said signal (3) is reconstructed in a form chosen from a voxel, an image, a stack of images and an MRI volume.

72. Method according to claim 42, wherein said signal (3) is reconstructed in the form of at least one spectrum of at least one NMR spectroscopy voxel.

73. Method according to claim 42, wherein said signal (3) is reconstructed in a form chosen from a voxel, an image, a stack of images and a volume from a medical images acquisition process chosen from X-ray tomography, positron emission tomography, single photon emission tomography and ultrasonography.

74. MRI device implementing the method according to claim 42, further comprising means of acquiring and processing MRI images.

75. NMR spectroscopy device implementing the method according to claim 42, further comprising means of acquiring and processing NMR spectra.

76. Medical imaging device implementing the method according to claim 42, further comprising means of acquiring and processing images.