Use of images providing surrogated representations of reduced-amounts of contrast agent in an MRI system

Abstract

A solution is proposed for imaging a body-part of a patient in an MRI system (120,130). A corresponding method (500) comprises acquiring (510-548) a high-contrast image and one or more reduced-contrast images of the body-part comprising a contrast agent; the high-contrast image is acquired (510-512;520-522,532;534-536,544-548) with a high-contrast acquisition protocol and the reduced-contrast images are acquired (514-518;524-532;538-548) with corresponding reduced-contrast acquisition protocols providing corresponding total or partial flattenings between representations of locations of the body-part containing the contrast agent and without it. An output image of the body-part is generated (556;568) based on the high-contrast image and one or more reduced-contrast images, each being used as a surrogated representation of the body-part without the contrast agent or comprising a corresponding reduced-amount of the contrast agent. Corresponding computer program (400) and computer program product for implementing the method (500) are proposed. Moreover, an MRI system (120,130) for performing the method (500) is proposed. A corresponding medical method is further proposed. A method (500) for training a machine learning model with these images is also proposed.

Description

[0001] USE OF IMAGES PROVIDING SURROGATED REPRESENTATIONS OF REDUCED-AMOUNTS OF CONTRAST AGENT IN AN MRI SYSTEM DESCRIPTION Acknowledgement

[0002] This research was made under the National Recovery and Resilience Plan (NRRP), Mission 4, Component 2 Investment 1.3, Research funded by the European Commission under the NextGenerationEU program, Partenariato Esteso PNRR PEI 3 - "FAIR - Future Artificial Intelligence Research”.

[0003] Technical field

[0004] The present disclosure relates to the field of medical applications. More specifically, this disclosure relates to Magnetic-Resonance Imaging (MRI) techniques.

[0005] Background

[0006] The background of the present disclosure is introduced hereinafter with the discussion of techniques relating to its context. However, even when this discussion refers to documents, acts, artifacts and the like, it does not suggest or represent that the discussed techniques are part of the prior art or are common general knowledge in the field relevant to the present disclosure. Particularly, it is expressly understood that possible drawbacks mentioned herein should not be considered to have been previously recognized in the prior art.

[0007] MRI is a well-established imaging technique, which allows inspecting bodyparts of patients through images providing visual representations thereof (typically, in a substantially non-invasive manner even if the body-parts are not visible directly). In general terms, the MRI technique is based on the exposure to a magnetic field of the body-part of each patient to be imaged. As locations of the body-part react to the magnetic field in a different way according to their characteristics, by measuring a response signal of the body-part to the magnetic field it is possible to infer morphological and / or physiological information of the body-part that is used to create images thereof.

[0008] The images of the body-part may be acquired with different acquisition protocols. For example, a Tl-weighted acquisition protocol is used to acquire (Tl) images highlighting fat in the body-part, so as to provide information about its anatomy. Instead, a T2-weighted acquisition protocol is used to acquire (T2) images highlighting fluid in the body -part, so as to provide information about possible lesions within it. In order to reduce a corresponding acquisition time, US-A-2005 / 033151 proposes acquiring only a portion of a T1 image, only a portion of a T2 image and a remaining common portion that is shared between them; each T1 / T2 image is then reconstructed from the corresponding portion and the common portion.

[0009] Typically, a contrast agent is administered to the patient for providing a (contrast) enhancement of the representation of a corresponding target containing it, z.e., a structure with specific characteristics (such as a known lesion), so as to make the target more conspicuous in the corresponding (contrast) images. In this case, a (nocontrast) image of the body-part may also be acquired before administration of the contrast agent to the patient for providing a background representation of the bodypart without it.

[0010] The no-contrast image and each contrast image may be applied to a machine learning model that generates a corresponding (simulated) image simulating an increase of a dose of the contrast agent (either to reduce an amount of the contrast agent that is administered to the patient or to boost its contrast enhancement). For this purpose, the machine learning model may have been trained with multiple training sets, each comprising a zero-dose image acquired before administration of the contrast agent, a reduced-dose image acquired after administration of a reduced-dose of the contrast agent lower than a full-dose that is standard in clinical practice and a full-dose image acquired after a further administration of the contrast agent for obtaining its fulldose; the machine learning model is trained to optimize its capability to generate the full-dose image from the zero-dose image and the reduced-dose image of each training set. Alternatively, WO-A-2023 / 073165 proposes to omit the use of the zero-dose image; particularly, in this case a neural network is trained to optimize its capability to generate the full-dose image from two reduced-dose images of each training set, wherein the reduced-dose images have been acquired with different sequence protocols (for example, T1 -weighted, T2-weighted and so on). Lescher Stephanie et al: "Quantitative T1 and T2 mapping in recurrent glioblastomas under bevacizumab: earlier detection of tumor progression compared to conventional MRI", Neuroradiology, Springer Berlin Heidelberg, Berlin / Heidelberg, vol. 57, no. 1, 7 October 2014 (2014-10-07), pages 11-20, instead discloses a variable flip angle method, wherein two T1 images are acquired with different flip angles, and then a T1 map is generated with each value calculated by fitting the steady-state spoiled gradient echo signal equation to the signal intensities of the two T1 images.

[0011] However, during each imaging session a relatively long time (such as from some minutes to a few hours) elapses between the acquisition of the no-contrast image and the acquisition of the contrast images, during which time a corresponding patient should remain as far as possible motionless. In any case, it is very difficult (if not impossible) to have the (no-contrast / contrast) images perfectly aligned to each other, since they are inevitably subjected to misalignments caused by unavoidable movements / deformations of the body-part (for example, due to breathing, digestion or specific pathologies). Therefore, registration techniques are commonly used to reduce the misalignments of the images. The registration techniques are based on the application of transformations attempting to bring the images into spatial correspondence (z.e., referring them to a common reference space).

[0012] However, the registration techniques are not completely satisfactory in most practical situations. The residual misalignments of the images then adversely affect the quality of every application based on a combination of the no-contrast image and the contrast images. Particularly, in the above-mentioned example (simulated increase of the dose of the contrast agent), this generates artifacts in the simulated image that may distort or obscure anatomical structures adversely affecting sensitivity and specificity, with a risk of misinterpretations (especially of small lesions) impacting patient outcomes (such as in diagnostic applications). Likewise, this impairs the training of the machine learning model, with negative consequences on its robustness (and then on its capability of generating the simulated images).

[0013] Summary

[0014] The present invention is set out in the appended claims; particularly, one or more aspects of the present invention are defined in the independent claims and advantageous features thereof are defined in the dependent claims.

[0015] A simplified summary of the present disclosure is herein presented in order to provide a basic understanding thereof; however, the sole purpose of this summary is to introduce some concepts of the disclosure in a simplified form as a prelude to its following more detailed description, and it is not to be interpreted as an identification of its key elements nor as a delineation of its scope.

[0016] In general terms, the present disclosure is based on the idea of using images providing a surrogated representation of a reduced-amount of the contrast agent (down to none).

[0017] Particularly, an aspect provides a method for imaging a body -part of a patient in an MRI system. The method comprises acquiring a high-contrast image and one or more reduced-contrast images of the body-part comprising a contrast agent; the high- contrast image is acquired with a high-contrast acquisition protocol and the reduced- contrast images are acquired with corresponding reduced-contrast acquisition protocols providing corresponding total or partial flattenings between representations of locations of the body -part containing the contrast agent and without it. An output image of the body-part is generated based on the high-contrast image and the reduced- contrast images, each being used as a surrogated representation of the body-part without the contrast agent or comprising a corresponding reduced-amount of the contrast agent.

[0018] A further aspect provides a software program for implementing the method.

[0019] A further aspect provides a corresponding software program product.

[0020] A further aspect provides an MRI system for performing the method.

[0021] A further aspect provides a corresponding medical method.

[0022] A further aspect provides a method for training a machine learning model with these images.

[0023] Brief description of the drawings

[0024] The solution of the present disclosure, as well as further features and the respective advantages, will be better understood with reference to the following detailed description thereof, to be read in conjunction with the accompanying drawings (wherein, for the sake of simplicity, corresponding elements are denoted with equal or similar references and their explanation is not repeated, and the name of each entity is generally used to denote both its type and its attributes, like value, content and representation).

[0025] In this respect, it should be expressly understood that features described in each complete sentence may be implemented independently of the features described in the other sentences, except for those strictly necessary functionally. Moreover, whenever an example of implementation is provided for a general feature, for the sake of clarity, it is possible that reference is made subsequently to that example of implementation; in such a case, it should be expressly understood that what is described in relation to the example of implementation also applies to the general feature and to any other implementation thereof. In any case, any explanation provided in relation to an aspect of the present invention applies mutatis mutandis to any other aspect thereof.

[0026] Particularly:

[0027] FIG.1 shows a schematic block diagram of an MRI installation wherein the solution according to an embodiment of the present disclosure may be practiced,

[0028] FIG.2 shows qualitative time diagrams of exemplary imaging sessions according to the prior art,

[0029] FIG.3A-FIG.3B show qualitative time diagrams of exemplary imaging sessions relating to the solution according to different embodiments of the present disclosure,

[0030] FIG.4 shows the main software components that may be used to implement the solution according to an embodiment of the present disclosure,

[0031] FIG.5A-FIG-5B show an activity diagram describing the flow of activities relating to the solution according to an embodiment of the present disclosure,

[0032] FIG.6 shows representative examples of in-vitro experimental results relating to the solution according to an embodiment of the present disclosure, and

[0033] FIG.7A-FIG.7E show representative examples of in-vivo experimental results relating to the solution according to an embodiment of the present disclosure. Detailed description

[0034] With reference in particular to FIG. l, a schematic block diagram is shown of an MRI installation 100 wherein the solution according to an embodiment of the present disclosure may be practiced.

[0035] For example, the MRI installation 100 comprises a scanner room 105, an equipment room 110 and a control room 115. The scanner room 105 houses an MRI scanner 120, so as to shield Radio-Frequency (RF) radiations, magnetic fields and acoustic noises. The equipment room 110 houses equipment 125 supporting operation of the MRI scanner 120 (for example, providing power supply, exchanging electrical signals, circulating cooling liquids and so on). The control room 115 houses a control unit of the MRI scanner 120, for example, a control computer 130. The control room 115 is also provided with a door 135 for accessing the scanner room 105 and with a window 140 for visual inspection thereof.

[0036] For example, the MRI scanner 120 comprises the following components. A gantry 145 (for example, in the form of a hollow cylinder) is used to receive a patient undergoing an imaging session (not shown in the figure). The gantry 145 houses, not visible in the figure, a superconducting (or permanent) magnet (for generating a stationary magnetic field B0, for example, of the order of 0.25-7.00 T), multiple sets of gradient coils for different axes (coupled with the superconducting magnet for applying a magnetic gradient adjusting the stationary magnetic field B0) and an RF coil (with a specific structure, such as surface coil, saddle coil or Helmholts coil, for applying magnetic pulses Bl to a corresponding type of body-part to be imaged and for receiving corresponding response signals), all of them surrounded by an RF / magnetic shield. A table 150 is used for laying down the patient. The table 150 is mounted on a base 155, which is slidable horizontally in and out the gantry 145 (typically, by means of a motor not visible in the figure). A driver 160 comprises all the components required to drive the gradient coils and the RF coil (for example, an RF transmitter, an output amplifier and the like) and to acquire the response signals from the body-part (for example, an input amplifier, an Analog-To-Digital Converter (ADC) and the like).

[0037] For example, the control computer 130 comprises the following components connected among them through a bus structure (not shown in the figure). Particularly, a microprocessor (or more) provides a logic capability of the control computer 130. A non-volatile memory (ROM) stores basic code for a bootstrap of the control computer 130 and a volatile memory (RAM) is used as a working memory by the microprocessor. The control computer 130 is provided with a mass-memory for storing programs and data (for example, a Solid State Disk or SSD). Moreover, the control computer 130 comprises a number of controllers for peripherals units (for example, a keyboard, a mouse, a monitor, an adapter for connecting to the MRI scanner 120, a driver for reading / writing removable storage units (such as of USB type), a network adapted for connecting to a network (such as a LAN), and so on).

[0038] For example, the MRI scanner 120 is used in a standard MRI technique. In this case, during each imaging scan of a body -part of a patient the gradient coil controls the (total) magnetic field so as to obtain a specific resonance (or Larmor) frequency of water in a selected slice of the body-part to be imaged. The magnetic field magnetizes the water in the body -part, so that in an equilibrium condition the spins of the two protons of each water molecule align with the magnetic field (slightly more in a low-energy state along a same direction than in a high-energy state against the direction of the magnetic field). The RF coil then applies a magnetic pulse at this resonance frequency to the bodypart. As a consequence, the spins of the protons in the selected slice absorb energy that moves them from the low-energy state to the high-energy state. The spins then return to their equilibrium condition, thereby emitting energy. However, the spins return to their equilibrium condition through different relaxation processes depending on the characteristics of their locations in the body-part. Therefore, by measuring corresponding response signals (after a certain period following the application of the magnetic pulse) it is possible to obtain values providing information about the locations. Particularly, a T1 (longitudinal) relaxation time reflects a rate at which the spins return to be aligned with the magnetic field; instead, a T2 (transversal) relaxation time (either natural T2 or observed T2*) reflects a rate at which the spins decay transversally to the magnetic field.

[0039] With reference now to FIG.2, qualitative time diagrams are shown of exemplary imaging sessions according to the prior art.

[0040] In general, during each imaging session defined by a whole procedure for analyzing a body-part of a patient, one or more imaging scans are performed with an MRI system comprising an MRI scanner and a control unit (for example, comprised in the MRI installation described above). Each imaging scan involves the acquisition of one or more images of the body -part with a corresponding acquisition protocol. The acquisition protocol is defined by an acquisition method and corresponding values of one or more acquisition parameters of the MRI scanner. The acquisition method specifies a (general) technique that regulates the acquisition of the images by the MRI scanner. Particularly, the acquisition method is defined by the magnetic gradient, the magnetic pulses and their timing that are used to operate the MRI scanner. The acquisition protocol specifies a (particular) setting of the MRI scanner that is used to acquire the images with the acquisition method. Particularly, the acquisition protocol may be defined by the values of repetition time, flip angle, echo time, b-value and so on. For example, the acquisition protocol is T1 -weighted, based on a gradient echo acquisition method (Tl-weighted gradient echo acquisition protocol) or on a spin echo acquisition method (Tl-weighted spin echo acquisition protocol).

[0041] As far as relevant to the present disclosure, the imaging session is based on a contrast agent. The contrast agent provides a (contrast) enhancement of the representation in the images of a corresponding target containing it (such as a lesion), by influencing a magnetic response of the protons of the water molecules around it. In the example at issue (Tl-weighted acquisition protocol), the contrast agent is of paramagnetic type; in this case, the contrast agent significantly reduces the T1 relaxation time, thereby increasing an intensity of the corresponding response signals shortly after the application of the magnetic pulses (for example, after less than 1.5-2.0 times the T1 relaxation time). The contrast agent may be either of endogen type (z.e., produced inside the patient) or of exogen type (z.e., administered to the patient).

[0042] Particularly, a time diagram 200a relates to an exemplary imaging scan based on an (exogen) contrast agent. In this case, a (no-contrast) image 205 is acquired of the body-part without the contrast agent, or at least with no significant amount thereof (since no contrast agent has ever been administered to the patient or a relatively long time has elapsed from a previous administration of any contrast agent to the patient ensuring that it has been substantially cleared). Later on, the contrast agent is administered to the patient. After a corresponding technical time (for example, from some minutes to a few hours, being required to instruct the patient, to obtain an optimal contrast-enhancement of the target and so on), a (contrast) image 210 is acquired of the body-part comprising the contrast agent. The no-contrast image 205 and the contrast image 210 are registered to reduce their misalignment caused by unavoidable movements / deformations of the body-part during the relatively long time elapsed between their acquisitions. The (registered) no-contrast / contrast images 205,210 are used to generate an output image 215 based thereon (providing an increased contrastenhancement with respect to the contrast image 210, for example, by simulating an increase of the contrast agent by means of a simulation model), which output image is then output (for example, displayed on the monitor of the MRI system).

[0043] A time diagram 200b relates to another exemplary imaging scan again based on an (exogen) contrast agent. In this case as well, the no-contrast image 205 is acquired and the contrast agent is administered to the patient at a low-dose. After a corresponding technical time as above, a contrast image, differentiated as low-dose (contrast) image 2101, is acquired of the body -part comprising the low-dose of the contrast agent. Later on, a further amount of the contrast agent is administered to the patient so as to obtain a high-dose (higher than the low-dose) as a total. After a corresponding technical time as above, a further contrast image, differentiated as high-dose (contrast) image 210h, is acquired. The no-contrast image 205, the low-dose image 2101 and the high-dose image 21 Oh are registered to reduce their misalignment. In this case, in addition or in alternative to generate the corresponding output image as above (for example, by applying the no-contrast image 205 and the high-dose image 21 Oh to the simulation model), the no-contrast image 205, the low-dose image 2101 and the high-dose image 21 Oh may be used as a training set for training a machine learning model (not shown in the figure) implementing the simulation model.

[0044] Instead, in case of a contrast agent of endogen type (not shown in the figure), only a contrast image of the body-part comprising the contrast agent may be acquired (at any moment); conversely, since the contrast agent cannot be removed from the body-part, no image may be obtained of the body-part with a different amount of the contrast agent or without it.

[0045] With reference now to FIG.3A-FIG.3B, qualitative time diagrams are shown of exemplary imaging sessions relating to the solution according to different embodiments of the present disclosure.

[0046] In this case, the same MRI system is used to acquire a high-contrast image and one or more reduced-contrast images of the body -part during an imaging scan thereof.

[0047] The high-contrast image is acquired from the body-part comprising a high- amount of the contrast agent (z.e., after the technical time from its administration if the contrast agent is of exogen type or at any moment if the contrast agent is of endogen type). The high-contrast image is acquired with a high-contrast acquisition protocol. Particularly, the high-contrast acquisition protocol is optimal for providing the corresponding contrast-enhancement given by the effect of the contrast agent (for example, a standard T1 -weighted gradient / spin echo acquisition protocol).

[0048] The reduced-contrast images are acquired from the body-part comprising the same high-amount of the contrast agent. Particularly, if the contrast agent is endogen the body-part intrinsically comprises the same high-amount of the contrast agent (during the whole imaging session). Instead, if the contrast agent is exogen the bodypart substantially comprises the same high-amount of the contrast agent in a period (generally lasting for several minutes) between a wash-in phase (required by the contrast agent to perfuse the body -part) and a wash-out phase (when the contrast agent is filtered out by the patient), apart for fluctuations caused by the circulation of the contrast agent that are negligible in practice. This is especially true when the high- contrast image and the reduced-contrast images are acquired close to each other (for example, in a total acquisition time equal at most to 2.0 times, preferably 1.5 times and still more preferably 1.2 times, such as at most substantially equal to a sum of corresponding individual acquisition times that are required to acquire each of them). However, each reduced-contrast image is acquired with a corresponding reduced- contrast acquisition protocol having a contrast weighting that is lower with respect to the high-contrast acquisition protocol. Particularly, the reduced-contrast acquisition protocol is sub-optimal for providing the corresponding contrast-enhancement given by the effect of the contrast agent. The reduced-contrast acquisition protocol causes a flattening of an effect of the contrast agent with respect to the high-contrast acquisition protocol, given by a flattening between the response signals of the contrast agent (z.e., of locations of the body -part containing the contrast agent) and the response signals of a background (z.e., of locations of the body -part without the contrast agent); this results in a corresponding flattening between the representation of the contrast agent and the representation of the background (z.e., of their contrast) in the reduced-contrast images with respect to the high-contrast image. The flattening of the effect of the contrast agent may be either total or partial, depending on the reduction of the contrast weighting. This means that a difference between a descriptor of the representation of the contrast agent and a descriptor of the representation of the background summarizing them (for example, their averages) becomes lower (such as less than 90%, preferably 70% and still more preferably 50%), down to substantially none (such as less than 5%, preferably 3% and still more preferably 1%). For example, in the high- contrast image the contrast agent may be bright and the background may be dark; in the reduced-contrast image providing the total flattening of the effect of the contrast agent (referred to as totally flattened reduced-contrast image) both the representation of the contrast agent and the representation of the background may become very dark (similar to each other), whereas in the reduced-contrast images providing corresponding partial flattenings of the effect of the contrast agent (referred to as partially flattened reduced-contrast images) the representation of the contrast agent may become less bright and the representation of the background may remain substantially the same or become less dark. Particularly, in the case of the T1 -weighted (gradient / spin echo) acquisition protocol the reduced-contrast acquisition protocols are less Tl-weighted than the high-contrast acquisition protocol.

[0049] The reduced-contrast images are used as surrogated images, comprising a (surrogated) no-contrast image (surrogating a representation of the body -part without the contrast agent) and / or one or more (surrogated) low-contrast images (surrogating a representation of the body-part comprising corresponding low-amounts of the contrast agent lower than the high-amount). The high-contrast image and the surrogated images (z.e., the corresponding reduced-contrast images) are used to generate the same output image based thereon that is then output (and possibly to train the same machine learning model) as above. Particularly, the totally flattened reduced- contrast image is used as surrogate of an (actual) no-contrast image that would have been acquired with the high-contrast acquisition protocol from the body-part without the contrast agent. In fact, the totally flattened reduced-contrast image excludes a contribution of the contrast agent, thereby providing a good approximation of the actual no-contrast image. Likewise, each partially flattened reduced-contrast image is used as surrogate of an (actual) low-contrast image that would have been acquired with the high-contrast acquisition protocol from the body-part comprising a corresponding low-amount of the contrast agent. In fact, the partially flattened reduced-contrast image limits a contribution of the contrast agent, thereby providing a good approximation of the actual low-contrast image. In both cases, the (totally / partially flattened) reduced-contrast image may substitute the corresponding actual (no- contrast / low-contrast) image with an acceptable degree of accuracy in many applications, despite its slightly reduced capability of making details stand out (due to the flattening of its contrast in general) that is however acceptable in many practical situations (in any case, with this reduced capability that is far outweighed by the following benefits). In other words, the output image is generated by applying a processing to be based on the high-contrast image (providing the representation of the body-part comprising the high-amount of the contrast agent) and a no-contrast image (providing the representation of the body-part without the contrast agent) and / or one or more low-contrast images (providing the representation of the body -part comprising corresponding low-amounts of the contrast agent lower than the high-amount). However, the processing is applied using the high-contrast image and corresponding surrogated images defined by the reduced-contrast images; particularly, a totally flattened reduced-contrast image is used as surrogate of the (actual) no-contrast image and one or more partially flattened reduced-contrast images are used as surrogate of corresponding (actual) low-contrast images.

[0050] The above-mentioned solution provides a number of benefits. Particularly, the (high-contrast / reduced-contrast) images are acquired from the body-part in the same condition (z.e., comprising the same high-amount of the contrast agent); therefore, the images may be (and preferably are) acquired close to each other. This reduces a time during which the patient is required to remain motionless (for example, only a few seconds), thereby making the imaging scan more comfortable.

[0051] At the same time, movements / deformations of the body-part may be avoided or at least substantially reduced (for example, by asking the patient not to breathe during the acquisition of the images). As a consequence, the images are automatically aligned to each other (without requiring any registration), or at most they have limited misalignments (far easier to remove with their registration).

[0052] All of the above significantly increases the quality of every application based on these images. Particularly, the good alignment of the high-contrast / reduced-contrast images has a beneficial effect on the output image based thereon, for example, reducing artifacts that may distort or obscure anatomical structures, thereby increasing sensitivity and specificity; this limits a risk of misinterpretations (especially of small lesions) that might adversely impact patient outcomes (such as in diagnostic applications). In addition or in alternative, this allows collecting a relatively high number of training sets (with good alignment of their images) for training the machine learning model (since the corresponding images may be acquired in standard imaging sessions without requiring multiple administrations of the contrast agent); as a consequence, the quality of the training of the machine learning model is significantly incremented, with a positive impact on its robustness (and then on its capability of generating the simulated images).

[0053] Particularly, when the contrast agent is of endogen type, the reduced-contrast images are used as surrogate of (virtual) no-contrast / low-contrast images that would not be attainable in practice (since the contrast agent cannot be removed from the bodypart). This allows any use of the no-contrast / low-contrast images in this case as well.

[0054] Preferably, the images are acquired in direct succession. This means that the images are acquired as far as possible close to each other, z.e., with no intentional delays added between their acquisitions, apart from the ones required for switching among the different (high-contrast / reduced-contrast) acquisition protocols. This result may be achieved in different ways. Particularly, time diagrams 300a, 300b and 300c relate to exemplary imaging scans wherein the high-contrast image, denoted with the reference 305, and a single reduced-contrast image, denoted with the reference 310, are acquired. In all cases, the high-contrast image 305 and the reduced-contrast-image 310 (as surrogate of a no- contrast / low-contrast image) are used, generally without requiring any registration thereof, to generate the output image based thereon, denoted with the reference 315.

[0055] More specifically, the time diagram 300a relates to the acquisition of the (high- contrast / reduced-contrast) images 305,310 in a succession mode, wherein the whole images 305,310 are acquired in direct succession, z.e., with the acquisition of one of the images 305,310 that starts as soon the acquisition of the other one of them ends. For example, the figure shows the acquisition in succession of the reduced-contrast image 310 and the high-contrast image 305 (with similar considerations that apply to the acquisition of the images 305,310 in different order). In this way, the images 305,310 are acquired in a total acquisition time that is substantially equal to the sum of their individual acquisition times, apart from the small delay caused by the switching between the different (high-contrast / reduced-contrast) acquisition protocols, for example, equal to 1.0-1.1 times it.

[0056] The time diagram 300b relates to the acquisition of the images 305,310 in an interleaved mode, wherein a plurality of (image) portions of the images 305,310 are acquired in direct succession, interleaved to each other, z.e., again with the acquisition of each image portion that starts as soon the acquisition of another image portion ends. Particularly, a plurality of high-contrast portions of the high-contrast image 305 and a plurality of reduced-contrast portions of the reduced-contrast image 310 are acquired (for example, as corresponding segments in a k-space). For example, the figure shows the acquisition in succession of a first reduced-contrast portion Rl, a first high-contrast portion Hl, a second reduced-contrast portion R2 and a second high-contrast portion H2 (with similar considerations that apply to the acquisition of the image portions in different number and / or order). The high-contrast portions H1-H2 are combined into the high-contrast image 305 and the reduced-contrast portions R1-R2 are combined into the reduced-contrast image 310. In this way, corresponding image portions of the images 305,310 may be acquired closer to each other, with a negligible increase of the total acquisition time due to the higher number of switchings between the different acquisition protocols (for example, equal to 1.0-1.2 times the sum of the individual acquisition times of the images 305,310).

[0057] The time diagram 300c relates to the acquisition of the images 305,310 in a shared mode, wherein the image portions (acquired in direct succession, interleaved to each other) are in part shared between the high-contrast image 305 and the reduced- contrast image 310. Particularly, one or more high-contrast portions of the high- contrast image 305, one or more reduced-contrast portions of the reduced-contrast image 310 and one or more common portions are acquired in direct succession, interleaved to each other (for example, again as corresponding segments in the k- space). The common portions are of a common image of the body-part comprising the same high-amount of the contrast agent (for example, corresponding to a rest of the images 305,310 different from their high-contrast / reduced-contrast portions); the common portions are acquired with a common acquisition protocol, which is based on the high-contrast acquisition protocol and / or the reduced-contrast acquisition protocols (for example, equal to one of them or changing gradually from one to another of them). For example, the figure shows the acquisition in succession of a single reduced-contrast portion Rl, a single high-contrast portion Hl and a single common portion Cl (with similar considerations that apply to the acquisition of the image portions in different number and / or order). The high-contrast image 305 is generated from the high-contrast portion Hl and the common portion Cl, and the reduced- contrast image 310 is generated from the reduced-contrast portion Rl and the common portion Cl. In this way, the images 305,310 are acquired in a total acquisition time that is shorter than the sum of their individual acquisition times (for example, equal to 0.2-0.8 times it), with a negligible degrade of their quality due to the missing complete acquisitions thereof.

[0058] Time diagrams 300d, 300e and 300f instead relate to exemplary imaging scans wherein the high-contrast image 305 and two reduced-contrast images, z.e., a totally flattened reduced-contrast image 3 lOt and a partially flattened reduced-contrast image 310p, are acquired. In all cases, the high-contrast image 305, the totally flattened reduced-contrast image 3 lOt (as surrogate of the no-contrast image) and the partially flattened reduced-contrast image 31 Op (as surrogate of a corresponding low-contrast image) are then used, generally without requiring any registration thereof, to train the machine learning model as above (in addition or in alternative to use the high-contrast image 305 and the totally flattened reduced-contrast image 3 lOt to generate the output image as above, not shown in the figure).

[0059] More specifically, the time diagram 300d relates to the acquisition of the (high- contrast / reduced-contrast) images 305,310t,310p in the succession mode. For example, the figure shows the acquisition in succession of the totally flattened reduced-contrast image 3 lOt, the partially flattened reduced-contrast image 3 lOp and the high-contrast image 305 (with similar considerations that apply to the acquisition of the images 305,310t,310p in different order).

[0060] The time diagram 300e relates to the acquisition of the images 305,310t,3 lOp in the interleaved mode. Particularly, a plurality of high-contrast portions of the high- contrast image 305, a plurality of totally flattened reduced-contrast portions of the totally flattened reduced-contrast image 3 lOt and a plurality of partially flattened reduced-contrast portions of the partially flattened reduced-contrast image 31 Op are acquired in direct succession, interleaved to each other. For example, the figure shows the acquisition in succession of a first totally flattened reduced-contrast portion Rtl, a first partially flattened reduced-contrast portion Rpl, a first high-contrast portion Hl, a second totally flattened reduced-contrast portion Rt2, a second partially flattened reduced-contrast portion Rp2 and a second high-contrast portion H2 (with similar considerations that apply to the acquisition of the image portions in different number and / or order). The high-contrast portions H1-H2 are combined into the high-contrast image 305, the totally flattened reduced-contrast portions Rtl-Rt2 are combined into the totally flattened reduced-contrast image 3 lOt and the partially flattened reduced- contrast portions Rpl-Rp2 are combined into the partially flattened reduced-contrast image 31 Op.

[0061] The time diagram 300f relates to the acquisition of the images 305,310t,3 lOp in the shared mode. Particularly, one or more high-contrast portions of the high-contrast image 305, one or more totally flattened reduced-contrast portions of the totally flattened reduced-contrast image 3 lOt, one or more partially flattened reduced-contrast portions of the partially flattened reduced-contrast image 31 Op, and one or more common portions as above are acquired in direct succession, interleaved to each other. For example, the figure shows the acquisition in succession of a single totally flattened reduced-contrast portion Rtl, a single partially flattened reduced-contrast portion Rpl, a single high-contrast portion Hl and a single common portion Cl (with similar considerations that apply to the acquisition of the image portions in different number and / or order). The high-contrast image 305 is generated from the high-contrast portion Hl and the common portion Cl, the totally flattened reduced-contrast image 3 lOt is generated from the totally flattened reduced-contrast portion Rtl and the common portion Cl, and the partially flattened reduced-contrast image 31 Op is generated from the partially flattened reduced-contrast portion Rtl and the common portion Cl.

[0062] With reference now to FIG.4, the main software components are shown that may be used to implement the solution according to an embodiment of the present disclosure.

[0063] Particularly, all the software components (programs and data) are denoted as a whole with the reference 400. The software components are typically stored in the mass memory and loaded (at least partially) into the working memory of the control unit of the MRI system when the programs are running, in addition to an operating system and to other application programs not directly relevant to the solution of the present disclosure (thus omitted in the figure for the sake of simplicity). The programs are initially installed into the mass memory, for example, from one or more removable storage units or from the network. In this respect, each program may be a module, segment or portion of code, which comprises one or more executable instructions for implementing the specified logical function.

[0064] An acquirer 405 acquires the (high-contrast / reduced-contrast) images of a body -part of a patient during each imaging scan thereof (by driving the corresponding components of the MRI scanner accordingly). For example, the acquirer 405 acquires the response signals from the body-part as raw data in the k-space (to which reference will be made in the following, with the same considerations that apply to the acquisition of the images in any other form). Each point in the k-space contains information about a spatial frequency, or wavenumber k (repetition rate of same values) and phase (starting values) of the response signals of the whole body-part. In this case, the acquirer 405 controls a recorder 410 to record a selected segment of the k-space of each image (for example, a row for all the spatial frequencies of a selected phase).

[0065] When the images are acquired in the succession mode, the acquirer 405 writes a k-space high-contrast image repository 415 and a k-space reduced-contrast image repository 420. The k-space high-contrast image repository 415 and the k-space reduced-contrast image repository 420 store one or more k-space high-contrast images and one or more k-space reduced-contrast images, respectively, that have been acquired during the imaging scan. Particularly, in case of k-space (high- contrast / reduced-contrast) images in 2-dimensions, or 2D (to which reference will be made in the following, with the same considerations that apply of any other dimension), each k-space image is defined by a matrix of cells, with a horizontal axis corresponding to the spatial frequency and a vertical axis corresponding to the phase of the response signals; each cell contains a complex number defining the intensity of the corresponding response signal (whereas in case of images in 3-dimensions, or 3D, an additional axis would be added corresponding to a further phase in the additional dimension).

[0066] When the images are acquired in the interleaved mode, the acquirer 405 writes a high-contrast portions repository 425 and a reduced-contrast portions repository 430. The high-contrast portions repository 425 and the reduced-contrast portions repository 430 store a plurality of high-contrast portions of each high-contrast image and a plurality of reduced-contrast portions of each reduced-contrast image, respectively, that have been acquired during the imaging scan. For example, in the k-space each (high-contrast / reduced-contrast) portion is defined by one or more rows of the corresponding k-space image (for example, 8-64 rows for corresponding phases). A combiner 435 combines the corresponding high-contrast portions into each high- contrast image and the corresponding reduced-contrast portions into each reduced- contrast image. For this purpose, the combiner 435 reads the high-contrast portions repository 425 and the reduced-contrast portions repository 430, and it writes the k- space high-contrast image repository 415 and the k-space reduced-contrast image repository 420. When the images are acquired in the shared mode, the acquirer 405 writes the high-contrast portion repository 425, the reduced-contrast portion repository 430 and a common portion repository 440. In this case, the high-contrast portions repository 425 and the reduced-contrast portions repository 430 store one or more high-contrast portions of each high-contrast image and one or more reduced-contrast portions of each reduced-contrast image, respectively, that have been acquired during the imaging scan; moreover, the common portion repository 440 stores one or more common portions of the common image of the body-part (for example, covering a rest of the high-contrast / reduced-contrast images not covered by their high-contrast / reduced- contrast portions). For example, in the k-space each common portion is again defined by one or more rows (for example, 8-64 rows for corresponding phases). A generator 445 generates each k-space high-contrast image from the corresponding high-contrast portions and the common portions, and it generates each k-space reduced-contrast image from the corresponding reduced-contrast portions and the common portions. For this purpose, the generator 445 reads the high-contrast portions repository 425, the reduced-contrast portions repository 430 and the common portions repository 440, and it writes the k-space high-contrast image repository 415 and the k-space reduced- contrast image repository 420.

[0067] In the example at issue (images acquired in the k-space), a converter 450 converts the k-space images into a magnitude form providing corresponding visual representations of the body-part (otherwise obtained in a different way or directly in alternative implementations). For this purpose, the converter 450 reads the k-space high-contrast image repository 415 and the k-space reduced-contrast image repository 420, and it writes a magnitude high-contrast image repository 455 and a magnitude reduced-contrast image repository 460. The magnitude high-contrast image repository 455 and the magnitude reduced-contrast image repository 460 store one or more magnitude high-contrast images and one or more magnitude reduced-contrast images, respectively, corresponding to the k-space high-contrast images and to the k-space reduced-contrast images, respectively. Each magnitude (high-contrast / reduced- contrast) image is defined by a matrix of cells (for example, with 512 rows and 512 columns) each containing a value of a basic picture element representative of a corresponding location of the body-part; particularly, in this case of 2D magnitude images, the basic picture element is a pixel that represents a corresponding basic area of the body-part (whereas in case of 3D magnitude images the basic picture element would be a voxel that represents a corresponding basic volume of the body -part); each pixel value defines a brightness of the pixel (for example, in gray-scale) as a function of the intensity of the response signal of the corresponding location.

[0068] Optionally, a normalizer 465 normalizes the representation of the background in the magnitude images. For example, the normalizer 465 updates the magnitude reduced-contrast images (to which reference will be made in the following with the same considerations that apply to any altemative / additional updates of the magnitude high-contrast images). For this purpose, the normalizer 465 reads the magnitude high- contrast image repository 455 and it reads / writes the magnitude reduced-contrast image repository 460.

[0069] A consumer 470 consumes the (magnitude) high-contrast images and reduced- contrast images, possibly normalized (to which reference will be made in the following with the same considerations that apply to the use of the high-contrast / reduced-contrast images in any other form). For this purpose, the consumer 470 reads the magnitude high-contrast image repository 455 and the magnitude reduced-contrast image repository 460. For example, the consumer 470 comprises a simulator (for example, implemented by a machine leaning model), which simulates an increase of the contrast agent (or an increase of a delay from administration of the contrast agent to the patient and / or an increase of an efficiency of the contrast agent in the medical imaging application); in addition or in alternative, the consumer 470 comprises a highlighter, which highlights the representation of the contrast agent. In any case, the consumer 470 also comprises an output module, which outputs a resulting output image (for example, by driving the monitor of the MRI system). Optionally, the consumer 470 also comprises a logger (which logs training sets resulting from corresponding imaging scans, each comprising a high-contrast image and corresponding reduced-contrast images, to be used as surrogate of a no-contrast image and possibly a low-contrast image, for training the machine learning model) and an exporter (which exports the training sets to a training computing system, not shown in the figure, wherein the machine learning model is trained). For example, the training (computing) system is a training computer with a similar structure as the control computer of the abovedescribed MRI system, suitably scaled according to its function.

[0070] With reference now to FIG.5A-FIG.5B, an activity diagram is shown describing the flow of activities relating to the solution according to an embodiment of the present disclosure.

[0071] In this respect, each block may correspond to one or more executable instructions for implementing the specified logical function on the corresponding computing machine (control unit of the MRI system and training system). Particularly, the activity diagram represents an exemplary process for imaging a body-part of a patient with a method 500.

[0072] Starting from the swim-lane of the control unit of the MRI system, the process begins at the black start circle 502 as soon as a (new) imaging session is started after the patient has reached a proper position within the MRI scanner, as indicated by a corresponding start command, for example, entered by a (healthcare) operator, like a radiographer, via a user interface of the acquirer. In response thereto, the flow of activity branches at block 504 according to a configuration of the MRI system (for example, set in advance, selected dynamically with a possible default value, or the only one available). If the imaging session is based on a contrast agent of exogen type, at block 506 a message is output (for example, on the monitor of the control unit) prompting the operator to administer the contrast agent to the patient (operation that is not necessary when the contrast agent has already been administered to the patient in advance). For example, the contrast agent is administered at a full-dose (standard in clinical practice) or at a reduced-dose (lower than the full-dose). The process then continues to block 508; the same point is reached directly from block 504 when the imaging session is based on a contrast agent of endogen type.

[0073] A corresponding imaging scan of the body-part is then started for acquiring (high-contrast / reduced-contrast) images of the body -part. Particularly, in the following reference is made to a single k-space high-contrast image and one or more k-space reduced-contrast images (with the same considerations that apply to any other number and type of images). At this point, the flow of activity again branches according to the configuration of the MRI system. Particularly, when the MRI system is configured to acquire the k-space images in the succession mode blocks 510-518 are executed, when the MRI system is configured to acquire the k-space images in the interleaved mode blocks 520-532 are executed, and when the MRI system is configured to acquire the k-space images in the shared mode blocks 534-548 are executed; in all cases, the flow of activity then merges again at block 550.

[0074] Considering now block 510 (succession mode), the MRI scanner is configured according to the high-contrast acquisition protocol, as defined by its acquisition method and high-contrast values of one or more acquisition parameters of the MRI scanner. For example, the acquisition method is gradient echo (using magnetic gradients to dephase and rephase the protons) or spin echo (using 180° magnetic pulses to dephase and rephase the protons), and the acquisition protocol is T1 -weighted. In general, the Tl- weighted acquisition protocol is defined by a low Time to Echo TE (between magnetic pulses and response signals) and a low Repetition Time TR (between successive magnetic pulses); moreover, the T1 -weighted acquisition protocol is further defined by a low / intermediate Flip Angle FA (at which the protons are rotated away from the magnetic field by the magnetic pulses) in case of the gradient echo acquisition method (whereas the flip angle is generally 90° in the case of the spin echo acquisition method). Particularly, the T1 -weighted gradient echo acquisition protocol may have a time to echo TE<10ms, a repetition time TR<50ms and a flip angle FA<40°, whereas the Tl- weighted spin echo acquisition protocol may have a time to echo TE<50ms and a repetition time TR<1, 000ms. In this case, the high-contrast values of the acquisition parameters are a low repetition time and, for the gradient echo acquisition method, a low / intermediate flip angle. For example, in the case of the Tl-weighted gradient echo acquisition protocol the repetition time may be TR=10-50ms, like TR=20ms, and the flip angle may be FA=25-35°, like FA=30°, whereas in the case of the Tl-weighted spin echo acquisition protocol the repetition time may be TR=300- 1.000ms, like TR=500ms. The MRI scanner is controlled at block 512 to acquire the k-space high- contrast image and to save it into the corresponding repository. As soon as the acquisition of the k-space high-contrast image has been completed, the MRI scanner is configured at block 514 according to the reduced-contrast acquisition protocol of a (current) k-space reduced-contrast image. For example, the reduced-contrast acquisition protocol is defined by the same acquisition method and by reduced-contrast values of the acquisition parameters of the MRI scanner (at least in part different from the high-contrast values). In this case, the acquisition method is again gradient echo or spin echo, with a lower Tl- weighting with respect to the high-contrast acquisition protocol. For example, in the case of the T1 -weighted gradient echo acquisition protocol this result may be achieved by reducing the repetition time (so that the protons have less time to relax longitudinally) and / or by reducing the flip angle (so that the protons move less from the low-energy state); in both cases, the response signal of the contrast agent and of the background have less possibility to differentiate according to their T1 relaxation times and then became more comparable. Similarly, in the case of the T1 -weighted spin echo acquisition protocol this result may be achieved by increasing the repetition time (so that the protons have more time to fully relax longitudinally); in this case, the response signal of the contrast agent and of the background both have more possibility to relax independently of their T1 relaxation times and then became more comparable. The reduced-contrast values of the acquisition parameters are selected according to the desired total / partial flattening of the effect of the contrast agent (defining a corresponding reduction of the contrast weighting of the reduced-contrast acquisition protocol). For example, in order to obtain a total flattening of the effect of the contrast agent in the Tl- weighted gradient echo acquisition protocol, a very low value of the repetition time and a very low value of the flip angle are used; particularly, the reduced-contrast value of the repetition time is 20-60%, preferably 30-50% and still more preferably 35-45%, such as 40% of the corresponding high-contrast value (for example, 1-10ms, like 8ms), and the reduced-contrast value of the flip angle is 1-20%, preferably 1-10% and still more preferably 1-5%, such as 3% of the corresponding high-contrast value (for example, 0.1-5°, like 1°). Instead, in order to obtain a total flattening of the effect of the contrast agent in the T1 -weighted spin echo acquisition protocol, a very high value of the repetition time is used; particularly, the reduced-contrast value of the repetition time is 5-15 times, preferably 7-13 times and still more preferably 9-11 times, such as 10 times the corresponding high-contrast value (for example, 3, 000-10, 000ms, like 5,000ms). On the other hand, in order to obtain a partial flattening of the effect of the contrast agent in the T1 -weighted gradient echo acquisition protocol, a low repetition time and a low flip angle are used (in any case higher than the ones required for obtaining the total flattening of the effect of the contrast agent); particularly, the reduced-contrast value of the repetition time is 80-100%, preferably 90-100% and still more preferably 95-100%, such as equal to the corresponding high-contrast value (for example, 10-50ms, like 20ms) and the reduced-contrast value of the flip angle is 10- 50%, preferably 25-40% and still more preferably 30-35%, such as 33% of the corresponding high-contrast value (for example, 5-20°, like 10°). Instead, in order to obtain a partial flattening of the effect of the contrast agent in the T1 -weighted spin echo acquisition protocol, a high value of the repetition time is used (in any case lower than the one required for obtaining the total flattening of the effect of the contrast agent); particularly, the reduced-contrast value of the repetition time is 2-4 times, preferably 2.5-3.5 times and still more preferably 2.8-3.2 times, such as 3 times the corresponding high-contrast value (for example, 900-3, 000ms, like 1,500ms). The MRI scanner is controlled at block 516 to acquire the k-space reduced-contrast image and to save it into the corresponding repository. Continuing to block 518, if a further k-space reduced-contrast image is to be acquired (with a different flattening of the effect of the contrast agent), the process returns to block 514 to repeat the same operations with a corresponding (different) reduced-contrast acquisition protocol of the MRI scanner. Conversely, once all the k-space reduced-contrast images have been acquired, the flow of activity descends into block 550. Similar considerations apply if the k-space images are acquired in a different order (not shown in the figure).

[0075] For example, the MRI scanner acquires each k-space image row-by-row in direct succession, such as in a zig-zag way for phases of alternated sign increasing in absolute value from zero to a maximum value thereof (M, such as 128-512), like: 0, +1, -1, +2, -2 +M-2, -M+2, +M-1, -M+l, +M, -M.

[0076] Considering instead block 520 (interleaved mode), a loop is entered for acquiring the image portions in direct succession, interleaved to each other. For this purpose, the MRI scanner is configured according to the high-contrast acquisition protocol as above. The MRI scanner is controlled at block 522 to acquire a (current) high-contrast portion and to save it into the corresponding repository. As soon as the acquisition of the high- contrast portion has been completed, the MRI scanner is configured at block 524 according to the reduced-contrast acquisition protocol of a (current) k-space reduced- contrast image as above (according to the desired flattening of the effect of the contrast agent). The MRI scanner is controlled at block 526 to acquire a (current) reduced-contrast portion of the k-space reduced-contrast image and to save it into the corresponding repository (for example, corresponding to the high-contrast portion just acquired). The process continues to block 528 as soon as the acquisition of the reduced-contrast portion has been completed. If a further k-space reduced-contrast image is to be acquired (with a different flattening of the effect of the contrast agent), the process returns to block 524 to repeat the same operations with a corresponding (different) reduced-contrast acquisition protocol of the MRI scanner. Conversely, once all the k-space reduced- contrast images have been taken into account, the flow of activity descends into block 530. If a further set of (corresponding) image portions is still to be acquired, the process returns to block 520 to repeat the same operations for it. Conversely, once all the image portions have been acquired the process continues to block 532, wherein the corresponding high-contrast portions are combined into the k-space high-contrast image and the corresponding reduced-contrast portions are combined into each k-space reduced- contrast image (from / to the corresponding repositories). The flow of activity then descends into block 550. Similar considerations apply if the image portions are acquired in a different order (not shown in the figure).

[0077] For example, each image portion is formed by one or more rows of the corresponding k-space (such as 8-128 rows). The MRI scanner acquires each image portion row-by-row in direct succession, such as in a zig-zag way for phases of alternated sign between corresponding lower limit and upper limit in absolute value. Preferably, at least one (following) image portion following a first image portion (along their acquisition succession) is acquired for phases decreasing in absolute value from its upper limit to its lower limit. For example, in case of only two high-contrast portions (H1,H2) and two reduced-contrast portions (R1,R2), they may be acquired in the following order:

[0078] Rl: 0, +1, -1, +2, -2 +L-2, -L+2, +L-1, -L+l, +L, -L,

[0079] Hl: -L, +L, -L+l, +L-1, -L+2, +L-2 ... -2, +2, -1, +1, 0,

[0080] R2: +M, -M, +M-1, -M+l, +M-2, -M+2, ... +L-2, -L+2, +L-1, -L+l, +L, -L,

[0081] H2: +M, -M, +M-1, -M+l, +M-2, -M+2, ... +L-2, -L+2, +L-1, -L+l, +L, -L, (with L<M, such as L=M / 2). In fact, in the k-space most contrast information is encoded at its center (around the zero value of the phase), whereas its periphery (towards the maximum in absolute value of the phase) mainly encodes spatial resolution information only. Therefore, the acquisition of the rows of the k-space of each following image portion for decreasing phases (in absolute value) ensures that the lower the phases (in absolute value) of the k-space rows, the more far away their acquisition from the one of a preceding image portion (along their acquisition succession); in this way, the k-space rows for lower phases (providing more contrast information) are less adversely affected by the different (high-contrast / reduced-contrast) acquisition protocol of the preceding image portion.

[0082] Considering instead block 534 (shared mode), a loop is entered for acquiring the image portions now partially shared in direct succession, interleaved to each other (in the following, for the sake of simplicity reference is made to a single high- contrast / reduced-contrast portion of each k-space high-contrast / reduced-contrast image and a single common portion, with the same considerations that apply to any other number thereof). For this purpose, the MRI scanner is configured according to the high- contrast acquisition protocol as above. The MRI scanner is controlled at block 536 to acquire the high-contrast portion and to save it into the corresponding repository. As soon as the acquisition of the high-contrast portion has been completed, the MRI scanner is configured at block 538 according to the reduced-contrast acquisition protocol of a (current) k-space reduced-contrast image as above (according to the desired flattening of the effect of the contrast agent). The MRI scanner is controlled at block 540 to acquire the reduced-contrast portion of the k-space reduced-contrast image and to save it into the corresponding repository. The process continues to block 542 as soon as the acquisition of the reduced-contrast portion has been completed. If a further k-space reduced- contrast image is to be acquired (with a different flattening of the effect of the contrast agent), the process returns to block 538 to repeat the same operations with a corresponding (different) reduced-contrast acquisition protocol of the MRI scanner. Conversely, once all the k-space reduced-contrast images have been taken into account, the flow of activity descends into block 544. At this point, the MRI scanner is configured according to the common acquisition protocol (based on the high-contrast acquisition protocol and / or the reduced-contrast acquisition protocols). For example, the common acquisition protocol is set to the reduced-contrast acquisition protocol providing the total flattening of the effect of the contrast agent. The MRI scanner is controlled at block 546 to acquire the common portion and to save it into the corresponding repository. Alternatively (not shown in the figure), the common acquisition protocol changes continually, for example, from the reduced-control acquisition protocol providing the total flattening of the effect of the contrast agent to the high-contrast acquisition protocol (such as with a sigmoidal law). In this case, a loop is performed for acquiring the common portion, wherein at each iteration of the loop the configuration of the MRI scanner is changed and then the MRI scanner is controlled to acquire a row (or more) of the common portion. If further image portions are still to be acquired (not shown in the figure), the same operations described above are repeated for each of them. With reference now to block 548, the high-contrast image is generated from the high-contrast portion and the common portion and each reduced-contrast image is generated from the corresponding reduced-contrast portion and the common portion (from / to the corresponding repositories). Particularly, for this purpose it is possible to exploit Artificial Intelligence (Al)-driven approaches (for example, based on deep learning techniques) or more traditional methods (for example, based on compressed sensing), for example, as described in US-A-2005 / 033151. The flow of activity then descends into block 550. Similar considerations apply if the image portions are acquired in a different order (not shown in the figure).

[0083] Preferably, the high-contrast / reduced-contrast portions relate to a core of the k- space (extending at its center), whereas the common portion relates to a residual of the k-space (extending at its periphery); therefore, even if a segment of the k-space high- contrast / reduced-contrast images corresponding to the common portion is not acquired, this is less critical since the (missing) segment mainly encodes spatial resolution information of the corresponding magnitude high-contrast / reduced-contrast images.

[0084] For example, as above each image portion is formed by one or more rows of the corresponding k-space (such as 8-128 rows). The MRI scanner acquires each image portion row-by-row in direct succession, such as in a zig-zag way for phases of alternated sign between corresponding lower limit and upper limit in absolute value. Preferably, at least one (following) image portion following a first image portion (along their acquisition succession) is acquired for phases decreasing in absolute value from the corresponding upper limit to the corresponding lower limit (to ensure that the k-space rows for lower phases, providing more contrast information, are less adversely affected by the different (high-contrast / reduced-contrast / common) acquisition protocol of the preceding image portion). For example, in case of only a single high-contrast portion (Hl), a single reduced-contrast portion (Rl) and a single common portion (Cl), they may be acquired in the following order:

[0085] Rl: 0, +1, -1, +2, -2 +L’-2, -L’+2, +L’-1, -L’+l, +L’, -L’,

[0086] Hl: -L’, +L’, -L’+l, +L’-1, -L’+2, +L’-2 ... -2, +2, -1, +1, 0,

[0087] Cl: -L’, +L’, -L’+l, +L’-1, -L’+2, +L’-2, ... -M+2, +M-2, -M+l, +M-1, -M,

[0088] +M,

[0089] (with L’=8-64, such as L’=16).

[0090] With reference now to block 550, in the example at issue the k-space high- contrast / reduced-contrast images are converted into the corresponding magnitude high- contrast / reduced-contrast images (from / to the corresponding repositories). For this purpose, the k-space images are converted to complex form by applying an inverse Fourier transform thereto. This generates corresponding complex high- contrast / reduced-contrast images (being saved into a corresponding temporary memory structure). Each complex (high-contrast / reduced-contrast) image is defined by a matrix of cells for the locations of the body -part (with the same size of the magnitude images); each cell contains a complex number representing the response signal being received from the corresponding location of the body-part. The complex high- contrast / reduced-contrast images are then converted into the corresponding magnitude high-contrast / reduced-contrast images, by setting each pixel value thereof to the modulus of the complex number of the same cell in the complex high- contrast / reduced-contrast images. Similar considerations apply if the magnitude high- contrast / reduced-contrast images are generated directly from the corresponding k- space high-contrast / reduced-contrast images (without passing through the complex high- contrast / reduced-contrast images).

[0091] Preferably, a representation of the background is normalized in the (magnitude) high-contrast / reduced-contrast images at block 552. In fact, the (total / partial) flattening of the effect of the contrast agent generally changes the representation of the background in the reduced-contrast images. For example, in the case of the gradient echo acquisition method, in the totally flattened reduced-contrast image the representation of the background may become darker (while the representation of the contrast agent becomes the same) and in the partially flattened reduced-contrast images the representation of the background may become less dark (while the representation of the contrast agent becomes less bright). As a consequence, the representation of the background in the high-contrast / reduced-contrast images may differ. Its normalization then substantially restores the same representation of the background in the high- contrast / reduced-contrast images (but maintaining the flattening of the effect of the contrast agent in the reduced-contrast images), with a beneficial effect on any comparison among the high-contrast / reduced-contrast images that is involved by their use. For example, for this purpose the high-contrast image and each possible partially flattened reduced-contrast image are segmented into two classes of cells representative of the contrast agent (contrast class) and of the background (background class), such as by applying the Otsu's method (with all the cells of a possible totally flattened reduced-contrast image that are automatically assigned to the background class). A (high-contrast) descriptor of the pixel values of the background class of the high- contrast image and a corresponding (reduced-contrast) descriptor of the pixel values of the background class of each reduced-contrast image summarizing them (for example, their averages) are calculated. The pixel values of the reduced-contrast images (or at least the ones of the background class) are then rescaled (for example, linearly) so as to equalize the reduced-contrast / high-contrast descriptors. For this purpose, a scaling factor is calculated for each reduced-contrast image as the ratio between the high-contrast descriptor and its reduced-contrast descriptor; the pixel values of the reduced-contrast image are then multiplied by the scaling factor. For example, let us consider a high-contrast image wherein its pixel values representative of the contrast agent have an average of 250 and its pixel values representative of the background have an average of 40; moreover, let us consider a totally flattened reduced-contrast image wherein all its pixel values (representative of the contrast agent and the background) have an average of 20, and a partially flattened reduced-contrast image wherein its pixel values representative of the contrast agent have an average of 150 and its pixel values representative of the background have an average of 80. In this case, the totally flattened reduced-contrast image is rescaled according to a scaling factor equal to 40 / 20=2, so as to have the average of its pixel values (representative of the contrast agent and of the background) equal to 20-2=40; likewise, the partially flattened reduced-contrast image is rescaled according to a scaling factor equal to 40 / 80=0.5, so as to have the average of its pixel values representative of the contrast agent equal to 150-0.5=75 and the average of its pixel values representative of the background equal to 80-0.5=40. In both cases, in each (totally / partially flattened) reduced-contrast image the average of its pixel values representative of the background becomes the same as in the high-contrast image (40); at the same time, in the totally flattened reduced contrast image the difference between the average of its pixel values representative of the contrast agent and the average of its pixel values representative of background remains zero (40-40=0), whereas in the partially flattened reduced contrast image the difference between the average of its pixel values representative of the contrast agent and the average of its pixel values representative of the background remains lower than in the high-contrast image (75-40=35 with respect to 250-40=210).

[0092] Continuing to block 554, the (possibly normalized) high-contrast image and reduced-contrast image(s) are then used to generate the output image as usual. For this purpose, the flow of activity branches according to the configuration of the MRI system. For example, when the MRI system is configured to simulate an increase of the contrast agent (or of the delay from the administration of the contrast agent and / or of the efficiency of the contrast agent) blocks 556-558 are executed, whereas when the MRI system is configured to highlight the representation of the contrast agent blocks 560-562 are executed; in both cases, the flow of activity then merges again at block 564.

[0093] Considering now block 556 (contrast agent increasing), an increased-contrast image (simulating a representation of the body-part comprising an increased-amount of the contrast agent, higher than the high-amount) is generated by applying the high- contrast image and the totally flattened reduced-contrast image (surrogate of the nocontrast image) to a simulation model (for example, a machine learning model, such as described in WO-A-2023 / 062202). Particularly, when the contrast agent is of exogen type, the high-contrast / reduced-contrast images have been acquired after administration of the contrast agent to the patient (at the high-dose), with the totally flattened reduced-contrast image surrogate of the no-contrast image that would have been acquired before administration of the contrast agent to the patient. If the contrast agent that has been administered is less than the full-dose, the increased-contrast image may simulate its increase to the full-dose (thereby allowing administering less contrast agent, especially useful when it may be dangerous for the patient, at the same time restoring the contrast-enhancement to the desired level that would have been provided by the contrast agent at the full-dose); alternatively, if the contrast agent that has been administered is equal to the full-dose, the increased-contrast image simulates an increase of the contrast agent to a boosted-dose, higher than the full-dose (thereby allowing obtaining a contrast-enhancement being higher than the one actually attainable in practice, at the same time without affecting the standard of care). An output image based on the increased-contrast image, for example, defined by the increased-contrast image itself or by a combined image generated by applying High Dynamic Range (HDR) techniques to the increased-contrast image and the no-contrast image, is then output at block 558 (for example, displayed on the monitor of the MRI system). The flow of activity then descends into block 564. Similar considerations apply if the output image is generated by applying the high-contrast image and the totally flattened reduced-contrast image (surrogate of the no-contrast image) to a (further) simulation model configured to simulate an increase of the delay from the administration of the contrast agent to the patient or an increase of the efficiency of the contrast agent in the medical imaging application (for example, as described in WO2025 / 137795). Considering instead block 560 (contrast highlighting), an isolated- contrast image (representative of only the contrast agent in the body -part) is generated by subtracting the totally flattened reduced-contrast image (surrogate of the nocontrast image) from the high-contrast image. An output image based on the isolated contrast image (for example, a combined image generated by superimposing the isolated-contrast image in color on the no-contrast image in black-and-white) is then output at block 562 as above. This facilitates detailed anatomical comparisons, functional assessments, and pathological analyses by reducing discrepancies in spatial positioning. The flow of activity then descends into block 564.

[0094] With reference now to block 564, optionally a training set is logged into a log repository (for example, storing a result of the most recent imaging scans of the MRI system in the control unit). The training set may be incomplete, z.e., only comprising the no-contrast image that is surrogated by the totally flattened reduced-contrast image and the high-contrast image, or complete, z.e., further comprising a low-contrast image (or more) that is surrogated by a corresponding partially-flattened reduced-contrast image. One or more training sets are exported at block 566 (for example, periodically) to the training system (for example, via the network or one or more removable storage units). The process then ends at the concentric white / black stop circles 568.

[0095] Moving to the swim-lane of the training system, in a completely independent way the training sets of one or more MRI systems are collected at block 570. Optionally, the training system may also collect further training sets that are actually acquired, and particularly further complete training sets (for example, acquired in laboratories with pre-clinical studies on animals) and / or further incomplete training sets (for example, acquired in corresponding standard imaging sessions). In any case, if necessary, each incomplete training set is completed by simulating the corresponding low-contrast image from its no-contrast image and high-contrast image (as described in WO-A-2023 / 062196). Wherever it is necessary (for example, in response to any significant change of operative conditions of the MRI systems, in case of release of a new version of the machine learning model and so on), the machine learning mode is trained at block 572 by means of the training sets (being all complete), so as to optimize its capability of generating the high-contrast image from the nocontrast image and the low-contrast image of each training set (for example, again as described in WO-A-2023 / 062196); the (trained) machine learning is then deployed to the MRI systems (for example, via the network or one or more removable storage units). The process then returns to block 570 to repeat the same operations continually. The above-mentioned generation of the training sets (comprising one or more reduced- contrast images) allows collecting a relatively high number of them in the normal clinical practice, without requiring multiple administrations of the contrast agent for the corresponding low-contrast images (and then without affecting the standard of care); moreover, when the high-contrast / reduced-contrast images are acquired close to each other, they are substantially aligned automatically. All of the above significantly simplifies the collection of the training sets and increases the quality of the training of the machine learning model (with a positive impact on its robustness).

[0096] With reference now to FIG.6, representative examples are shown of in-vitro experimental results relating to the solution according to an embodiment of the present disclosure.

[0097] Particularly, a sample holder with three compartments (enriched with small plastic structures, such as grids) was used. A phantom was prepared by filling the compartments with water solution having different concentrations (0 mM, 0.25 mM and 1 mM) of a contrast agent (ProHance® by Bracco Imaging S.p. A., trademark thereof). Multiple sets of images of the phantom were acquired. Each set comprises a totally flattened reduced-contrast image (surrogate of a no-contrast image) acquired with a reduced-contrast acquisition protocol providing a total flattening of the effect of the contrast agent (3D T1 -weighted gradient echo with repetition time TR=5.8ms and flip angle FA=1°), a partially flattened reduced-contrast image (surrogate of a low-contrast image) acquired with a reduced-contrast acquisition protocol providing a partial flattening of the effect of the contrast agent (3D T1 -weighted gradient echo with repetition time TR=20ms and flip angle FA=10°), and a high-contrast image acquired using a high-contrast acquisition protocol providing an optimal contrast weighting (3D Tl-weighted gradient echo with repetition time TR=20ms and flip angle FA=30°).

[0098] Particularly, a no-contrast image 605o, a low-contrast image 6051 and a high- contrast image 605h were acquired in the succession mode. The images 605o,6051,605h where acquired in the k-space row-by-row, in a zig-zag way for phases of alternated sign increasing in absolute value from zero to a maximum value thereof equal to 128.

[0099] Moreover, multiple sets of images were acquired in the shared mode with different sizes of the corresponding image portions. For this purpose, a single high- contrast portion of the high-contrast image, a single reduced-contrast portion of the low-contrast image, a single reduced-contrast portion of the no-contrast image, and a single common portion were acquired in succession; the common portion was acquired with a common acquisition protocol equal to the same reduced-contrast acquisition protocol providing the total flattening of the effect of the contrast agent as above (3D Tl-weighted gradient echo with repetition time TR=5.8ms and flip angle FA=1°). The image portions where acquired for corresponding phases in depth (along a z-axis) and for all the phases in width (along a y-axis), in both cases row-by-row, in a zig-zag way for phases of alternated sign. Along the z-axis, each high-contrast / reduced-contrast portion extends at the center of the k-space (for phases ranging between 0 and an intermediate value in absolute value), whereas the common portion extends at the periphery of the k-space (for phases ranging between the intermediate value and the maximum value in absolute value); particularly, the high-contrast portion was acquired for phases increasing in absolute value from zero to the intermediate value, the reduced-contrast portion of the low-contrast image was acquired for phases decreasing in absolute value from the intermediate value to zero, the reduced-contrast portion of the no-contrast image was acquired for phases decreasing in absolute value from the intermediate value to zero, and the common portion were acquired for phases increasing in absolute value from the intermediate value to the maximum value. Along the y-axis, each image portion was acquired for phases increasing in absolute value from zero to the maximum value. The figure shows a set of images comprising a nocontrast image 610o, a low-contrast image 6101 and a high-contrast image 61 Oh acquired with the intermediate value equal to 16, a set of images comprising a nocontrast image 615o, a low-contrast image 6151 and a high-contrast image 615h acquired with the intermediate value equal to 32, and a set of images comprising a nocontrast image 620o, a low-contrast image 6201 and a high-contrast image 620h acquired with the intermediate value equal to 64.

[0100] As can been seen, the different compartments of the phantom are well represented in the images 610o,6101,610h, in the images 615o,6151,615h and in the images 620o,6201,620h acquired in the shared mode (independently of the size of the corresponding common portions), substantially the same as in the images 605o,6051,650h acquired in the succession mode (with a small worsening of quality, negligible in practice, as the sizes of the common portions increase).

[0101] With reference now to FIG.7A-FIG.7E, representative in-vivo examples are shown of experimental results relating to the solution according to an embodiment of the present disclosure.

[0102] Particularly, a dedicated pre-clinical study was carried out on three (Wistar) rats. The rats underwent a surgical procedure to orthotopically induce a tumor (C6 glioma) in their brain. After 12-14 days from the induction of the tumor, the rats underwent an imaging session of the brain using a 7T preclinical MRI scanner (Biospec 70 / 16, Bruker). For each rat, the following images were acquired.

[0103] As shown in FIG.7A for a representative rat, before administration of any contrast agent, a high-contrast acquisition protocol providing an optimal contrast weighting (3D T1 -weighted gradient echo with repetition time TR=20ms and flip angle FA=30°) was used to acquire a corresponding (actual) no-contrast image 705o. After administration of a contrast agent (ProHance® by Bracco Imaging S.p.A., trademark thereof) at its full-dose (0.1 mmol Gd / kg), the same high-contrast acquisition protocol was used to acquire a corresponding (actual) high-contrast image 705h. Moreover, a reduced-contrast acquisition protocol providing a total flattening of the effect of the contrast agent (3D T1 -weighted gradient echo with repetition time TR=8.6ms and flip angle FA=1°) was used to acquire a totally flattened reduced-contrast image to be used as surrogate of a (surrogated) no-contrast image 710o, and another reduced-contrast acquisition protocol providing a partial flattening of the effect of the contrast agent (3D Tl-weighted gradient echo with repetition time TR=20ms and flip angle FA=10°) was used to acquire a partially flattened reduced-contrast image to be used as surrogate of a (surrogated) low-contrast image 7101. As can be seen, the actual no-contrast image 705o and the surrogated no-contrast image 710o are substantially the same; moreover, the surrogated low-contrast image 7011 provides a reduced contrast enhancement with respect to the actual high-contrast image 705h.

[0104] Moving to FIG.7B, the surrogated no-contrast image 710o and the actual high- contrast image 705h were normalized to each other as described above, and then to a common range of voxel values from 0 to 1. Multiple (simulated boosted-contrast) images were generated to simulate different increases of the contrast agent from the (normalized) surrogated no-contrast image 710o and actual high-contrast image 705h. Each simulated boosted-contrast image is simulated by setting each voxel value thereof to:

[0105] Vb=V0+(Vh-Vo) k wherein Vb is the voxel value of the simulated boosted-contrast image, VO is the voxel value of the surrogated no-contrast image 710o, Vh is the voxel value of the actual high-contrast image 705h and k ’ is calculated by applying the following formula: wherein k is an increasing factor (of a desired increase of the contrast agent to be simulated), Fg is a function applying a Gaussian filter to its argument and si, sh are tunable parameters (sl=50 and 5 / 1=0.02); the arguments of the Gaussian filter are the difference (Vh-Vo), between the voxel values of the actual high-contrast image 705h and of the surrogated no-contrast image 710o, and a tunable standard deviation sg (for example, sg=4 voxels). Particularly, the figure shows three simulated boosted-contrast images that were generated with different values of the increasing factor k, denoted with the references 715b2 (for k=2), 715b3 (for k=3) and 715b4 (for k=4). As can been seen, the representation of the tumor is more conspicuous in the simulated boosted- contrast images 715b2-715b4 than it is in the actual high-contrast image 705h (with the higher the increasing factor k the more conspicuous the representation of the tumor).

[0106] Moving to FIG.7C, different metrics were calculated of the actual high-contrast image and the simulated boosted-contrast images. Particularly, each (actual high- contrast / simulated boosted-contrast) image was segmented manually into a (tumor) area representative of the tumor and a corresponding (healthy) area in a contralateral region of a parenchyma of the brain. A contrast to noise ratio (CNR) was calculated as:

[0107] CNR=(Mt-Mh) / SDh, wherein Mt is the mean of the voxel values in the tumor area, Mh is the mean of the voxel values in the healthy area and SDh is the standard deviation of the voxel values in the healthy area. Moreover, a contrast enhancement percentage (CEP) was calculated as:

[0108] CEP =100- (Mt-Mo) / Mo, wherein Mo is the mean of the voxel values corresponding to the tumor area in the actual no-contrast image. The values of the CNR and the CEP so calculated are shown in corresponding diagrams 720a and 720b, respectively (plotting them on the ordinate axis for the actual high-contrast image (full) and the simulated boosted-contrast images (k=2, k=3 and k=4) on the abscissa axis). As can been seen, both the CNR and the CEP are better in the simulated boosted-contrast images than they are in the actual high- contrast image (with the higher the increasing factor k the higher the CNR and the CEP).

[0109] Moving to FIG.7D, different simulated boosted-contrast images were simulated from the same (normalized) surrogated no-contrast image 710o and actual high-contrast image 705h by means of the neural network described in WO-A- 2023 / 062202. Particularly, the figure shows three simulated boosted-contrast images that were generated with different values of the increasing factor k, denoted with the references 715b2’ (for k=2), 715b3’ (for k=3) and 715b4’ (for A=4). As can been seen, in this case as well the representation of the tumor is more conspicuous in the simulated boosted-contrast images 715b2’-715b4’ than it is in the actual high-contrast image 705h (with the higher the increasing factor k the more conspicuous the representation of the tumor).

[0110] Moving to FIG.7E, the CNR and the CEP of the simulated boosted-contrast images were calculated as above (in addition to the ones of the actual high-contrast image). The values of the CNR and the CEP so calculated are shown in corresponding diagrams 720a’ and 720b’ as above. As can been seen, both the CNR and the CEP are again better in the simulated boosted-contrast images (k=2, k=3 and k=4) than they are in the actual high-contrast image (full), with the higher the increasing factor k the higher the CNR and the CEP.

[0111] Modifications

[0112] In order to satisfy local and specific requirements, a person skilled in the art may apply many logical and / or physical modifications to the present disclosure, provided that it remains within the scope of the claims. Particularly, the present disclosure may be practiced even without the specific details (such as the numerical values) set forth in the preceding description to provide a more thorough understanding thereof; conversely, well-known features may have been omitted or simplified in order not to obscure the description with unnecessary particulars. Specific features described in connection with any embodiment of the present disclosure may be incorporated in any other embodiment as a matter of general design choice. Moreover, items presented in a same group and different embodiments, examples or alternatives are not to be construed as de facto equivalent to each other (but they are separate and autonomous entities). In any case, each numerical value should be read as modified according to the applicable tolerances; particularly, unless otherwise indicated, the terms “substantially”, “about”, “approximately” and the like should be intended as within 10%, preferably 5% and still more preferably 1%. Moreover, each range of numerical values should be intended as expressly specifying any possible number along the continuum within the range (comprising its end points). Ordinal or other qualifiers are merely used as labels to distinguish elements with the same name but do not by themselves connote any priority, precedence or order. The terms include, comprise, have, contain, involve and the like should be intended with an open, non-exhaustive meaning (z.e., not limited to the recited items), the terms based on, dependent on, in agreement with, according to, function of and the like should be intended as a nonexclusive relationship (z.e., with possible further variables involved), the term a / an should be intended as one or more items (unless expressly indicated otherwise), the terms and / or, at least one of, one or more of and the like with respect to a list of two or more entities should be understood comprising each one of such entities individually and any combination of any number of such entities (with the possible addition of other entities), and the term means for (or similar functional formulation) should be intended as any structure adapted or configured for carrying out the relevant function.

[0113] More specifically, each of the following modifications may be applied (alone or in combination with any other modification) to the corresponding features mentioned above. Particularly, it is expressly understood that each feature mentioned above may be replaced by any of its alternatives or it may be generalized to the corresponding genus as set out in the following.

[0114] For example, an embodiment provides an (imaging) method for imaging a body-part of a patient with an MRI system comprising an MRI scanner and a control unit. However, the MRI system may be of any type (see below) for imaging any bodypart (for example, organs, regions thereof, tissues, bones, joints and the like in any condition, such as healthy, pathological with any lesions and so on) of any patient (for example, persons, animals and so on). Moreover, the method may be used in any medical application (see below). In any case, although the method may facilitate the task of a physician, it only provides intermediate results that may help him / her but with the medical activity stricto sensu that is always made by the physician himself / herself.

[0115] In an embodiment, the method comprises the following steps under the control of the control unit. However, the control unit may be of any type (see below).

[0116] In an embodiment, the method comprises controlling (by the control unit) the MRI scanner. However, the MRI scanner may be of any type (see below). In an embodiment, the method comprises controlling (by the control unit) the MRI scanner to acquire a high-contrast image of the body-part comprising a high- amount of a contrast agent. However, the high-contrast image may be of any type (for example, providing a spatial / frequency representation, in 2D / 3D, with any size and resolution, with pixel / voxels values having any chromaticity and bit depth, and so on) and acquired in any way (for example, in k-space traversing it along any trajectory, such as row-by-row, spiral, radial and the like, and so on). Moreover, the body-part may comprise any type of contrast agent (for example, exogen, endogen, of paramagnetic type, superparamagnetic type and so on) in the high-amount having any value.

[0117] In an embodiment, the high-contrast image is acquired with a high-contrast acquisition protocol. However, the high-contrast acquisition protocol may be of any type (for example, based on any acquisition method, defined by any number and type of acquisition parameters having any values, and so on).

[0118] In an embodiment, the method comprises controlling (by the control unit) the MRI scanner to acquire one or more reduced-contrast images of the body-part comprising the high-amount of the contrast agent. However, the reduced-contrast images may be in any number and of any type (for example, either the same or different with respect to the high-contrast image, and so on).

[0119] In an embodiment, the reduced-contrast images are acquired with corresponding reduced-contrast acquisition protocols having corresponding contrast weightings (lower with respect to the high-contrast acquisition protocol) providing corresponding total or partial flattenings, between a representation of locations of the body -part comprising the contrast agent and a representation of locations of the bodypart without the contrast agent, in the reduced-contrast images with respect to the high- contrast image. However, the reduced-contrast acquisition protocols may be of any type (for example, based on any acquisition method and defined by any number and type of acquisition parameters, either the same or different with respect to the high- contrast acquisition protocol, with any values of the acquisition parameters, and so on); moreover, the contrast weightings of the reduced-contrast acquisition protocols may be lower than the one of the high-contrast acquisition protocol in any way (for example, by any absolute / relative value and so on) for providing any flattening of the effect of the contrast agent (for example, quantified by any absolute / relative reduction of a difference between any descriptors of the representations of the contrast agent and of the background, such as any central tendency indicators of the corresponding values, such as their mean, median, mode, range and the like, providing a total flattening, any partial flattening in absolute or relative terms, and so on).

[0120] In an embodiment, the method comprises generating (by the control unit) an output image of the body-part providing an increased contrast-enhancement with respect to the high-contrast image. However, the output image may provide any type of increased contrast-enhancement (for example, simulating an increased amount of the contrast agent, highlighting the representation of the contrast agent, simulating an increased delay from administration of the contrast agent (z.e., with the high-contrast image that is acquired with an original delay from the administration of the contrast agent to the patient and the output image that simulates an increased delay from the administration of the contrast agent to the patient higher than the original delay), and / or simulating an increased efficiency of the contrast agent (z.e., with the contrast agent that provides an original efficiency in the medical imaging application and the output image that simulates administration to the patient of a further contrast agent providing an increased efficiency in the medical imaging application higher than the original efficiency, such as with the contrast agent that is specific for another medical imaging application and the output image that simulates the further contrast agent being specific for the medical imaging application), in the form of a representation of the body-part and / or of a parametric map, and so on).

[0121] In an embodiment, the output image is based on the high-contrast image and one or more surrogated images comprising a no-contrast image (surrogating a representation of the body-part without the contrast agent) and / or at least one low- contrast images (surrogating a representation of the body-part comprising a corresponding low-amount of the contrast agent lower than the high-amount). However, the surrogated images may be in any number (for example, only the nocontrast image, one or more low-contrast images, any combination thereof and so on) and of any type (for example, surrogating an actual no-contrast image that would have been acquired before administration of an exogen contrast agent, a virtual no-contrast image of the body-part with an endogen contrast agent mimicking its removal, any actual low-contrast image that would have been acquired after administration of any reduced-amount of an exogen contrast agent, any virtual low-contrast image of the body-part with an endogen contrast agent mimicking any decrease thereof, corresponding to low-amounts of the contrast agent having any value in either absolute terms or relative terms with respect to the high-amount, and so on).

[0122] In an embodiment, the no-contrast image is surrogated by one of the reduced- contrast images providing the total flattening and the low-contrast image is surrogated by one of the reduced-contrast images providing a corresponding one of the partial flattenings. However, the reduced-contrast images may be used in place of any surrogated images (for example, only the no-contrast image, only any number of low- contrast images, both of them, and so on).

[0123] In an embodiment, the method comprises outputting (by the control unit) the output image. However, the output image may be output in any way (for example, displayed on any device, such as a monitor, virtual reality glasses and the like, or more generally output in real-time or off-line in any way, such as printed, transmitted remotely and so on).

[0124] Further embodiments provide additional advantageous features, which may however be omitted at all in a basic implementation. In this respect, it is expressly understood that the features of each of the following embodiments may be combined with the above features either alone or in combination with the features of any number of the other following embodiments.

[0125] In an embodiment, the method comprises generating (by the control unit) the output image by applying the high-contrast image and the surrogated images to a simulation model being configured to simulate a representation of the body-part comprising an increased-amount of the contrast agent higher than the high-amount. However, the output image may be generated by applying any number and type of surrogated images (see above) to any simulation model (for example, based on any machine learning model like a neural network, a convolutional neural network, a dense neural network, a linear regression, a gaussian process, a decision tree and the like, a noise reduction filter and similar, or based on any analytic techniques like involving additions, multiplications, subtractions, thresholding, linear / non-linear functions and the like, and so on) being configured to simulate any increased-amount of the contrast agent (for example, in either absolute or relative terms with respect to the high-dose, and so on).

[0126] In an embodiment, the method comprises controlling (by the control unit) the MRI scanner to acquire the high-contrast image and the reduced-contrast images after administration to the patient of the contrast agent at a high-dose. However, the contrast agent may be of any type (for example, any targeted contrast agent based on specific or non-specific interactions, such as gadobenate, gadoxetate and the like, any nontargeted contrast agent, such as gadoteridol, gadopiclenol, gadobutrol, gadoterate, gadopentetate, gadodiamide and the like, and so on) and it may be administered in any way (for example, by a syringe, a pump and so on) at any high-dose (for example, equal to the full-dose, lower than the full-dose by any amount, and so on); moreover, the contrast agent may have been administered at any administration instant (for example, at the beginning of a corresponding imaging session, in advance and so on) and the images may be acquired at any acquisition instant following it, between a wash-in phase and a wash-out phase of the contrast agent (for example, immediately, with any delay and so on).

[0127] In an embodiment, the method comprises generating (by the control unit) the output image simulating a representation of the body-part after administration to the patient of the contrast agent at an increased-dose higher than the high-dose. However, the increased-dose may have any value (for example, equal to the full dose, any boosted-dose higher than the full-dose, and so on).

[0128] In an embodiment, the output image is generated by applying the high-contrast image and the surrogated images, comprising the no-contrast image surrogating a representation of the body-part before administration to the patient of the contrast agent and / or the low-contrast images surrogating a representation of the body-part after administration to the patient of corresponding low-doses of the contrast agent lower than the high-dose, to the simulation model. However, the output image may be generated from any surrogated images (for example, the no-contrast image only, any number of low-contrast images corresponding to any low-doses of the contrast agent, any combination thereof, and so on).

[0129] In an embodiment, the high-dose is lower than a full-dose being a standard in clinical practice and the increased-dose is equal to the full-dose. However, the fulldose may be of any type (for example, depending on the type of the medical imaging procedure, on the type of the body-part, on the type, weight, age and the like of the patient, fixed, and so on) and the high-dose may have any value (for example, in either absolute or relative terms with respect to the full-dose).

[0130] In an embodiment, the high-dose is equal to the full-dose and the increased- dose is higher than the standard in clinical practice. However, the increased-dose may have any value (for example, in either absolute or relative terms with respect to the full-dose).

[0131] In an embodiment, the contrast agent is of endogen type. However, the endogen contrast agent may be of any type (for example, based on diamagnetic T1 of tissues, diamagnetic T2 of tissues, water diffusion coefficient, bold effect and so on).

[0132] In an embodiment, the method comprises generating (by the control unit) an isolated-contrast image by subtracting the no-contrast image from the high-contrast image. However, the isolated-contrast image may be generated in any way (for example, by subtracting the no-contrast image from the high-contrast image at the level of pixel / voxel values, groups thereof and so on).

[0133] In an embodiment, the output image is based on the isolated-contrast image for highlighting the representation of the locations of the body -part containing the contrast agent with respect to the high-contrast image. However, the output image may be based on the isolated-contrast image in any way (for example, by superimposing the isolated-contrast image on any baseline image, such as the actual / surrogated nocontrast image, with any highlighting of the representation of the contrast agent with respect to the background, such as any color and black-and-white, different colors, different intensities and so on).

[0134] In an embodiment, the method comprises generating (by the control unit) the output image by superimposing the isolated-contrast image in color on the no-contrast image in black-and-white. However, the isolated-contrast image may be superimposed on the no-contrast image in any way (for example, for any pixel / voxels of the isolated- contrast image different from zero, higher than any threshold, as so on).

[0135] In an embodiment, the method comprises exporting (by the control unit) one or more training sets to a training computing system. However, the training sets may be in any number and exported in any way (for example, periodically, when they reach a predefined number or in response to a manual command, from the control unit or any other computing system to which they have been uploaded, via any number of removable storage units or any network connection, and so on) to any training computing system (for example, a personal computer, a server, a virtual machine, a cloud service and so on).

[0136] In an embodiment, each of the training sets comprises the high-contrast image and the reduced-contrast images resulting from a corresponding iteration of said controlling the MRI scanner to acquire the high-contrast image and said controlling the MRI scanner to acquire the reduced-contrast images, respectively. However, the training set may be of any type (for example, comprising the no-contrast image and any number of low-contrast images, with no low-contrast image to be simulated from the no-contrast image and the high-contrast image, and so on).

[0137] In an embodiment, the training sets are for use by the training computing system to train a machine learning model to optimize a capability thereof to generate the high-contrast image from the reduced-contrast images of each training set. However, the machine learning model may be of any type (see above) and it may be trained in any way (for example, by selecting any training / verification sets from the training sets, using any algorithm, with or without any number of additional training sets, such as each comprising all its images being acquired, the high-contrast image and the low-contrast image being acquired and the low-contrast image being simulated therefrom, and so on).

[0138] In an embodiment, the high-contrast acquisition protocol is defined by an acquisition method and corresponding high-contrast values of one or more acquisition parameters. However, the acquisition method may be of any type (for example, gradient echo, spin echo, fast spin echo, echo planar, compressed sensing and so on) in any acquisition protocol (for example, T1 -weighed, T2-weighed, diffusion- weighted, proton density and so on); moreover, the acquisition parameters may be in any number, of any type (for example, repetition time, flip angle, echo time, b-value and so on) and with any high-contrast values.

[0139] In an embodiment, each of the reduced-contrast acquisition protocols is defined by the acquisition method and corresponding reduced-contrast values of the acquisition parameters. However, the reduced-contrast values of each reduced-contrast protocol may be of any type (for example, completely / partially different from the corresponding high-contrast values and from the corresponding reduced-contrast values of the other reduced-contrast acquisition protocols, with any differences, and so on). Possible examples are, in addition to the ones described above, lower echo time in a T2-weigthed acquisition protocol, lower b-value in a diffusion-weighted acquisition protocol, lower repetition time and / or higher echo time in a proton density acquisition protocol, and so on.

[0140] In an embodiment, the high-contrast acquisition protocol is Tl-weigthed. However, the Tl-weighted acquisition protocol may be of any type (for example, with any high-contrast values of any acquisition parameters falling within corresponding ranges defining it, such as time to echo, repetition time, flip angle and so on).

[0141] In an embodiment, the reduced-contrast acquisition protocols are less Tl- weighted than the high-contrast acquisition protocol. However, the reduced-contrast acquisition protocols may be less Tl-weighted than the high-contrast acquisition protocol in any way (for example, sub-optimal / optimal, with any difference, and so on).

[0142] In an embodiment, the acquisition method is gradient echo. However, the gradient echo acquisition method may be of any type (for example, with any parameters falling within corresponding ranges defining it, such as gradient amplitude and duration, spoiling and so on).

[0143] In an embodiment, the reduced-contrast acquisition protocol providing the total flattening has a repetition time equal to the repetition time of the high-contrast acquisition protocol and a flip angle lower than the flip angle of the high-contrast acquisition protocol. However, these flip angles may differ in any way (for example, by any extent, in absolute / relative terms, and so on), and these repetition times may be either the same or slightly different.

[0144] In an embodiment, the reduced-contrast acquisition protocol providing the partial flattening has the repetition time lower than the repetition time of the high- contrast acquisition protocol and the flip angle lower than the flip angle of the high- contrast acquisition protocol. However, these repetition times and these flip angles may differ in any way (for example, by any extent, in absolute / relative terms, and so on).

[0145] In an embodiment, the flip angle of the reduced-contrast acquisition protocol providing the partial flattening is higher than the flip angle of the reduced-contrast acquisition protocol providing the total flattening. However, these flip angles may differ in any way (for example, by any extent, in absolute / relative terms, and so on).

[0146] In an embodiment, the acquisition method is spin echo. However, the spin echo acquisition method may be of any type (for example, with any parameters falling within corresponding ranges defining it, such as number of refocusing pulses, echo train length and so on).

[0147] In an embodiment, the repetition time of the reduced-contrast acquisition protocols is higher than the repetition time of the high-contrast acquisition protocol. However, these repetition times may differ in any way (for example, by any extent, in absolute / relative terms, and so on).

[0148] In an embodiment, the repetition time of the reduced-contrast acquisition protocol providing the partial flattening is lower than the repetition time of the reduced-contrast acquisition protocol providing the total flattening. However, these repetition times may differ in any way (for example, by any extent, in absolute / relative terms, and so on).

[0149] In an embodiment, the method comprises normalizing (by the control unit) the representation of the locations of the body -part without the contrast agent in the high- contrast image and in the reduced-contrast images. However, the high- contrast / reduced-contrast images may be normalized in any way (for example, detecting the representation of the background by applying any segmentation technique, updating the high-contrast image and / or the reduced-contrast images, rescaling the whole images or only the representation of the background in any way, such as linearly, logarithmically and the like, and so on).

[0150] In an embodiment, the normalization is performed by scaling each reduced- contrast image according to a corresponding scaling factor depending on a ratio between a high-contrast descriptor of the representation of the locations of the bodypart without the contrast agent in the high-contrast image and a corresponding reduced- contrast descriptor of the representation of the locations of the body-part without the contrast agent in the reduced-contrast image. However, the descriptors may be of any type (for example, any central tendency indicators of the corresponding values, such as mean, median, mode and the like, their range and so on) and the scaling factor may be calculated and used in any way from them; for example, whenever any (high- contrast / reduced-contrast) image is to be updated to obtain a target value of its descriptor (for example, equal to the descriptor of another image remaining unchanged or to any different value), the scaling factor is calculated as the ratio between the target value and the descriptor of the image, and the scaling factor is then used to multiply the image.

[0151] In an embodiment, the method comprises controlling (by the control unit) the MRI scanner to acquire the high-contrast image and the reduced-contrast images in direct succession. However, the high-contrast image and the reduced-contrast images may be acquired in direct succession in any way (for example, in any order); in any case, the possibility of introducing a short delay between the acquisition of the images is not excluded.

[0152] In an embodiment, the method comprises controlling (by the control unit) the MRI scanner to acquire a plurality of image portions comprising a plurality of high- contrast portions of the high-contrast image and a plurality of reduced-contrast portions of each of the reduced-contrast images. However, these image portions may be in any number (for example, with any number of high-contrast portions and any number of reduced-contrast portions of each reduced-contrast image, either the same or different, and so on) and of any type (for example, with any constant / varying extent, as a whole covering each image completely or only partially, and so on).

[0153] In an embodiment, the high-contrast portions and the reduced-contrast portions are acquired in direct succession interleaved to each other. However, these image portions may be acquired in direct succession interleaved to each other in any way (for example, in any order); in any case, the possibility of introducing a short delay between the acquisition of the image portions is not excluded.

[0154] In an embodiment, the method comprises combining (by the control unit) the high-contrast portions into the high-contrast image and the corresponding reduced- contrast portions into each reduced-contrast image. However, the corresponding high- contrast / reduced-contrast portions may be combined into each high-contrast / reduced- contrast image in any way (for example, by simply juxtaposing them, by applying interpolation techniques, Al-based techniques and so on).

[0155] In an embodiment, the method comprises controlling (by the control unit) the MRI scanner to acquire a plurality of image portions comprising one or more high- contrast portions of the high-contrast image, one or more reduced-contrast portions of each of the reduced-contrast images and one or more common portions of a common image of the body-part comprising the high-amount of the contrast agent. However, these image portions may be in any number (for example, with any number of high- contrast portions, reduced-contrast portions of each reduced-contrast image and common portions, either the same or different, and so on) and of any type (for example, with any constant / varying extent, as a whole covering each image completely or only partially, with the high-contrast / reduced-contrast portions and the common portions being disjoint or partially overlapped, and so on).

[0156] In an embodiment, the common portions are acquired with a common acquisition protocol based on the high-contrast acquisition protocol and / or the reduced-contrast acquisition protocols. However the common acquisition protocol may be of any type (for example, fixed, varying among the common portions and / or within each of them, and so on).

[0157] In an embodiment, the high-contrast portions, the reduced-contrast portions and the common portions are acquired in direct succession interleaved to each other. However, these image portions may be acquired in direct succession interleaved to each other in any way (for example, in any order); in any case, the possibility of introducing a short delay between the acquisition of the image portions is not excluded.

[0158] In an embodiment, the method comprises generating (by the control unit) the high-contrast image from the high-contrast portions and the common portions and each reduced-contrast image from the corresponding reduced-contrast portions and the common portions. However, each high-contrast / reduced-contrast image may be generated from the corresponding high-contrast / reduced-contrast portions and the common portions in any way (for example, by applying Al-based techniques, interpolation techniques, by simply juxtaposing them and so on).

[0159] In an embodiment, the high-contrast portions and the reduced-contrast portions cover a center of a k-space corresponding to phases extending in absolute value between zero and an intermediate value and wherein the common portions cover a periphery of the k-space corresponding to phases extending in absolute value between the intermediate value and a maximum value higher than the intermediate value. However, the k-space may have any extent (defined by any maximum value) and its center / periphery may have any extent (defined by any intermediate value); moreover, the image portions may be acquired to cover the center / periphery of the k-space in any way (for example, each image portion is acquired for phases being adjacent / non- adjacent, increasing / decreasing in absolute value, of alternated signs or of the same sign, and so on).

[0160] In an embodiment, the method comprises controlling (by the control unit) the MRI scanner to acquire the common portions with the common acquisition protocol equal to the reduced-contrast acquisition protocol providing the total flattening. However, the possibility is not excluded of having the common acquisition protocol equal to any other fixed acquisition protocol (for example, a reduced-contrast acquisition protocol providing a partial flattening, the high-contrast acquisition protocol, any other acquisition protocol interposed among the reduced-contrast acquisition protocols and the high-contrast acquisition protocol, and so on).

[0161] In an embodiment, the method comprises controlling (by the control unit) the MRI scanner to acquire the common portions with the common acquisition protocol changing from one of the reduced-contrast acquisition protocols to the high-contrast acquisition protocol. However, the common acquisition protocol may change between any pair of acquisition protocols (for example, between any of the reduced-contrast acquisition protocols and the high-contrast acquisition protocol, between two reduced- contrast acquisition protocols of any type, in any order, with any law, such as sigmoidal, sinusoidal and the like, with any pitch and so on).

[0162] In an embodiment, the common acquisition protocol changes from the reduced- contrast acquisition protocol providing the total flattening to the high-contrast acquisition protocol. However, the common acquisition protocol may change between these acquisition protocols in any way (as above).

[0163] In an embodiment, the method comprises controlling (by the control unit) the MRI scanner to acquire each of the image portions comprising a plurality of k-space rows for phases ranging in absolute value between a corresponding lower limit and a corresponding upper limit. However, each image portion may comprise any number of rows for phases ranging between any lower / upper limit.

[0164] In an embodiment, at least one of the image portions following a first one of the image portions is acquired for phases decreasing in absolute value from the corresponding upper limit to the corresponding lower limit. However, this acquisition scheme may apply to any dimension (for example, in case of 3D images only in width, in depth, in both of them and so on) and to any following image portions (for example, only the ones extending around the zero phase, all of them and so on); each following image portion may be acquired for decreasing phases in any way (for example, for phases of alternated signs starting from a positive phase or from a negative phase, for phases of any sign and then of the other sign, and so on), with the first image portion that may be acquired in any way (for example, for phases increasing or decreasing in absolute value between the corresponding lower / upper limits, and so on).

[0165] Generally, similar considerations apply if the same solution is implemented with an equivalent method, provided that it remains within the scope of the claims (by using similar steps with the same functions of more steps or portions thereof, removing some non-essential steps or adding further optional steps); moreover, the steps may be performed in a different order, concurrently or in an interleaved way (at least in part).

[0166] An embodiment provides a computer program, which is configured for causing an MRI system to perform the method of above when the computer program is executed on a control unit of the MRI system. An embodiment provides a computer program product, the computer program product comprising one or more non- transitory computer readable storage media having program instructions collectively stored in the readable storage media, the program instructions readable by a control unit of an MRI system to cause the control unit to perform the same method. However, the (computer) program may be executed on any control unit (see below). The program may be implemented as a stand-alone module, as a plug-in for a pre-existing software program (for example, a control manager of the MRI system) or even directly in the latter.

[0167] Generally, similar considerations apply if the program is structured in a different way, or if additional modules or functions are provided (provided that it remains within the scope of the claims). Particularly, the program may take any form suitable to be used by the control unit thereby configuring it to perform the desired operations; the program may be in the form of external or resident software, firmware, or microcode (either in object code or in source code), for example, to be compiled or interpreted. Moreover, it is possible to provide the program on any computer readable storage medium. The storage medium is any tangible medium (different from transitory signals per se) that may retain and store instructions for use by the control unit. For example, the storage medium may be of electronic, magnetic, optical, electromagnetic, infrared, or semiconductor type; examples of such storage medium are fixed disks (where the program may be pre-loaded), removable disks, memory keys (for example, of USB type), and the like. The program may be downloaded to the control unit from the storage medium or via a network (for example, the Internet, a wide area network and / or a local area network comprising transmission cables, optical fibers, wireless connections, network devices); one or more network adapters in the control unit receive the program from the network and forward it for storage into one or more storage devices of the control unit. In any case, the solution according to an embodiment of the present disclosure lends itself to be implemented even with a hardware structure (for example, by electronic circuits integrated on one or more chips of semiconductor material), or with a combination of software and hardware suitably programmed or otherwise configured.

[0168] An embodiment provides an MRI system, which comprises an MRI scanner and a control unit configured for performing the steps of the above method. An embodiment provides an MRI system, which comprises an MRI scanner and a control unit comprising a circuit (z.e., any hardware suitably configured, for example, by software) for performing each step of the same method. However, the MRI system may be of any type (for example, working in standard mode, CEST mode and so on), and it may comprise any MRI scanner (for example, with magnets providing the magnetic field at any intensity, alternative gradient and / or RF coils) and any control unit (for example, a computer, a microcontroller and so on).

[0169] Generally, similar considerations apply if the MRI system has a different structure, comprises equivalent components or has other operative characteristics, provided that it remains within the scope of the claims. In any case, every component thereof may be separated into more elements, or two or more components may be combined together into a single element; moreover, each component may be replicated to support the execution of the corresponding operations in parallel. Moreover, unless specified otherwise, any interaction between different components generally does not need to be continuous, and it may be either direct or indirect through one or more intermediaries.

[0170] An embodiment provides a medical method. In an embodiment, the medical method comprises imaging a body-part of a patient as above. In an embodiment, the medical method comprises performing a medical procedure relating to the body-part according to the output image. However, the medical procedure may be of any type (for example, a diagnostic procedure, a therapeutic procedure, a surgical procedure and so on).

[0171] An embodiment provides a (training) method for training a machine learning model to simulate an increased contrast-enhancement in medical imaging applications. In an embodiment, the method comprises the following steps under the control of a training computing system. In an embodiment, the method comprises receiving (by the training computing system) a plurality of training sets, each of the training sets comprising the high-contrast image and the reduced-contrast images resulting from a corresponding iteration of said controlling the MRI scanner to acquire the high- contrast image and said controlling the MRI scanner to acquire the reduced-contrast images, respectively, of above. In an embodiment, the method comprises training (by the training computing system) the machine learning model to optimize a capability thereof to generate the high-contrast image from the reduced-contrast images of each of the training sets.

[0172] An embodiment provides a computer program, which is configured for causing a training computing system to perform the method of above when the computer program is executed on the training computing system. An embodiment provides a computer program product, the computer program product comprising one or more non-transitory computer readable storage media having program instructions collectively stored in the readable storage media, the program instructions readable by a training computing system to cause the training computing system to perform the same method.

[0173] An embodiment provides a training computing system configured for performing the steps of the above method. An embodiment provides a training computing system comprising a circuit (z.e., any hardware suitably configured, for example, by software) for performing each step of the same method.

[0174] However, the same considerations pointed out above with respect to the steps of the imaging method, the corresponding computer program and computer program product and the training computing system apply in this case as well.

Claims

1. 55CLAIMS1. A method (500) for imaging a body -part of a patient with an MRI system (120,130) comprising an MRI scanner (120) and a control unit (130), wherein the method (500) comprises, under the control of the control unit (130): controlling (510-512;520-522,532;534-536, 544-548), by the control unit (130), the MRI scanner (120) to acquire a high-contrast image of the body -part comprising a high-amount of a contrast agent, the high-contrast image being acquired with a high-contrast acquisition protocol, controlling (514-518; 524-532; 538-548), by the control unit (130), the MRI scanner to acquire one or more reduced-contrast images of the body-part comprising the high-amount of the contrast agent, the reduced-contrast images being acquired with corresponding reduced-contrast acquisition protocols having corresponding contrast weightings, lower with respect to the high-contrast acquisition protocol, providing corresponding total or partial flattenings, between a representation of locations of the body -part containing the contrast agent and a representation of locations of the bodypart without the contrast agent, in the reduced-contrast images with respect to the high- contrast image, generating (556;560), by the control unit (130), an output image of the bodypart providing an increased contrast-enhancement with respect to the high-contrast image, wherein the output image is based on the high-contrast image and one or more surrogated images comprising a no-contrast image surrogating a representation of the body -part without the contrast agent and / or at least one low-contrast image surrogating a representation of the body-part comprising a corresponding low-amount of the contrast agent lower than the high-amount, the no-contrast image being surrogated by one of the reduced-contrast images providing the total flattening and the low-contrast image being surrogated by one of the reduced-contrast images providing a corresponding one of the partial flattenings, and outputting (558; 562), by the control unit (130), the output image.

2. The method (500) according to claim 1, wherein the method (500) comprises:56 generating (556), by the control unit (130), the output image by applying the high-contrast image and the surrogated images to a simulation model being configured to simulate a representation of the body-part comprising an increased-amount of the contrast agent higher than the high-amount.

3. The method (500) according to claim 2, wherein the method (500) comprises: controlling (510-548), by the control unit (130), the MRI scanner (120) to acquire the high-contrast image and the reduced-contrast images after administration to the patient of the contrast agent at a high-dose, and generating (556), by the control unit (130), the output image simulating a representation of the body-part after administration to the patient of the contrast agent at an increased-dose higher than the high-dose, wherein the output image is generated by applying the high-contrast image and the surrogated images, comprising the nocontrast image surrogating a representation of the body-part before administration to the patient of the contrast agent and / or the low-contrast image surrogating a representation of the body-part after administration to the patient of a low-dose of the contrast agent lower than the high-dose, to the simulation model, particularly with the high-dose lower than a full-dose being a standard in clinical practice and the increased- dose equal to the full-dose or with the high-dose equal to the full-dose and the increased-dose higher than the standard in clinical practice.

4. The method (500) according to claim 2, wherein the contrast agent is of endogen type.

5. The method (500) according to any claim from 1 to 4, wherein the method (500) comprises: generating (560), by the control unit (130), an isolated-contrast image by subtracting the no-contrast image from the high-contrast image, the output image being based on the isolated-contrast image for highlighting the representation of the locations of the body -part containing the contrast agent with respect to the high-contrast image.

6. The method (500) according to claim 5, wherein the method (500) comprises:57 generating (560), by the control unit (130), the output image by superimposing the isolated-contrast image in color on the no-contrast image in black-and-white.

7. The method (500) according to any claim from 1 to 6, wherein the method (500) comprises: exporting (564-566), by the control unit, one or more training sets to a training computing system, each of the training sets comprising the high-contrast image and the reduced-contrast images resulting from a corresponding iteration of said controlling (510-512;520-522,532;534-536, 544-548) the MRI scanner (120) to acquire the high-contrast image and said controlling (514-518;524-532;538-548) the MRI scanner to acquire the reduced-contrast images, respectively, wherein the training sets are for use by the training computing system to train a machine learning model to optimize a capability thereof to generate the high-contrast image from the reduced- contrast images of each training set.

8. The method (500) according to any claim from 1 to 7, wherein the high- contrast acquisition protocol is defined by an acquisition method and corresponding high-contrast values of one or more acquisition parameters and each of the reduced- contrast acquisition protocols is defined by the acquisition method and corresponding reduced-contrast values of the acquisition parameters.

9. The method (500) according to claim 8, wherein the high-contrast acquisition protocol is Tl-weigthed and the reduced-contrast acquisition protocols are less T1 -weighted than the high-contrast acquisition protocol, and wherein the acquisition method is gradient echo, the reduced-contrast acquisition protocol providing the total flattening having a repetition time equal to the repetition time of the high-contrast acquisition protocol and a flip angle lower than the flip angle of the high-contrast acquisition protocol and the reduced-contrast acquisition protocol providing the partial flattening having the repetition time lower than the repetition time of the high-contrast acquisition protocol and the flip angle lower than the flip angle of the high-contrast acquisition protocol, the flip angle of the reduced-contrast acquisition protocol providing the partial flattening being higher than the flip angle of the reduced- contrast acquisition protocol providing the total flattening, or58 the acquisition method is spin echo, the reduced-contrast acquisition protocols having the repetition time higher than the repetition time of the high-contrast acquisition protocol, the repetition time of the reduced-contrast acquisition protocol providing the partial flattening being lower than the repetition time of the reduced- contrast acquisition protocol providing the total flattening.

10. The method (500) according to any claim from 1 to 9, wherein the method (500) comprises: normalizing (552), by the control unit (130), the representation of the locations of the body-part without the contrast agent in the high-contrast image and in the reduced-contrast images, particularly by scaling each reduced-contrast image according to a corresponding scaling factor depending on a ratio between a high- contrast descriptor of the representation of the locations of the body-part without the contrast agent in the high-contrast image and a corresponding reduced-contrast descriptor of the representation of the locations of the body-part without the contrast agent in the reduced-contrast image.

11. The method (500) according to any claim from 1 to 10, wherein the method (500) comprises: controlling (510-518), by the control unit (130), the MRI scanner (120) to acquire the high-contrast image and the reduced-contrast images in direct succession.

12. The method (500) according to any claim from 1 to 10, wherein the method (500) comprises: controlling (520-530), by the control unit (130), the MRI scanner (120) to acquire a plurality of image portions comprising a plurality of high-contrast portions of the high-contrast image and a plurality of reduced-contrast portions of each of the reduced-contrast images, the high-contrast portions and the reduced-contrast portions being acquired in direct succession interleaved to each other, and combining (532), by the control unit (130), the high-contrast portions into the high-contrast image and the corresponding reduced-contrast portions into each reduced-contrast image.

13. The method (500) according to any claim from 1 to 10, wherein the method(500) comprises: controlling (534-546), by the control unit (130), the MRI scanner (120) to acquire a plurality of image portions comprising one or more high-contrast portions of the high-contrast image, one or more reduced-contrast portions of each of the reduced- contrast images and one or more common portions of a common image of the bodypart comprising the high-amount of the contrast agent, wherein the common portions are acquired with a common acquisition protocol based on the high-contrast acquisition protocol and / or the reduced-contrast acquisition protocols, the high- contrast portions, the reduced-contrast portions and the common portions being acquired in direct succession interleaved to each other, and generating (548), by the control unit (130), the high-contrast image from the high-contrast portions and the common portions and each reduced-contrast image from the corresponding reduced-contrast portions and the common portions.

14. The method (500) according to claim 13, wherein the high-contrast portions and the reduced-contrast portions cover a center of a k-space corresponding to phases extending in absolute value between zero and an intermediate value, and wherein the common portions cover a periphery of the k-space corresponding to phases extending in absolute value between the intermediate value and a maximum value higher than the intermediate value.

15. The method (500) according to claim 13 or 14, wherein the method (500) comprises: controlling (544-546), by the control unit (130), the MRI scanner (120) to acquire the common portions with the common acquisition protocol equal to the reduced-contrast acquisition protocol providing the total flattening or with the common acquisition protocol changing from one of the reduced-contrast acquisition protocols, particularly the reduced-contrast acquisition protocol providing the total flattening, to the high-contrast acquisition protocol.

16. The method (500) according to any claim from 12 to 15, wherein the method (500) comprises: controlling (520-530;534-546), by the control unit (130), the MRI scanner(120) to acquire each of the image portions comprising a plurality of k-space rows for phases ranging in absolute value between a corresponding lower limit and a corresponding upper limit, at least one of the image portions following a first one of the image portions being acquired for phases decreasing in absolute value from the corresponding upper limit to the corresponding lower limit.

17. A computer program (400) configured for causing an MRI system (120,130) comprising an MRI scanner (120) and a control unit (130) to perform the method (500) of any claim from 1 to 16 when the computer program (400) is executed on the control unit (130).

18. A computer program product, the computer program product comprising one or more non-transitory computer readable storage media having program instructions collectively stored in the readable storage media, the program instructions readable by a control unit of an MRI system to cause the control unit to perform the method of claim 1.

19. An MRI system (120,130) comprising an MRI scanner (120) and a control unit (130) configured for performing the steps of the method (500) according to any claim from 1 to 16.

20. An MRI system comprising an MRI scanner and a control unit, the control unit comprising a circuit for performing each step of the method according to claim 1.

21. A medical method comprising: imaging a body -part of a patient with the method of claim 1, and performing a medical procedure relating to the body-part according to the output image.

22. A method (500) for training a machine learning model to simulate an increased contrast-enhancement in medical imaging applications, wherein the method (500) comprises, under the control of a training computing system: receiving (570), by the training computing system, a plurality of training sets, each of the training sets comprising the high-contrast image and the reduced-contrast images resulting from a corresponding iteration of said controlling (510-512;520- 522, 532;534-536, 544-548) the MRI scanner (120) to acquire the high-contrast imageand said controlling (514-518;524-532;538-548) the MRI scanner to acquire the reduced-contrast images, respectively, of any claim from 1 to 16, and training (572), by the training computing system, the machine learning model to optimize a capability thereof to generate the high-contrast image from the reduced- contrast images of each of the training sets.

Images