Method for real-time correction of geometric distortions of magnetic-resonance-thermometry data, and magnetic-resonance-imaging system implementing this correction method

Abstract

The invention relates to a method for correcting geometric distortions of thermometry data in magnetic-resonance imaging obtained by means of fast echo-planar sequences, comprising the following steps: - a first series (I) of acquisitions of reference images by MRI, with phase encoding in one direction (e.g. R->L), - a second series (II) of acquisitions of thermometry images by MRI, with phase encoding in the opposite direction to step one (e.g. L->R), carried out with imaging parameters identical to those used in the first series, - an estimation (III, IV) of the geometric distortions, on the basis of the magnitude components of the reference images and of the thermometry images thus acquired, so as to determine geometric-distortion correction data, and - an application (V) of the geometric-distortion correction data to thermometry images and / or thermometry results obtained by processing these thermometry images.

Description

Method for real-time correction of geometric distortions of magnetic resonance thermometry data, and magnetic resonance imaging system implementing this correction method FIELD OF THE INVENTION

[0001] The present invention relates to a method for real-time correction of geometric distortions of magnetic resonance thermometry data. It also relates to a magnetic resonance imaging system implementing this correction method. The field of the invention is image processing, more particularly registration and non-linear deformations. STATE OF THE ART

[0002] Geometric distortions of thermometry data are a recurring problem when using fast MRI (Magnetic Resonance Imaging) sequences such as echo planar imaging (EPI). These sequences induce geometric distortions in the image they generate, in reference to the. EPI imaging is one of the most commonly used MRI techniques in neuroscience. These fast acquisition times allow, for example, the study of functional brain activation or efficient measurement of white matter tractography via diffusion-weighted MRI. This time efficiency is made possible by the use of a low bandwidth per pixel in the phase-encoding direction. This leads to significant geometric distortions in regions of inhomogeneity of the main magnetic field (B0).Lower bandwidth (i.e., longer inter-echo spacing) worsens distortions; distortions in some regions can be as large as 5–10 mm [1] “Sources of distortion in functional MRI data” depending on the phase encoding direction. B0 inhomogeneities and resulting distortions are most significant at the interfaces of different tissue types (e.g., brain, bone, air) in regions such as the orbitofrontal cortex and temporal lobes. Inhomogeneities also scale linearly with B0 field strength, so geometric distortions may be more severe at 7 Tesla than at 3 Tesla.

[0003] This known issue is the subject of numerous attempts to fix it. We can thus cite the publication [2] “Real-time geometric distortion correction for interventional imaging with echo-planar imaging (EPI)”, [3] “How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion tensor imaging”, [4] the publication “Block-Matching Distortion Correction of Echo-Planar Images With Opposite Phase Encoding Directions”, [5] “Fiber tracking in the cervical spine and inferior brain regions with reversed gradient diffusion tensor imaging”, [6] “Evaluation of six phase encoding based susceptibility distortion correction methods for diffusion MRI. »

[0004] Several aspects of image quality / characteristics need to be taken into account, given that the correction methods proposed in the state of the art concern the brain and are not easily transposable to other organs, for the reasons explained below.

[0005] For most organs involved in imaging involving thermometry, the field of view is reduced in the slice dimension compared to conventional brain applications. These often acquire a 3D volume with 20-30 slices to cover a good part of the brain, with low anisotropy between the in-plane and in-slice resolution. In interventional applications targeted in the brain, liver, heart, prostate, kidney or thyroid, only 8-12 slices are acquired with relatively high anisotropy, for example 2mm² in the plane compared to 3-5mm in the slice dimension.

[0006] Furthermore, the brain is an organ with very comparable characteristics between individuals. This is not the case for the abdomen, in which the organs present great complexity and variety: male / female, body size, organ characteristics / pathologies. The use of such algorithms in interventional imaging will have to demonstrate great robustness in these respects.

[0007] Furthermore, it appears necessary to propose a correction method that can be applied to MRI scanners without modifying the sequences. The correction method disclosed in the prior art document [2] requires modifying the sequence, which makes it very difficult to exploit for commercial purposes.

[0008] Furthermore, the problem of geometric distortions is critical for interventional imaging when it comes to superimposing / correlating / comparing / analyzing images with low distortions (anatomy) with temperature images with higher distortions. Anatomical inconsistencies are then visible and can mislead the practitioner during the therapeutic procedure. For example, in the case where the practitioner performs a segmentation of the area to be targeted on an image without distortions and performs temperature and thermal dose monitoring without corrections to define desired ablation margins: phenomena of under- or over-treatment of the target are then inevitable.

[0009] The objective of the present invention is thus to propose a method for real-time correction of geometric distortions in magnetic resonance thermometry data, which does not require modification of the sequences and which is adapted to the characteristics of the imaged organs. It is then possible to use the thermometric information precisely to treat patients precisely.

[0010] DEFINITIONS

[0011] Phase coding: In MRI, the image is coded in 3 dimensions: reading, phase, and slice. These directions are characteristic in MRI because they define how the image is constructed. In a 2D MRI image, we always find reading and phase encoding according to the columns and rows or vice versa. The slice direction is used to characterize the 3rd dimension. It is according to the phase direction that distortion phenomena are most visible.

[0012] Real-time: Compatible with online use during treatment at the foot of the MRI: the initial calculation takes a few tens of seconds and the application of corrections is less than a second.

[0013] Dynamic: A temporal volume, among all the image volumes acquired during a treatment follow-up. Direction L->R or R->L:

[0014] In the example of an axial image with column-wise phase encoding, we are in the Left-Right (L->R) direction of the patient. At acquisition, the image can be acquired (in the phase direction) from left to right or from right to left (L->R or R->L), with reference to the document "Evaluation of six phase encoding based susceptibility distortion correction methods for diffusion MRI" [5] or to the document "MRI Physics: Spatial Localization" [6]. What is demonstrated in the literature is that the distortions are opposite in these two acquisitions. The same applies to the anteroposterior direction A->P or P->A and head-foot H->F or F->H. Echo planar imaging:

[0015] A type of magnetic resonance acquisition that aims to acquire multiple lines of frequency space after each radiofrequency pulse emitted by the scanner. This sequence differs from the gradient echo sequence where a single line is acquired, considerably increasing image acquisition time.

[0016] Segmented: Echo-planar imaging acquiring multiple lines of frequency space

[0017] Single Shot: Echo-planar imaging acquiring all lines in frequency space

[0018] Volume: A set of slices or block of voxels, which can be acquired in various ways.

[0019] This objective is achieved with a method for correcting geometric distortions of thermometry data in magnetic resonance imaging obtained by means of rapid echo planar type sequences, comprising the following steps: a first series (I) of acquisitions of reference images by MRI, with phase coding in one direction (eg R->L), a second series (II) of acquisitions of thermometry images by MRI, with phase coding in the other direction (egL->R), carried out with imaging parameters identical to those used in the first series, an estimation (III, IV) of the geometric distortions, from the magnitude components of said reference images and said thermometry images thus acquired, comprising a first step (III) of coarse estimation of the geometric distortion implementing an inverted gradient algorithm and a second step (IV) of fine estimation of the geometric distortion implementing a block matching algorithm for echo planar imaging (EPI) distortion correction, so as to determine geometric distortion correction data, and an application (V) of the geometric distortion correction data to thermometry images and / or thermometry results from the processing of these thermometry images.

[0020] In a particular configuration, the method for real-time correction of geometric distortions of magnetic resonance thermometry data obtained by means of rapid echo planar type sequences, comprises the following steps: a first series of acquisitions of reference images by MRI, with phase coding in one direction (eg R->L), a second series of acquisitions of MRI thermometry images, with phase coding in the opposite direction (eg L->R), carried out with imaging parameters identical to those used in the first series, a verification of the validity of the two acquisitions previously described to carry out the correction.The information relating to the geometry of the acquisition and its properties are analyzed, an estimation of the geometric distortions, from the magnitude components of said reference images and said thermometry images thus acquired, so as to determine geometric distortion correction data, and an application of the geometric distortion correction data to thermometry images and / or thermometry results resulting from the processing of these thermometry images.

[0021] After correction, the spatial information of the thermometry is then in agreement with the anatomical information obtained through undistorted sequences.

[0022] Distortion correction involves finding very large and local distortions. These displacements can be extremely difficult to estimate with conventional registration algorithms.

[0023] The first coarse estimation step can advantageously implement an algorithm disclosed in the article “Fiber tracking in the cervical spine and inferior brain regions with reversed gradient diffusion tensor imaging” by HU Voss et al. []. The geometric distortions Δy are directed only in the phase encoding direction y. Thus for two images with reversed phase encoding gradients, Δy is also reversed. Therefore, for each voxel on the frequency encoding axis, the reversed gradient procedure yields an image of intensity I + (y + ) with voxels in y position + =y+Δy and another intensity image I − (y − ) with voxels in y position − =y−Δy. After acquiring the two images and defining Δy, in the post-processing step, the distortion-corrected image y can be calculated.

[0024] The second step of fine estimation can advantageously implement an algorithm disclosed in the article “Block-Matching Distortion Correction of Echo-Planar Images With Opposite Phase Encoding Directions” by R. Hédouin et al. [https: / inria.hal.science / hal-01436561 / document].

[0025] In a variant of the invention, the distortion correction method may further comprise a step of temporal averaging of the reference images as well as temporal averaging of the thermometry images. The averaging step improves the quality of the distortion estimation and thus the final result of the correction.

[0026] In a variant of the invention, the estimation method can be carried out after masking a part of the anatomy located close to the treatment area or in an area which is of interest for thermometric monitoring.

[0027] The distortion correction method according to the invention can implement two image acquisition modes to provide geometric distortion correction on a static organ as well as on a mobile organ.

[0028] In an implementation of the distortion correction method according to the invention for imaging a mobile organ of a subject (liver, kidney), a first acquisition mode is provided in which a volume of several sections of the mobile organ is acquired at each respiratory cycle of the subject.

[0029] In a second acquisition mode, a section of the moving organ is acquired using EPI (Echo Planar Imaging) in a single shot of sufficiently short duration so that the movement of said moving organ during the acquisition is negligible.

[0030] In this particular configuration, the distortion correction method according to the invention further comprises, following the acquisition step, a step for carrying out a temporal resetting to freeze the movement of the mobile organ in a common respiratory state.

[0031] We can thus propose a correction of geometric distortions on mobile organs such as the liver, with the following two acquisition modes:with synchronization: the volume acquisition is carried out during the pause period of physiological movement.without synchronization: a very rapid acquisition is made in EPI sequence, "single shot", a slice in approximately 100 ms, several slices are then acquired and a temporal registration by multi-slice optical blur registration algorithm is applied to freeze the organ in a reference state.

[0032] The distortion correction method according to the invention can be applied to temperature imaging and / or thermal dose imaging, carried out with or without acquisition synchronization. In all application cases, a step of verifying the characteristics is necessary, it will include among other things a step of verifying the geometry of the input images.

[0033] When the distortion correction method according to the invention is applied to temperature imaging, at the end of this correction, the temperature information is spatially correlated to an underlying anatomy, thus allowing the definition of margins with respect to critical anatomical structures to be preserved. The distortion correction method according to the invention can then further comprise a step of interpolating a transformation matrix on thermometric images.

[0034] The distortion correction method according to the invention can be advantageously applied to thermal dose images, produced with or without synchronization of the acquisition, so that thermal dose information is spatially correlated to an underlying anatomy, thus allowing the definition of margins with respect to critical anatomical structures to be preserved or, conversely, to verify the existence of sufficient carcinological margins in the case of tumor ablation.

[0035] The distortion correction method can also be used during hyperthermia procedures to verify that the temperature rise in specific anatomical areas does not exceed the threshold for tissue damage.

[0036] According to another aspect of the invention, a magnetic resonance imaging system is proposed, implementing the method for correcting geometric distortions of thermometry data obtained by means of rapid echo planar type sequences according to the invention, comprising: means for acquiring a first series (I) of reference images by MRI, with phase encoding in one direction (eg R->L), means for acquiring a second series (II) of thermometry images by MRI, with phase encoding in the other direction (egL->R), carried out with imaging parameters identical to those used in the first series, means (III, IV) for estimating geometric distortions, from the magnitude components of said reference images and said thermometry images thus acquired, so as to determine geometric distortion correction data, said means for estimating a geometric distortion comprising first means (III) for coarse estimation of the geometric distortion implementing an inverted gradient algorithm and first means (IV) for fine estimation of the geometric distortion implementing a block matching algorithm for echo planar imaging (EPI) distortion correction, and means for applying geometric distortion correction data to thermometry images and / or thermometry results from the processing of these thermometry images.

[0037] In a first embodiment, its first and second means for estimating geometric distortions (III, IV) can be arranged in a computer remote from the image acquisition means and the application means and cooperate with these first and second acquisition means and these application means via a computing cloud (Cloud).

[0038] In a second embodiment, the first and second means for estimating geometric distortions can be integrated within a magnetic resonance imager (MRI). DESCRIPTION OF FIGURES

[0039] Laillustrates the problem of image distortions, known in the prior art;Larepresents the basic principle of the distortion correction method according to the invention;Laillustrates schematically the steps of an exemplary embodiment of the distortion correction method according to the invention; andLadescribes an exemplary embodiment of an algorithm providing a fine estimation of the geometric distortion, disclosed in document [4].

[0040] DETAILED DESCRIPTION OF AN EXAMPLE OF IMPLEMENTATION

[0041] The method for correcting geometric distortion of images according to the invention is built around a basic principle aiming, with reference to, to acquire, on the one hand, reference scans using a phase encoding direction (eg R->L), and on the other hand, thermometry scans using an opposite phase encoding direction (eg L->R).

[0042] In an exemplary embodiment illustrated by the, the distortion correction method comprises a step I of acquiring reference images comprising for example 5 dynamics acquired with phase encoding (eg R->L), a step II of acquiring thermometry images comprising for example 100 dynamics acquired with phase encoding (eg L->R).

[0043] The images thus acquired are then subjected to a step III of coarse estimation and then to a step IV of fine estimation of the geometric distortion. The geometric distortion corrections are then applied (step V) to thermometry volumes (magnitude, temperature, thermal dose).

[0044] For each line in the phase encoding direction, for both encoding directions: Calculation of the normalized cumulative intensity. where L1 and L2 are the line intensities of the phase encoding images, α1 and α2 are constants. A cubic interpolation is used to find y 1,n and there 2,n such thatN1(y 1,n ) = N2(y 2,n ) = X n with x_n between 0 and 1For each position y n = (y 1,n + y 2,n ) / 2, the transformation is calculated as follows:

[0045] With reference to the, the fine estimation step can implement an algorithm comprising a set of iterations on original images I F , I B(steps 1,3) and forward transformations (step 2.1) and backward (or inverse) transformations (step 2.2), block matching steps in log-Euclidean space involving a step (4) of resampling the reference images and a step (5) of resampling the thermometry images, then steps (6-8) to symmetrize the transforms and return to regular space.

[0046] Given below is a practical example of implementing a block-matching algorithm for EPI distortion correction, implemented in the fine estimation step, disclosed in [4]:

[0047] ______________________________________________________________________________for p = 1...P, iterate over levels of the pyramid, dofor l = 1...L, iterations, doResample the images to obtain I F,l−1 and I B,l−1 Estimate the local transformations for each block on I B,l−1: A+ ← block match (I B,l−1 , I F,l−1 ) Estimate the local transformations for each block on I F,l−1 : A− ← block matching (I F,l−1 , I B,l−1 )Extrapolate dense and asymmetric SVF (Singular Value Filter) updates from A+ and A−:δS+ ← extrapolate(A+),δS− ← extrapolate(A−)Compute a symmetric SVF update: δS, and compose it with the current transformationsEnsure that T +,l and T −,l are symmetrical oppositeRegularize (elastic type) T +,l and T −,l

[0048] ______________________________________________________________________________

[0049] The distortion correction method according to the invention can use an Anima library such as https: / github.com / Inria-Empenn / Anima-Public.

[0050] It should be noted that in interventional MRI, there has been no use of the Anima library, which is mainly used in clinical research in neuroscience and neurodiagnostics. In the state of the art, the company Profound Medical Corp implements segmented acquisition, but with low acceleration factors, in practice less than 8.

[0051] With the correction method according to the invention, it becomes possible to achieve a higher acceleration factor, up to the single-shot EPI sequence

[0052] Of course, the present invention is not limited to the exemplary embodiment which has just been described and many other embodiments can be envisaged without departing from the scope of the invention. REFERENCES

[0053] https: / onlinelibrary.wiley.com / action / showCitFormats?doi=10.1002%2F%28SICI%291097-0193%281999%298%3A2%2F3%3C80%3A%3AAID-HBM2%3E3.0.CO%3B2-Chttps: / hal.archives-ouvertes.fr / hal-01503905 / documenthttps: / pubmed.ncbi.nlm.nih.gov / 14568458 / https: / hal.inria.fr / hal-01436561 / documenthttps: / pubmed.ncbi.nlm.nih.gov / 16563951 / https: / www.biorxiv.org / content / 10.1101 / 766139v1.full.pdfhttp: / xrayphysics.com / spatial.html#:~:text=The%20y%2Daxis%20of%20K,and%20below%20the%20center%20axis.

Claims

Method for correcting geometric distortions of thermometry data in magnetic resonance imaging obtained by means of rapid echo planar type sequences, comprising the following steps:a first series (I) of acquisitions of reference images by MRI, with phase encoding in one direction (eg R->L),a second series (II) of acquisitions of thermometry images by MRI, with phase encoding in the other direction (egL->R), carried out with imaging parameters identical to those used in the first seriesan estimation (III, IV) of the geometric distortions, from the magnitude components of said reference images and said thermometry images thus acquired, comprising a first step (III) of coarse estimation of the geometric distortion implementing an inverted gradient algorithm and a second step (IV) of fine estimation of the geometric distortion implementing a block matching algorithm for echo planar imaging (EPI) distortion correction, so as to determine geometric distortion correction data, andan application (V) of the geometric distortion correction data to thermometry images and / or thermometry results from the processing of these thermometry images. Distortion correction method according to the preceding claim, characterized in that it further comprises a step for temporally averaging reference images and a step for temporally averaging thermometry images. A distortion correction method according to any one of the preceding claims, characterized in that it is applied to temperature images, so that temperature information is spatially correlated to underlying anatomy. Distortion correction method according to the preceding claim, characterized in that it further comprises a step of interpolating a transformation matrix on thermometric images. A distortion correction method according to any one of the preceding claims, characterized in that it is applied to thermal dose images, such that thermal dose information is spatially correlated to underlying anatomy, Distortion correction method according to any one of the preceding claims, implementing two image acquisition modes to provide geometric distortion correction on a static organ as well as on a mobile organ. Distortion correction method according to the preceding claim, implemented for imaging a moving organ of a subject, characterized in that in a first acquisition mode, a volume of several sections of said moving organ is acquired at each respiratory cycle of said subject. Distortion correction method according to the preceding claim, characterized in that in a second acquisition mode, a section of the moving member is acquired in EPI imaging (Echo Planar Imaging) in a single shot of sufficiently short duration so that the movement of said moving member during the acquisition is negligible. Distortion correction method according to the preceding claim, characterized in that it further comprises, following the acquisition step, a step for carrying out a temporal resetting to freeze the movement of the mobile organ in a common respiratory state. Magnetic resonance imaging system, implementing the method of correcting geometric distortions of thermometry data obtained by means of rapid sequences of the echo planar type, comprising:means for acquiring a first series (I) of reference images by MRI, with phase encoding in one direction (eg R->L),means for acquiring a second series (II) of thermometry images by MRI, with phase encoding in the other direction (egL->R), carried out with imaging parameters identical to those used in the first series, means (III, IV) for estimating geometric distortions, from the magnitude components of said reference images and said thermometry images thus acquired, so as to determine geometric distortion correction data, said means for estimating a geometric distortion comprising first means (III) for coarse estimation of the geometric distortion implementing an inverted gradient algorithm and first means (IV) for fine estimation of the geometric distortion implementing a block matching algorithm for echo planar imaging (EPI) distortion correction, and means (V) for applying geometric distortion correction data to thermometry images and / or thermometry results from the processing of these thermometry images. Imaging system according to one of the two preceding claims, characterized in that the first and second means for estimating the geometric distortions (III, IV) are arranged in a computer remote from the image acquisition means and the application means and cooperate with said first and second acquisition means and said application means via a computer cloud (Cloud). Imaging system according to one of claims 10 or 11, characterized in that the first and second means for estimating geometric distortions are integrated within a magnetic resonance imager (MRI).

Images