A method for post-processing acquired sequences to correct magnetic field inhomogeneity in magnetic resonance imaging systems.

Abstract

The present invention relates to a method (100) for post-processing a biomarker (MBMK) generated from a set of M volumes Z(Δωj), each of which contains N voxels representing the magnitude of an experimental signal sampled with a frequency shift Δωj relative to the resonance frequency of water during an acquisition sequence by a magnetic resonance imaging apparatus (1). The method is provided to be performed by a processing unit of an imaging analysis system. Such a method (100) includes a correction step (130, 140) by decorrelating the biomarker (MBMK) to the excitation magnetic field (B1) of the imaging apparatus (1). The method (100) relates to the use of any magnetic resonance imaging where correction for magnetic field inhomogeneity is required.

Description

[Technical Field]

[0001] The present invention relates to a method for post-processing an acquisition sequence, optionally a plurality of acquisition sequences, wherein the acquisition sequence requires correction for inhomogeneity of the B0 static magnetic field and the B1 excitation magnetic field of a magnetic resonance imaging apparatus, more commonly known by the acronym MRI. Such post-processing methods are of course used in particular in the framework of CEST (Chemical Exchange Saturation Transfer) imaging. The use of the post-processing method according to the invention, when applied to CEST imaging, makes it possible to use this technology in the clinical field, and thus ultimately enables quantification of biomarkers associated with chemical species of interest in human or animal organs, for characterizing various lesions with altered metabolic properties such as tumors, ischemic tissues, or multiple sclerosis. The most widely used CEST technology in clinical research is called amide proton transfer-weighted MRI (APTw MRI), because its saturation mechanism generates a sufficient signal-to-noise ratio to map amide-related signals on 3-tesla clinical MRI scanners. [Background Art]

[0002] Magnetic resonance imaging (MMRI) is based on the analysis of the response of water molecule protons when excited by a magnetic field. This response depends on the environment of such protons, thereby making it possible to distinguish between several types of tissues. A nuclear magnetic resonance imaging apparatus 1, as shown in Figures 1 and 2 as non-limiting examples, is commonly used. This provides, as a non-limiting example, multiple digital image sequences 12 of one or more parts of a patient's body, such as the brain, heart, or lungs. For this purpose, the apparatus applies a combination of high-frequency electromagnetic waves to the body part under consideration and measures the signal re-emitted by a specific atom, such as hydrogen, for nuclear magnetic resonance imaging, as a non-limiting example. This allows the apparatus to determine the magnetic properties, and therefore the chemical composition of biological tissue, and thus their properties, in each basic volume, commonly called a voxel, of the imaged volume. The magnetic resonance imaging apparatus 1 is controlled by a console 2. Thus, a user 6, for example, an operator, practitioner, or researcher, can select commands 11 to control the apparatus 1 based on parameters or settings 16 entered via the input human-machine interface 8 of the analysis system AS. Such a human-machine interface 8 may consist, for example, a computer keyboard, a pointing device, a touchscreen, a microphone, or, more generally, any interface provided to convert gestures or commands given by a human 6 into control or configuration data. From the information 10 generated by the device 1, a sequence of digital images 12 of parts of a human or animal body is obtained. Such information 10 or images 12 may also be called “experimental data”.

[0003] The image sequence 12 may optionally be stored in a server 3, i.e., a computer equipped with its own storage means, to constitute a patient's medical file 13. Such a file 13 may include various types of images, such as functional images highlighting tissue activity or anatomical images showing tissue characteristics. The image sequence 12, or more generally the experimental data, is analyzed by a processing unit 4, which is provided for this purpose. Such a processing unit 4 may consist of one or more microprocessors or microcontrollers that execute appropriate application program instructions stored in storage means, such as the non-volatile memory of the imaging analysis system. The processing unit 4 includes means for communicating with the outside world to collect images. Through the communication means, the processing unit 4 can ultimately provide to the user 6 of the imaging analysis system AS, via an output human-machine interface 5, an estimated or quantitative representation of biomarkers generated by the processing unit 4 from the experimental data 10 and / or 12 resulting from magnetic resonance imaging, such as a graphic and / or audible representation. Throughout this document, the term “output human-machine interface” refers to any device, used alone or in combination, that enables the output or provision of a reconstructed physiological signal, in this case a biomarker, in a graphical, haptic, audible, or more generally human-perceptible representation, to the user 6 of the magnetic resonance imaging analysis system AS. Such a human-machine output interface 5 may, but not exhaustively, consist of one or more screens, loudspeakers, or other suitable alternative means. Thus, the user 6 of the analysis system AS may confirm or reject a diagnosis, decide on appropriate treatment actions, perform further research, fine-tune the adjustment parameters of the measuring device, etc. Optionally, the user 6 may also configure the operation of the processing unit 4 or the output human-machine interface 5 by operation and / or acquisition parameters 16. This makes it possible, for example, to define display thresholds or select biomarkers, indicators, or estimated or quantified parameters to be displayed.To do this, the user uses the aforementioned input human-machine interface 8 or a second input interface provided for this purpose. Advantageously, the input human-machine interface 8 and the output human-machine interface 5 may be one and the same physical entity. Furthermore, the input human-machine interface 8 and the output human-machine interface 5 of the imaging analysis system AS may be integrated into the acquisition console 2. There is a modification described in relation to Figure 2. In this modification, the imaging system AS as described above further includes a preprocessing unit 7 to analyze the image sequence 12, from which an experimental signal 15 is estimated, and then sent to a processing unit 4, which is thus relieved from this task.

[0004] Techniques or modalities based on magnetic resonance imaging include chemical exchange saturation transfer (CEST). Such techniques involve applying high-frequency pulses at different defined resonance frequencies ω so that a chemical species of interest reaches a saturation state. The resonance frequencies ω are, respectively, associated with different chemical species of interest, such as amide groups (NH), amine groups (NH2), or hydroxyl groups (OH). Their thus excited, unstable hydrogen atoms are exchanged with unexcited hydrogen atoms from water. By continuously repeating such application of high-frequency pulses at defined resonance frequencies ω different from the water-related frequency ω0 for a defined duration, e.g., several seconds, water saturation of the chemical species of interest accumulates. The concentration of the chemical species of interest can be indirectly measured by a decrease in the water signal, a phenomenon known as the "CEST effect," which is analyzed in the form of a Z spectrum. Such decreases can be readily detected by fast magnetic resonance imaging acquisition sequences, such as single-shot echo-planar imaging, or by other acquisition techniques, such as gradient echo imaging, fast spin echo imaging, or turbo spin echo imaging. Figure 3 shows an example of a Z spectrum in the form of a set of samples Z(Δω) following a frequency shift Δω relative to the frequency of water, where such a shift Δω is expressed in ppm. If the samples Z(Δω) are ordered according to a decreasing Δω, as shown in Figure 3, the Z spectrum may be presented in the form of a discrete signal showing a measurement of the continuous signal provided by the measuring device 1 relative to the fundamental volume of the organ, which is favorably normalized by the signal measured without high-frequency saturation and expressed as a percentage. According to Figure 3, this set of samples constitutes a discrete signal Z of substantially the shape of "V" when the samples are ordered according to a decreasing frequency shift Δω, and its minimum value Z(Δω)m is associated with a frequency shift Δω = 1 ppm. Logically, if the device is perfectly tuned and / or the static magnetic field is uniform, the minimum value Z(Δω)m should be associated with a frequency shift Δω of zero.

[0005] CEST technology improves the detection of certain metabolites in the human body whose concentrations are insufficient to detect with conventional magnetic resonance imaging sequences. Therefore, CEST technology provides valuable information for practitioners seeking to establish diagnoses and make treatment decisions in the treatment of lesions. However, in the clinical field, such biomarkers must be generated quickly and accurately. However, according to conventional techniques, performing CEST technology requires sampling multiple volumes to compensate for the heterogeneity of B0 and B1. This large number of samples results in a total acquisition time that is particularly incompatible with the requirements imposed by the clinical field.

[0006] Theoretically, to extract a signal from a molecule of interest, it is sufficient to apply high-frequency saturation to the following three positions in the Z spectrum. - The molecular resonance frequency (3.5 ppm relative to the water frequency in the case of APTw imaging). This volume can be called S(3.5 ppm) or "label". - A frequency opposite to the molecular frequency relative to the water frequency (in the case of APTw imaging, -3.5 ppm relative to the water frequency). This volume can be called S(-3.5 ppm) or "reference". - To obtain a volume without the CEST effect, the first two volumes must be normalized to a frequency far removed from the water frequency (for example, in the case of APTw imaging, the frequency may be 300 ppm relative to the water frequency). This volume can be called S0.

[0007] During the continuous application of high-frequency pulses, many effects, such as direct water saturation (DS) and magnetization transfer contrast (MT), interact with saturation transfer. Other phenomena, such as magnetic field heterogeneity, are coupled with these effects and are inherent in the execution of magnetic resonance imaging techniques. CEST contrast is based on the chemical exchange between unstable protons and water. MT contrast, as far as it is concerned, arises from the exchange of magnetization between water protons and protons in a solid or semi-solid environment. Therefore, correction is necessary to obtain a reliable CEST signal. Furthermore, theoretically, water saturation requires only three spectral positions (S(3.5 ppm), S(-3.5 ppm), and S0 in the case of APTw imaging) to extract a signal characterizing the concentration of the molecule of interest; however, in practice, such extraction is complicated due to the presence of B0 static magnetic field heterogeneity in the magnetic resonance imaging apparatus. These heterogeneities are partly caused by the patient's tissue characteristics and result in shifts in the acquired spectral samples, as shown in Figure 3.

[0008] To overcome this problem, dense sampling is performed around the resonance frequency of water, including the frequency of the molecule of interest and the inverse frequency of the molecule. Then, the measured values ​​S(Δωj) obtained at different frequencies Δωj are normalized by the measured values ​​(volume S0) obtained at frequencies far from the water frequency. With appropriate mathematical interpolation as shown in Figure 4, a Z spectrum can be obtained for each voxel over a range of frequency shifts of interest, where Ζ(Δω) = S(Δω) / S0. The samples, after ordering according to a decreasing frequency shift Δω, exhibit discrete signals with a substantially "V-shaped" shape or curve. Then, the Z spectrum is "repositioned to the center" so that its minimum value Z(Δω)m corresponds to the water frequency, as shown in Figure 5. Thus, the Z spectrum can be expressed in relative form, where the water frequency is equal to a zero frequency shift Δω, and the frequency ω is expressed in ppm as a positive or negative relative frequency shift Δω with respect to the water frequency. Such relative frequency shifts are generally expressed in ppm. Therefore, the B0 parameter correction, as illustrated in Figure 5, can be applied to multiple voxels or basic volumes. Using the shift position of the minimum Z(Δω)m value of the Z spectrum for a set of basic volumes, a map called the eigenMB0 map, i.e., a raster image, can be calculated and displayed.

[0009] Figure 6 illustrates the determination of a first non-restrictive example of a biomarker generated from the Z spectra described above in relation to Figures 3 to 5. Such a biomarker may, alternatively, correspond to the raw Z spectrum. Alternatively, such a biomarker may correspond to the difference between a first area ZA1, generally called the "label," and a second area ZA2, generally called the "reference," under the interpolation curve of the Z spectrum from the sample Z(Δω), for each range of frequency shifts Δω that are symmetric with respect to a zero frequency shift. In this case, in the example in Figure 6, area ZA1 is associated with frequency shifts between +3 ppm and +4 ppm, and area ZA2 is associated with frequency shifts between -3 ppm and -4 ppm. The selection of such ranges is typically associated with amides, which are biomarkers of interest in APTw imaging. Indeed, the resonance frequencies of amides generally correspond to a frequency shift Δω = 3.5 ppm on average. Alternatively, if we are interested in amines whose resonance frequencies generally correspond to an average frequency shift Δω = 2 ppm, the ranges may be selected such that they range from 1.5 ppm to 2.5 ppm for area ZA1 and from -1.5 ppm to -2.5 ppm for area ZA2, respectively. A parameter map associated with this biomarker may be created and optionally displayed to show the biomarker for several base volumes of interest.

[0010] Figure 7 illustrates the determination of a second non-limiting example of a biomarker generated from the Z spectra described above in relation to Figures 3 to 5. Unlike the previous example in which the biomarker was obtained by subtracting two areas ZA1 and ZA2, the biomarker may be generated by subtracting two interpolation values ​​Z1 and Z2 of a sample Z spectrum, the sample being selected such that its associated frequency shifts Δω1 and Δω2 are symmetric with respect to a frequency shift Δω=0. The value Z1 may be called the “label,” and the value Z2 may be called the “reference.” If we are interested in amides, as in the previous example, then frequency shifts Δω1=3.5ppm and Δω2=-3.5ppm are preferred, as shown in Figure 7. Alternatively, if we are interested in amides, then frequency shifts Δω1=2ppm and Δω2=-2ppm may be selected. To extract the signal of the biomarker of interest, the value of Z1 (label) is subtracted from the value of Z2 (reference). This subtraction (reference-label) theoretically allows for the removal of parasitic effects such as the effects of DS and MT. Such biomarkers are called "magnetization-transfer asymmetry (MTRasym)," as disclosed in the reference "Zhou, J., Payen, JF, Wilson, DA, Traystman, RJ & Van Zijl, PCM Using the amide proton signals of intracellular proteins and peptides to detect pH effects in MRI. Nature Medicine. 2003. 9(8):1085-1090." An alternative methodology can be used to extract the signal of the biomarker of interest: the reciprocal of value Z2 (1 / reference) is subtracted from the reciprocal of value Z1 (1 / label). This subtraction (1 / label-1 / reference) theoretically allows for more effective removal of parasitic effects such as the effects of DS and MT.Such biomarkers have been named "magnetic transfer coefficient REX (MTRREX)" (see ZAISS, Moritz, et al. Inverse Z-spectrum analysis for spillover-, MT-, and T1-corrected steady-state pulsed CEST-MRI-application to pH-weighted MRI of acute stroke. NMR in biomedicine, 2014, 27.3: 240-252). Such selections may be subject to predetermined settings or choices entered by the user through an appropriate input human-machine interface.

[0011] While increasing the density of Z spectral samples leads to higher resolution in the calculated MB0 map, acquiring multiple Z spectral samples results in considerable total acquisition time on the MRI device and is not necessarily synonymous with optimal results for mapping B0 parameters across the set of volumes of interest.

[0012] In addition to the difficulties caused by the B0 static magnetic field inhomogeneity of the magnetic resonance imaging apparatus, the Z spectrum is also affected by the B1 excitation magnetic field inhomogeneity. To overcome this problem and successfully extract a reliable CEST signal, parameter B1 correction is also commonly performed.

[0013] To perform B1 correction, the volume, which has already been sampled at different frequencies for B0 correction, must also be sampled at different values ​​of high-frequency B1, further increasing the acquisition time on the MRI device, as clearly stated in the reference "WINDSCHUH, Johannes, et al. Correction of B1-inhomogeneities for relaxation-compensated CEST imaging at 7 T. NMR in biomedicine, 2015, 28.5: 529-537".

[0014] To overcome these shortcomings, the WASAB1 sequence was proposed to enable more accurate determination of the resonant frequencies of protons in bulk water. As mentioned above, the non-uniformity of the B0 static magnetic field and / or B1 excitation magnetic field is unavoidable in magnetic resonance imaging. They cause distortion in the intensity of the acquired images. Therefore, it is necessary to improve the validity of the intensity by performing a post-processing method for the WASAB1 sequence and accurately determining the distribution of the B0 and B1 magnetic fields of the magnetic resonance imaging apparatus and the associated spatial information. Document WO2022090644A1 illustrates such a post-processing method.

[0015] WASAB1 acquisition is a magnetic resonance imaging acquisition sequence that enables simultaneous mapping of the B0 static magnetic field and the B1 excited magnetic field based on Rabi oscillations induced by off-resonance irradiation, and off-resonance irradiation can be used to correct for the aforementioned heterogeneity. The document PAPAGEORGAKIS CHRISTOS ET AL “Fast WASABI post processing: Access to rapid B0 and B1 correction in clinical routine for CEST MRI”, MAGNETIC RESONANCE IMAGING - XP087365586 presents a method for accelerating the estimation of such B0 and B1 magnetic field heterogeneity maps based on the WASAB1 acquisition sequence without applying corrections to eliminate the effects of such heterogeneity on the acquired data.

[0016] Furthermore, there are other acquisition methods for mapping the B0 static magnetic field, such as WASSR sequencing (see KIM, Mina, et al. Water saturation shift referencing (WASSR) for Chemical exchange saturation transfer (CEST) experiments. Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine, 2009, 61.6: 1441-1450), Dixon sequencing (see ZHANG, Shu, et al. CEST-Dixon for human breast lesion characterization at 3T: A preliminary study. Magnetic resonance in medicine, 2018, 80.3: 895-903), or BSS sequencing for further mapping the excitation magnetic field (see SACOLICK, Laura I., et al. B1 mapping by Bloch-Siegert shift. Magnetic resonance in medicine, 2010, 63.5: 1315-1322).

[0017] These known techniques not only allow for better mapping of the B0 static magnetic field but also reduce the volume acquired for the Z spectrum. Thus, sampling around the "label" Z1 and the "reference" Z2 may be sufficient to perform correction of the B0 magnetic field at the frequency shift of interest (e.g., 3.5 and -3.5 ppm). However, the number of volumes acquired is still large. In general, to avoid performing useless or irrelevant interpolation to compensate for the effects of B0 field heterogeneity, five Z(Δω) volumes around the "reference" and five Z(Δω) volumes around the "label," e.g., Z(-4.5 ppm), Z(-4 ppm), Z(-3.5 ppm), Z(-3 ppm), Z(-2.5 ppm), Z(2.5 ppm), Z(3 ppm), Z(3.5 ppm), Z(4 ppm), Z(4.5 ppm), are needed. On average, the time required to retrieve a CEST volume is on the order of 15 seconds. These 10 volumes Z(Δω) are understood to be already normalized by volume S0 (the 11th volume to be retrieved) according to the formula Z(Δω)=S(Δω) / S0.

[0018] The numerous acquisitions performed to compensate for B0 and B1 magnetic field heterogeneity, which should be carried out for each basic volume, result in excessively long acquisition times for clinical applications. For example, the first step, which includes compensating for B0 static magnetic field heterogeneity as described above, can result in an acquisition time of 2 to 3 minutes on a magnetic resonance imaging (MRI) system. If the time required to compensate for B1 excitation magnetic field heterogeneity is added, it generally requires repeating 10 acquisitions for each volume of interest (i.e., 5 for the "label" and 5 for the "reference," which are assumed to be acquired at a nominal B1 value equal to 2 μT in the previous example) for at least 5 values ​​of excitation magnetic field B1 (e.g., 1.2 μT, 1.6 μT, 2 μT, 2.4 μT, 2.8 μT, etc.), but the acquisition time can rise to 15 minutes on the most efficient systems (such as those in the described example) and almost 1 hour on the least efficient ones. The techniques described in MUeLLER-LUTZ ANJA ET AL: “Comparison of B0 versus B0 and B1 field inhomogeneity correction for glycosaminoglycan Chemical exchange saturation transfer imaging”, Magnetic Materials In Physics, Biology and Medicine, XP06589490, or disclosed in US 2012 / 019245 A1, illustrate the need to use multiple acquisitions for different values ​​of B1, thus disadvantageous to acquisition time. For this reason, most practitioners omit the effort of correcting for B1 excitation field inhomogeneity.

[0019] Figure 8 visually illustrates the effects of the lack or partial correction of the aforementioned heterogeneity correction for the B0 and B1 magnetic fields. The left side of Figure 8 shows a parameter map MBMK illustrating a “reference-label” biomarker BMK generated from Z spectra as described above in relation to Figures 3 through 5 and 7, for a set of basic volumes. The biomarker is obtained by subtracting two interpolated sample values ​​Z1 and Z2 from the Z spectra, the samples also selected with frequency shifts Δω1=3.5 ppm and Δω2=-3.5 ppm, symmetric with respect to frequency shift Δω=0. The center and right sides of Figure 8 show MB0 and MrB1 heterogeneity maps for the B0 and B1 magnetic fields, respectively, for the same set of basic volumes. A B0 homogeneous static magnetic field would result in an MB0 homogeneity map where N voxels would show a value of 0 ppm. However, particularly on the MB0 map in Figure 8, we can observe four regions MB0-1, MB0-2, MB0-3, and MB0-4 where the values ​​are negative (between 0 and -0.3 ppm), as well as one region MB0-5 where the voxel value reaches 0.3 ppm. This heterogeneity of the B0 static magnetic field also results in four regions MBMK-1 through MBMK-4 on the MBMK map (left-hand portion of Figure 8) where the signal is very weak or absent. If a patient has disease in these regions, the diagnosis could be significantly altered. Furthermore, the MrB1 map (right-hand portion of Figure 8) clearly shows heterogeneity of the B1 excitation magnetic field (however, if the B1 excitation magnetic field were uniform, the voxels on the MrB1 map should substantially show values ​​equal to 1). However, the voxels in question show lower voxel values ​​outside the roughly central region labeled MrB1-C, particularly around the MrB1-C region, and especially in the left frontal region, which significantly alters the accuracy of the MBMK map.

[0020] Document US2022 / 342021 describes an effective technique for compensating for B0 magnetic field inhomogeneity, primarily by relying on a shift in the B0 magnetic field value. The document briefly mentions the possibility of using such a technique for compensating for B0 magnetic field inhomogeneity to address B1 excitation magnetic field inhomogeneity, without detailing the technical basis, modality, or suitability for use. However, in light of the technical teachings objectively derived from document US2022 / 342021, it is clear that any substitution of the technical teachings would not be effective in compensating for the effects of B1 excitation magnetic field inhomogeneity, or even introduce a serious bias into the CEST signal. Document US2022 / 342021 neither describes nor claims any examples of compensating for inhomogeneity in the B1 excitation magnetic field. [Overview of the project]

[0021] The present invention makes it possible to address all or some of the shortcomings raised by known or aforementioned solutions.

[0022] Among the many advantages provided by the present invention, the following can be mentioned. - The post-processing method according to the present invention for correcting B1 magnetic field inhomogeneity requires only the MrB1 map and the MBMK map of interest to be corrected, avoiding the acquisition of numerous other volumes required for interpolation in the high-frequency region expressed in units of μT, which is required by known methods (to handle B1 excitation magnetic field inhomogeneity), and the acquisition time is reduced to less than 1 / 25th thanks to the implementation of the present invention. - The implementation of such post-processing methods opens up CEST technology, which is particularly promising in the clinical field, especially in providing support for decision-making and diagnosis related to numerous lesions. - The post-processing method according to the present invention finds direct and effective use in any magnetic resonance imaging (MRI) applications that require correction of B1 excitation field heterogeneity, such as anatomical acquisitions known as T1-weighted images (T1WI) or T2-weighted images (T2WI). The present invention makes it possible to eliminate the use of prior art algorithms for correcting heterogeneity that require execution times of several tens of seconds or expensive hardware resources, such as the algorithm disclosed in the published document NJ Tustison, BB Avants, PA Cook, Y. Zheng, A. Egan, PA Yushkevich, and JC Gee. “N4ITK: Improved N3 Bias Correction” IEEE Transactions on Medical Imaging, 29(6):1310-1320, June 2010.

[0023] For this purpose, the present invention relates, firstly, to a method for post-processing a biomarker BMK generated from a plurality of M volumes Z(Δωj), each of which contains N voxels representing the magnitude of an experimental signal sampled with a frequency shift Δωj relative to the resonance frequency of water during an acquisition sequence by a magnetic resonance imaging apparatus, and the method is performed by a processing unit of an imaging analysis system. In particular in the clinical field, such a method opens up the possibility of compensating for the effect of magnetic field heterogeneity of the imaging apparatus on the biomarker product. - A step of generating a biomarker BMK in the form of a map MBMK having N voxels, where each voxel i represents a signal value MBMK[i] and i is between 1 and N, - A step of generating MrB1 maps showing the non-uniformity of the B1 excitation magnetic field of the magnetic resonance imaging apparatus for M volumes, wherein each map includes N voxels, - A step to estimate the parameters of the function Fc such that MBMK[i]=Fc(MrB1[i]), - generating a corrected BMK biomarker in the form of a map MBMK´ of N voxels such that MBMK´[i]=MBMK[i]-Fc(MrB1[i])+Fc(1);

[0024] According to an advantageous embodiment, the step of estimating the parameters of the function Fc such that MBMK[i]=Fc(MrB1[i]) may comprise performing a least squares minimization on all or part of the set of N voxels of the maps MBMK and MrB1.

[0025] If the generated biomarker is particularly affected by B0 magnetic field inhomogeneity, the method according to the present invention may comprise, prior to the step of generating the corrected biomarker, a step of correcting the influence of B0 static magnetic field inhomogeneity on said generated biomarker.

[0026] In order to provide a human-perceptible representation of the corrected biomarker, such a method according to the present invention may comprise the steps of: generating graphic content representing the parameter map MBMK´, wherein each of the N voxels of MBMK´ encodes a corrected biomarker value; and then causing display of said graphic content via an output human-machine interface of an imaging analysis system.

[0027] According to a second aspect, the present invention relates to an imaging analysis system comprising a processing unit, means for communicating with the outside world, and storage means. Such a system is characterized in that: - the communicating means is arranged to receive experimental data resulting from an acquisition sequence performed by a magnetic resonance imaging apparatus from said outside world, - the storage means comprises program instructions which, when interpreted by said processing unit, cause execution of the biomarker post-processing method according to the present invention.

[0028] According to the third aspect, the present invention relates to a computer program comprising one or more instructions interpretable by a processing unit of such an imaging analysis system, wherein the computer program may be stored in a storage means of the system and is configured such that the interpretation of the program instructions by the processing unit executes the biomarker post-processing method according to the present invention.

[0029] According to the fourth object, the present invention relates to a computer-readable storage medium containing program instructions for such a computer program.

[0030] Other features and advantages will become clear from reading the following description and examining the attached drawings. [Brief explanation of the drawing]

[0031] [Figure 1] Figure 1 illustrates a simplified explanation of the system for analyzing images obtained by nuclear magnetic resonance, which has already been described. [Figure 2] Figure 2 illustrates a simplified explanation of a modified version of the system for analyzing images obtained by nuclear magnetic resonance, which has already been described. [Figure 3] Figure 3 shows a first example of Z spectra, which have already been described, showing the magnitude of experimental signals sampled with different frequency shifts Δω relative to the resonance frequency of water during the acquisition sequence by a magnetic resonance imaging apparatus, and the frequency shifts Δω are ordered in descending order. [Figure 4] Figure 4 shows the same example of a Z spectrum in the form of a set of sample Z(Δω) ordered similarly to that described in Figure 3, where, according to prior art, the sample of the Z spectrum is superimposed on a continuous experimental signal estimated after fitting a defined model. [Figure 5]Figure 5 shows the same example of a Z spectrum in the form of an ordered set of Z(Δω) samples as described in Figure 3, which, after correction for B0 magnetic field inhomogeneity, is superimposed on the Z spectrum samples, which were estimated after fitting a defined model according to the prior art. [Figure 6] Figure 6 shows the estimation of the first biomarker from a sample of the Z spectrum according to the conventional technique, after the imaging device has been corrected and adjusted, as previously described. [Figure 7] Figure 7 shows the estimation of a second biomarker from a sample of Z spectra according to the conventional technique, after correction and adjustment of the imaging device, as previously described. [Figure 8] Figure 8, which has already been explained, shows the effect of static and excitation magnetic field heterogeneity on the biomarker product. [Figure 9] Figure 9 shows a first example of a correlation between the product of such a biomarker and the B1 excitation magnetic field of an imaging instrument. [Figure 10] Figure 10 shows an example of performing the decorrelation of the B1 excitation magnetic field for such corrected biomarker products according to the present invention. [Figure 11] Figure 11 shows a second example of the correlation between the products of such biomarkers and the B1 excitation magnetic field of the imaging instrument, using a Gaussian distribution fitted to a scatter plot. [Figure 12] Figure 12 shows probability distribution maps related to the Gaussian distributions exemplified in Figure 11, applied to the human brain. [Figure 13] Figure 13 shows a comparison of parameter maps for the generation of the first biomarker, which are corrected considering only the non-uniformity of the B0 static magnetic field of the imaging device, and also considering the non-uniformity of the B1 excitation magnetic field of the imaging device. [Figure 14]Figure 14 shows a comparison of parameter maps for the second biomarker, which is corrected considering only the non-uniformity of the B0 static magnetic field of the imaging device, and also considering the non-uniformity of the B1 excitation magnetic field of the imaging device. [Figure 15] Figure 15 shows an example of a functional algorithm for the post-processing method according to the present invention. [Modes for carrying out the invention]

[0032] In relation to Figure 15, an example of a post-processing method 100 according to the present invention of a parameter map MBMK showing a biomarker BMK generated from a plurality of M volumes Z(Δωj) is described here. Each of the M volumes Z(Δωj) contains N voxels, respectively, representing the magnitudes of experimental signals 10, 12 sampled with a frequency shift Δωj relative to the resonance frequency of water during an acquisition sequence by a magnetic resonance imaging apparatus (e.g., apparatus 1 described above in relation to Figure 1 or Figure 2), in relation to a set of basic volumes of interest of an organ. A CEST-type acquisition selected to illustrate the method according to the present invention results in several Z(Δωj) elements. However, M can also be reduced to 1 within the framework of anatomical acquisitions (T1WI, T2WI, etc.) or other multi-phase acquisitions, e.g., ASL (arterial spin labeling) acquisition or functional magnetic resonance acquisition (also known as "functional magnetic resonance imaging").

[0033] Such a method 100 includes a first step 110 of generating a value MBMK[i] of the biomarker BMK for a set of i voxels, calculated over M fundamental volumes of interest for each voxel i. Such a biomarker BMK may be a biomarker "reference-label" generated from the Zi spectrum (e.g., the Z spectrum previously shown in Figures 3 to 5 and 7) for a set of N voxels and M fundamental volumes for each voxel i. For the record, such a biomarker BMK is obtained by subtracting two interpolated values ​​Z1 and Z2 of a sample from the Zi spectrum of voxel i of the M volumes, the sample is also selected with frequency shifts Δω1 = 3.5 ppm and Δω2 = -3.5 ppm, which are symmetric with respect to frequency shift Δω = 0. Such biomarkers BMKs can further be defined as "1 / label-1 / reference" to correct data for spillover effects known in CEST imaging (see https: / / analyticalsciencejournals.onlinelibrary.wiley.com / doi / abs / 10.1002 / nbm.3054 - or Inverse Z-spectrum analysis for spillover-, MT-, and T1-corrected steady-state pulsed CEST-MRI - application to pH-weighted MRI of acute stroke. 2014. NMR Biomed, 27: 240-252. Zaiss, M., Xu, J., Goerke, S., Khan, IS, Singer, RJ, Gore, JC, Gochberg, DF and Bachert, P.).

[0034] A biomarker may further consist simply of a labeled volume or a reference volume. Whatever the biomarker may be, the value of the biomarker BMK for the i-th volume of interest is denoted as BMK[i].

[0035] Such a method 100 according to the present invention further comprises a step 120 of generating values of the B1 excitation magnetic field of an imaging apparatus 1. Accordingly, such a step 120 comprises generating an intrinsic map MrB1, that is, generating a value MrB1[i] of said map MrB1 for each of its N voxels i (1<i<N). For this purpose, the present invention provides for the application of any known technique for generating such an intrinsic map MrB1, for example, the technique described in the document PAPAGEORGAKIS CHRISTOS ET AL "Fast WASABI post-processing: Access to rapid B0 and B1 correction in clinical routine for CEST MRI", MAGNETIC RESONANCE IMAGING - XP087365586. If the method according to the present invention optionally comprises correcting, in addition to the effect of non-uniformity of the B1 excitation magnetic field, the effect induced by non-uniformity of the B0 static magnetic field by relying on a technique disclosed for example in the document US2022 / 342021 A1, said step 120 may comprise generating values of said B0 static magnetic field of the imaging apparatus 1. In this case, such a step 120 may further comprise generating an intrinsic map MB0, that is, generating a value MB0[i] for each of its N voxels i (1<i<N).

[0036] As shown in FIG. 9 and FIG. 11, the inventors have noticed that the product of a map MBMK of any biomarker BMK correlates, whatever the biomarker BMK is (for example, "reference-label"), with the values of the N voxels of an intrinsic map MrB1 of the B1 magnetic field of the imaging apparatus.

[0037] Figure 9 shows a scatter plot with the x-axis representing the values ​​of the map MrB1, which represents the inhomogeneity of the B1 excitation magnetic field, and the y-axis representing the values ​​of the parameter map MBMK, which represents the biomarker BMK ("reference-label"). The correlation is clear. Visually, a straight line CL symbolizing this correlation can be easily estimated. Thus, the inventors concluded that the voxel value MBMK[i] of the map MBMK for the biomarker BMK can be expressed as a function Fc of the voxel of the map MrB1, which represents the inhomogeneity of the B1 excitation magnetic field. If the parameters of such a function Fc are successfully estimated, it is sufficient to correct each value of the biomarker by removing the value obtained by applying the function Fc to the B1 excitation magnetic field for the same fundamental volume of interest, in order to, so to speak, eliminate the correlation between the biomarker BMK and the B1 excitation magnetic field. Thus, the corrected value of the biomarker can be expressed in the form of the formula MBMK'[i]=MBMK[i]-Fc(MrB1[i])+c, where c is a constant. Therefore, the method 100 according to the present invention includes a step 130 of estimating the parameters of a function Fc such that MBMK[i]=Fc(MrB1[i]).

[0038] As shown in Figure 15, Method 100 according to the present invention can be applied to suppress the effects of B1 excitation magnetic field heterogeneity independently of (i.e., alone or subsequently to) correction for effects induced by B0 magnetic field heterogeneity of the imaging apparatus. Accordingly, such Method 100 can be applied to the raw data MBMK of the biomarker BMK (uncorrected if the biomarker BMK is not affected by B0 magnetic field heterogeneity), or to the corrected data MBMK' of the biomarker after performing an optional step 140' of correcting the raw data MBMK to suppress the effects of B0 static magnetic field heterogeneity of the imaging apparatus, the optional step 140' is shown in Figure 15 with a dashed line to highlight its optional nature.

[0039] Advantageously, the constant c is selected to be equal to Fc(1), i.e., Fc(MrB1=1). In a modified example, such a constant c may be calculated as arising from the mean or median value of the biomarker BMK in a region of organ unaffected or only slightly affected by the B1 excitation field heterogeneity (the focus of CEST acquisition). Any other technique may be performed to avoid biasing the value of the biomarker MBMK' corrected by the present invention. The present invention provides an advantageous embodiment for segmenting the correlation between the biomarker BMK and the B1 excitation field heterogeneity, and thus optimizing the estimation of the Fc function parameters in step 130. Such advantageous techniques will be detailed later in relation to Figures 11 and 12. According to one embodiment of the present invention, the execution of such a segmentation technique may be used, as described above, to determine one or more regions of organ unaffected or only slightly affected by the B1 excitation field heterogeneity, and to calculate the value of the constant c as the mean, median, etc.

[0040] Next, Method 100 according to the present invention includes step 140 of generating, for each voxel i, the corrected biomarker BMK' value MBMK'[i] for each of the N voxel sets such that MBMK'[i] = MBMK[i] - Fc(MrB1[i]) + c, i.e., MBMK'[i] = a * MrB1[i] + b in the example related to Figures 9 and 13. Figure 10 shows a scatter plot corresponding to the previously corrected scatter plot values, with the new scatter plot superimposed on the previous scatter plot before correction. Thus, the new horizontal line DCL symbolizes that the application of step 140 has caused the correlation between the biomarker BMK and the B1 excitation magnetic field to be lost.

[0041] The properties of the function Fc can be variable (linear, quadratic, etc.), but it must be a polynomial relating to the monotonic inverse transform of the CEST signal under consideration. As an unrestricted example, such a function Fc may consist of an inverse function, an exponential function, or a logarithmic function.

[0042] Figure 11 illustrates the fact that the B1 excitation magnetic field can affect various tissues of the organ in question, in this case, brain tissue. If a tissue mask is not available, step 130 of method 100 according to the present invention may include a substep 131 comprising segmenting an image showing a correlation between the biomarker BMK and the heterogeneity of the B1 magnetic field using maps MBMK and MrB1. A bivariate Gaussian mixture model (GMM) may be used to fit a scatter plot formed by placing the values ​​of the B1 excitation magnetic field on a first horizontal axis and the values ​​of the biomarker BMK on a second vertical axis, as shown in Figure 11. To improve the determination of the GMM model, it may be advantageous to fit the values ​​of the B1 magnetic field to a range of values ​​of the biomarker BMK, as shown in B1' in Figure 11. The number of Gaussian distributions to be fitted is fixed at 3 in the example shown in Figure 11 (Gaussian distributions GD1, GD2, and GD3), but this number may vary based on the data. Several algorithms, including an expectation maximization algorithm, may be used to fit the GMM model. Figure 12 illustrates the probability maps DPM1, DPM2, and DPM3 associated with the three Gaussian distributions GD1, GD2, and GD3, respectively, obtained from the data shown in relation to Figure 11. It is visually readily apparent that the Gaussian distribution GD2, which is more compact and uniform than the other two GD1 and GD3, roughly corresponds to the white matter of brain tissue, as illustrated by map DPM2. Accordingly, step 130 of method 100 according to the present invention may include a substep of selecting the smallest identified Gaussian distribution (i.e., the one with the smallest product of its eigenvalues), in this case the distribution GD2, which extends to a sufficient number of voxels in the organ under consideration, in Figures 11 and 12, in this case the human brain, to avoid cases where it is difficult to determine and the validity is poor. Such minimum voxel coverage by such a Gaussian distribution may be set to 30% of the voxels in the brain. However, such a minimum coverage threshold of 30% is not limiting to the present invention. Such thresholds could be lower than 20% or higher than the 30% mentioned.Therefore, the estimation of the parameters of the function Fc in step 130 can favorably be performed only for voxels characterized in this way by the determination of the Gaussian distribution GD2. As mentioned above, such a determination of the GMM can be used to identify the voxels of interest used to calculate the constant c, if the constant c arises from the calculation of the mean or median rather than c=Fc(1).

[0043] Notwithstanding step 140 of generating the corrected BMK' biomarker value MBMK'[i] for each of the set of N voxels for each voxel i such that MBMK'[i] = MBMK[i] - Fc(MrB1[i]) + c, method 100 according to the present invention may include step 150 of generating graphic content showing a parameter map MBMK' having voxels encoding the corrected biomarker BMK' values ​​for all or part of a set M of basic volumes of interest; and step 160 of displaying or more generally outputting the graphic content associated with the map MBMK' in the form of an image by the output human-machine interface of an imaging analysis system AS.

[0044] Through the example described in relation to Figure 15, it can be pointed out and emphasized that the implementation of Method 100 according to the present invention requires only one defined value of the B1 excitation magnetic field, unlike previous solutions which require numerous acquisitions for different values ​​of B1. Accordingly, the present invention significantly reduces the time required to correct the biomarker BMK, thereby suppressing the effect of B1 excitation magnetic field heterogeneity on the biomarker BMK.

[0045] Figures 13 and 14 illustrate the contribution of the present invention to suppress the effect of B1 excitation field heterogeneity after an initial correction to suppress the effect of B0 static field heterogeneity. Figure 13 shows two parameter maps MBMK' of a human brain biomarker (in this case, the "reference-label") and an intrinsic map MrB1 showing strong B1 excitation field heterogeneity of the imaging device in question. On the left is the first map MBMK' after a correction to suppress the effect of B0 field heterogeneity alone according to any known technique. On this first map, regions labeled MBMK-1 and MBMK-2, where the signal is weak or absent, are clearly identifiable, considering the effect of B1 field heterogeneity indicated by map MrB1 located on the right side of Figure 13. In the center of Figure 13 is the second parameter map MBMK' of the same BMK biomarker after a correction according to the present invention to suppress B1 excitation field heterogeneity. The contribution of the correction according to the present invention can be observed. Regions MBMK-1 and MBMK-2 on the left map MBMK' have disappeared from the center map MBMK'. In fact, no region within the central map MBMK' contains pixels associated with non-existent signals. On the contrary, the corrected map MBMK' according to the present invention further highlights region MBMK-3 of interest to practitioner 6, which may influence the diagnosis made by practitioner 6. Such region MBMK-3 would not have been noticed without the correction to suppress the effects of heterogeneity in the B1 field.

[0046] Figure 14 also illustrates the contribution of the present invention in a clinical setting. Such Figure 14 shows two parameter maps MBMK and MBMK' of the human brain biomarker BMK (in this case, the "reference-label"). On the left side, the first map MBMK without correction is illustrated, and on the first map MBMK, numerous artifacts can be clearly identified, considering the effect of the non-uniformity of the B1 excitation magnetic field, as shown by map MrB1 on the right side of Figure 14. In the center of Figure 14, the parameter map MBMK' of the same biomarker BMK after correction according to the present invention is shown. It is possible to observe the effect of the correction provided by the present invention. The represented MBMK' no longer shows artifacts, its pixels are more uniform, and the gradation of their values ​​is richer.

[0047] In addition to the extremely high accuracy and improved validity of the graphic representation MBMK' available to practitioners, the implementation of a method for post-processing a first set Z of experimental signals sample Z(Δω) resulting from the acquisition sequence by the medical magnetic resonance imaging apparatus 1 makes it fully available in the clinical field, thus opening the door to the said clinical field, particularly CEST technology, for quantifying one or more biomarkers associated with chemical species of interest in human or animal organs in order to ultimately characterize different lesions with altered metabolic properties. The present invention is not limited to this use. Therefore, it finds a major position in any use of magnetic resonance imaging where correction for B1 excitation field heterogeneity is required.

Claims

1. A method (100) for post-processing biomarkers (BMKs) generated from a plurality of M volumes Z(Δωj), wherein each of the M volumes Z(Δωj) contains N voxels representing the magnitude of experimental signals sampled with a frequency shift Δωj relative to the resonance frequency of water during an acquisition sequence by a magnetic resonance imaging apparatus (1), and the method is performed by a processing unit (4) of an imaging analysis system (AS). Step (110) of generating the biomarker (BMK) in the form of a map MBMK having N voxels, wherein each voxel i represents a signal value MBMK[i] where i is between 1 and N, Step (120) of generating a map MrB1 that shows the non-uniformity of the excitation magnetic field (B1) of the magnetic resonance imaging apparatus (1) for the M volumes, wherein the map MrB1 includes N voxels, Step (130) is to estimate the parameters of the function Fc such that MBMK[i] = Fc(MrB1[i]), A method (100) comprising the step (140) of generating a corrected biomarker (BMK) in the form of a map MBMK' of N voxels such that MBMK'[i] = MBMK[i] - Fc(MrB1[i]) + Fc(1).

2. The method according to claim 1 (100), wherein the step (130) of estimating the parameters of the function Fc such that MBMK[i] = Fc(MrB1[i]) includes performing least-squares minimization on all or part of the set of N voxels of the map MBMK and MrB1.

3. The method according to claim 1 or 2, comprising the step (150) of generating graphic content showing a parameter map MBMK', wherein N voxels therein each encode the value of the corrected biomarker.

4. The method according to claim 3, further comprising the step (160) of displaying the graphic content by an output human-machine interface (5) of the imaging analysis system (AS).

5. The method according to any one of claims 1 to 4, further comprising a step (140') of correcting the effect of B0 static magnetic field heterogeneity on the generated (110) biomarker (BMK) before the step (140) of generating the corrected biomarker (BMK).

6. The system includes a processing unit (4), means for communicating with the outside world, and storage means. The communication means is provided to receive experimental data generated from the acquisition sequence by the magnetic resonance imaging apparatus (1) from the outside world. The imaging analysis system (AS) is characterized in that the storage means includes program instructions, and the method (100) according to any one of claims 1 to 5 is executed by interpretation of the program instructions by the processing unit.

7. A computer program comprising one or more instructions interpretable by the processing unit (4) of the imaging analysis system (AS) according to claim 6, which can be stored in the storage means of the system, characterized in that the interpretation of the program instructions by the processing unit results in the execution of the method (100) according to any one of claims 1 to 5.

8. A computer-readable storage medium comprising the program instructions of the computer program described in claim 7.

Images