Machine learning in the field of contrast radiology

Abstract

The present invention relates to the technical field of generating artificial contrast radiological images by machine learning methods.

Description

Technical Field

[0001] The present invention relates to the technical field of generating artificial contrast radiographic images by machine learning methods.

Background Art

[0002] Radiology is a medical field dealing with image processing for diagnostic and therapeutic purposes.

[0003] X-rays and X-ray sensitive films were previously mainly used in medical images, but today's radiology includes various different imaging methods such as computed tomography (CT), magnetic resonance imaging (MRI) or ultrasound examination.

[0004] Using all these methods, substances that facilitate the depiction or demarcation of specific structures within the examination object can be used. The said substances are called contrast agents.

[0005] In computed tomography, iodine-containing solutions are usually used as contrast agents. In magnetic resonance imaging (MRI), superparamagnetic substances (e.g., iron oxide nanoparticles, superparamagnetic iron platinum particles (SIPP)) or paramagnetic substances (e.g., gadolinium chelates, manganese chelates) are usually used as contrast agents.

[0006] Based on the pattern of spread within tissues, contrast agents can be broadly classified into the following categories: extracellular contrast agents, intracellular contrast agents, and intravascular contrast agents.

[0007] Examples of extracellular MRI contrast agents include, for example, gadolinium chelates gadobutrol (Gadovist (registered trademark)), gadoteridol (Prohance (registered trademark)), gadoteric acid (Dotarem (registered trademark)), gadopentetic acid (Magnevist (registered trademark)) and gadodiamide (Omnican (registered trademark)). The highly hydrophilic nature of the gadolinium chelates and their low molecular weight result in rapid diffusion into the interstitial space after intravenous administration. After a certain relatively short period of circulation in the bloodstream, they are excreted via the kidneys.

[0008] Intracellular contrast agents are taken up to some extent into the cells of tissues and subsequently excreted.

[0009] Intracellular MRI contrast agents based on gadoxetic acid are distinguished, for example, by the fact that they are proportionally and specifically taken up by hepatocytes (liver cells), accumulate in functional tissue (soft tissue), and enhance the contrast in healthy liver tissue before being excreted in the feces via the bile. Examples of such contrast agents based on gadoxetic acid are described in US6,039,931A; they are commercially available, for example, under the trade names Primovist (registered trademark) or Eovist (registered trademark). Furthermore, an MRI contrast agent with low uptake by hepatocytes is gadobenate dimeglumine (Multihance (registered trademark)).

[0010] The contrast enhancement effect of Primovist (registered trademark) / Eovist (registered trademark) is mediated by the stable gadolinium complex Gd-EOB-DTPA (gadolinium ethoxybenzyl diethylenetriaminepentaacetic acid). DTPA forms a complex with paramagnetic gadolinium ions with very high thermodynamic stability. The ethoxybenzyl (EOB) radical is a mediator of hepatic biliary uptake of the contrast agent.

[0011] Intravascular contrast agents are distinguished from extracellular contrast agents by their significantly longer residence time in the bloodstream. Gadofosveset, for example, is an intravascular MRI contrast agent based on gadolinium. It is used in trisodium salt monohydrate form (Ablavar®). It binds to serum albumin, thereby achieving a long residence time of the contrast agent in the bloodstream (half-life in the blood of approximately 17 hours).

[0012] Numerous radiological examinations involve administering a contrast agent to a patient, and tracking its dynamic spread within the body using imaging techniques. Examples include the detection and differential diagnosis of localized liver lesions using dynamic contrast-enhanced magnetic resonance imaging (DMU).

[0013] Primovist® can be used to detect tumors in the liver. While blood supply to healthy liver tissue is primarily achieved via the portal vein, the hepatic artery supplies most primary tumors. Therefore, after intravenous injection of a bolus of contrast agent, it is possible to observe a time delay between the signal elevation of healthy liver parenchyma and the signal elevation of tumors.

[0014] In addition to malignant tumors, benign lesions such as cysts, hemangiomas, and focal nodular hyperplasia (FNH) are frequently found in the liver. Proper treatment planning requires differentiating these from malignant tumors. Primovist® can be used to identify benign and malignant focal liver lesions. T1-weighted MRI provides information about the characteristics of these lesions. Differentiation is achieved by utilizing different blood supply to the liver and tumors, as well as temporal profiles of contrast enhancement.

[0015] The contrast enhancement achieved by Primovist® can be divided into at least two phases: a dynamic phase (including the so-called arterial, portal, and late phases) and a hepatobiliary phase, where significant uptake of Primovist® into hepatocytes has already occurred.

[0016] In cases of contrast enhancement achieved by Primovist® during the distribution phase, what is observed is a typical perfusion pattern that provides information for characterizing the lesion. Depicting angiogenesis helps to characterize the type of lesion and determine the spatial relationship between the tumor and the blood vessels.

[0017] In T1-weighted MRI images, Primovist® elicits a clear signal enhancement in healthy liver parenchyma 10–20 minutes after injection (hepatobiliary phase), while lesions containing no hepatocytes or only a few hepatocytes, such as metastases or moderately to poorly differentiated hepatocellular carcinoma (HCC), appear as darker areas. [Overview of the Initiative] [Problems that the invention aims to solve]

[0018] Therefore, tracking the temporal diffusion of contrast agent across the dynamic and hepatobiliary phases offers a good possibility for detecting and differentiating localized liver lesions, although the examination spans a relatively long time period. To minimize motion artifacts in MRI images, patient movement should be largely avoided throughout the aforementioned time span. Prolonged restriction of movement can be uncomfortable for the patient.

[0019] This problem and further problems are solved by the present invention. The present invention provides means that enable the synthetic generation of one or more radiographic images. The synthetically generated radiographic images are predicted by a machine learning model based on a time series of measured radiographic images showing contrast enhancement that changes over time. It has the advantage that radiographic examinations can be made faster because it is not necessary to measure all radiographic images that are important for diagnosis; one or more radiographic images can be predicted (calculated) based on the measured radiographic images; and therefore examination time can be reduced. Furthermore, the prediction of radiographic images is performed in frequency space (not in real space as is usually the case). As a result, it is possible to separate contrast information from detailed information in the radiological representation of the examination area, limit training and prediction to contrast information, and then reintroduce detailed information after prediction. This procedure reduces computational complexity, for example. Furthermore, working in frequency space means a higher tolerance for image registration defects. [Means for solving the problem]

[0020] In its first embodiment, the present invention - A step of receiving a plurality of first representations of the examination region to be examined in frequency space, wherein at least some of the first representations represent the examination region during a first time span after administration of contrast agent, - A step of supplying a prediction model with multiple first representations, wherein the prediction model is trained to generate one or more second reference representations representing examination areas during a second time span in frequency space from first reference representations, at least some of which represent examination areas during a first time span in frequency space after administration of contrast agent, based on first and second reference representations of examination areas of a number of examination subjects, - A step of receiving one or more predicted representations of the test region in frequency space from a prediction model, where one or more predicted representations represent the test region during a second time span. - A step of converting one or more predicted representations into one or more representations of the inspection domain in real space. - A step of outputting and / or storing one or more representations of the inspection region in real space. This provides a computer implementation method that includes [the following].

[0021] The present invention further, • Receiving unit and, • Control and computing units, and • Output unit and Equipped with, Here, the control and computing units are: - Prompting the receiving unit to receive multiple first representations of the examination area under examination in frequency space, wherein at least some of the first representations represent the examination area during a first time span after the administration of the contrast agent, - To supply the predictive model with multiple first representations, the predictive model is trained to generate one or more second reference representations representing the examination areas during a second time span in frequency space from first reference representations, at least some of which represent the examination areas during a first time span in frequency space after the administration of contrast agent, based on first and second reference representations of a number of examination areas under examination, - Receive one or more predicted representations of the inspection region in frequency space from the prediction model, where one or more predicted representations represent the inspection region during a second time span. - To convert one or more predicted representations into one or more representations of the inspection domain in real space, - Prompt the output unit to output and / or store one or more representations of the inspection area in real space. Provides a configured system.

[0022] The present invention further provides a computer program product comprising a computer program that can be loaded into the memory of a computer system, wherein it loads the following into the computer system: - A step of receiving a plurality of first representations of the examination region to be examined in frequency space, wherein at least some of the first representations represent the examination region during a first time span after administration of contrast agent, - A step of supplying a prediction model with multiple first representations, wherein the prediction model is trained to generate one or more second reference representations representing examination areas during a second time span in frequency space from first reference representations, at least some of which represent examination areas during a first time span in frequency space after administration of contrast agent, based on first and second reference representations of examination areas of a number of examination subjects, - A step of receiving one or more predicted representations of the test region in frequency space from a prediction model, where one or more predicted representations represent the test region during a second time span. - A step of converting one or more predicted representations into one or more representations of the inspection domain in real space. - A step of outputting and / or storing one or more representations of the inspection region in real space. Prompt them to do it.

[0023] The present invention further provides the use of a contrast agent in a method for predicting at least one radiographic image, wherein the method is as follows: - A step of generating a plurality of first representations of the examination region under examination in frequency space, wherein at least some of the first representations represent the examination region during a first time span after administration of contrast agent, - A step of supplying a prediction model with multiple first representations, wherein the prediction model is trained to generate one or more second reference representations representing examination areas during a second time span in frequency space from first reference representations, at least some of which represent examination areas during a first time span in frequency space after administration of contrast agent, based on first and second reference representations of examination areas of a number of examination subjects, - A step of receiving one or more predicted representations of the test region in frequency space from a prediction model, where one or more predicted representations represent the test region during a second time span. - A step of converting one or more predicted representations into one or more representations of the inspection domain in real space. - A step of outputting and / or storing one or more representations of the inspection region in real space. Includes.

[0024] Further provided are contrast agents for use in a method for predicting at least one radiographic image, wherein the method is as follows: - The step of administering the contrast agent, where the contrast agent spreads within the examination area being examined. - A step of generating a plurality of first representations of the examination region under examination in frequency space, wherein at least some of the first representations represent the examination region during a first time span after administration of contrast agent, - A step of supplying a prediction model with multiple first representations, wherein the prediction model is trained to generate one or more second reference representations representing examination areas during a second time span in frequency space from first reference representations, at least some of which represent examination areas during a first time span in frequency space after administration of contrast agent, based on first and second reference representations of examination areas of a number of examination subjects, - A step of receiving one or more predicted representations of the test region in frequency space from a prediction model, where one or more predicted representations represent the test region during a second time span. - A step of converting one or more predicted representations into one or more representations of the inspection domain in real space. - A step of outputting and / or storing one or more representations of the inspection region in real space. Includes.

[0025] Furthermore, a kit comprising a contrast agent and a computer program product according to the present invention is provided.

[0026] The present invention will be described in more detail below without distinguishing between the subject matter of the invention (methods, systems, computer program products, uses, contrast agents for use, kits). Rather, the following description is intended to apply equally to all subject matter of the invention, regardless of the context in which it arises (methods, systems, computer program products, uses, contrast agents for use, kits).

[0027] Where steps are described in order herein or in the claims, this does not necessarily mean that the invention is limited to the order in which they are described. Rather, unless one step is built upon another, this does not necessarily mean that the steps built upon the previous step are performed consecutively (although this will become apparent in the individual cases), but the steps may also be performed in a different order or in parallel with one another. Thus, the order in which they are described is a preferred embodiment of the invention.

[0028] This invention can shorten the time span of radiation examinations of the subject being examined.

[0029] The "subject of examination" is usually a living organism, preferably a mammal, and very preferably a human.

[0030] The "examination area" is typically a part of the object being examined, such as an organ or a part of an organ. Also called the image volume (field of view, FOV), the examination area is specifically the volume captured in a radiographic image. The examination area is typically defined by the radiologist, for example, on a general image (localizer). Of course, the examination area can also be defined alternatively or additionally, for example, automatically based on a selected protocol.

[0031] The term “radiological examination” is understood to mean all imaging methods that enable insight into an object being examined using electromagnetic waves, particle radiation, or mechanical waves for diagnostic, therapeutic, and / or scientific purposes. In the context of this invention, the term “radiology” specifically includes the following examination methods: computed tomography, magnetic resonance imaging, ultrasound, positron emission tomography, echocardiography, and scintigraphy.

[0032] In a preferred embodiment of the present invention, the radiological examination is magnetic resonance imaging.

[0033] Magnetic resonance imaging (MRI) is an imaging technique specifically used in medical diagnosis to depict the structure and function of tissues and organs within the body of a human or animal.

[0034] In MRI, the magnetic moments of protons in the object being examined are aligned in a fundamental magnetic field, resulting in macroscopic magnetization along the longitudinal direction. This is then deflected from its resting position by irradiation with a radio frequency (HF) pulse (excitation). Subsequently, the return from the excited state to the resting position (relaxation), i.e., the dynamics of the magnetization, is detected as a relaxation signal by one or more HF receiving coils.

[0035] For spatial coding, rapidly switched gradient magnetic fields are superimposed on the fundamental magnetic field. The captured relaxation signal, or detected and spatially resolved MRI data, initially exists as raw data in frequency space and can be transformed into real space (image space), for example, by a subsequent inverse Fourier transform.

[0036] In native MRI, tissue contrast is generated by different relaxation times (T1 and T2) and proton densities. T1 relaxation represents the transition of the longitudinal magnetization to its equilibrium state, where T1 is the time it takes to reach 63.21% of the equilibrium magnetization before resonant excitation. This is also called the longitudinal relaxation time or spin-lattice relaxation time. T2 relaxation similarly represents the transition of the transverse magnetization to its equilibrium state.

[0037] In radiological examinations, contrast agents are commonly used to enhance contrast.

[0038] A "contrast agent" is a substance or mixture of substances that improves the depiction of the structure and function of the body in imaging methods such as X-ray diagnosis, magnetic resonance imaging, and ultrasound.

[0039] Examples of contrast agents can be found in the literature (for example, ASL Jascinth et al.: Contrast Agents in computed tomography: A Review, Journal of Applied Dental and Medical Sciences, 2016, Vol.2, Issue 2, 143-149; H. Lusic et al.: X-ray-Computed Tomography Contrast Agents, Chem. Rev. 2013, 113, 3, 1641-1666; https: / / www.radiology.wisc.edu / wp-content / uploads / 2017 / 10 / contrast-agents-tutorial.pdf, MRNough et al.: Radiographic and magnetic resonances contrast agents: Essentials and tips for safe practices, World J Radiol. 2017 Sep 28; 9(9): 339-349; LCAbonyi et al.: Intravascular Contrast Media in Radiography: Historical Development & Review of Risk Factors for Adverse Reactions, South American Journal of Clinical Research, 2016, Vol.3, Issue1, 1-10; ACR Manual on Contrast Media, 2020, ISBN:978-1-55903-012-0; A. Ignee et al.: Ultrasound contrast agents, Endosc Ultrasound. 2016 Nov-Dec;5(6):355-362 (see reference).

[0040] MRI contrast agents exert their effects by altering the relaxation time of structures that take up the contrast agent. Two groups can be distinguished: paramagnetic and superparamagnetic materials. Both groups of materials have unpaired electrons that induce a magnetic field around individual atoms or molecules. Superparamagnetic contrast agents primarily shorten T2, while paramagnetic contrast agents primarily shorten T1. The effect of the contrast agent is indirect, as the contrast agent itself does not emit a signal but merely affects the intensity of signals in its vicinity. An example of a superparamagnetic contrast agent is iron oxide nanoparticles (SPIO, superparamagnetic iron oxide). Examples of paramagnetic contrast agents include gadopentetate dimeglumine (trade name: Magnevist®, etc.), gadoteric acid (Dotarem®, Dotagita®, Cyclolux®), gadodiamide (Omniscan®), gadoteridol (ProHance®), and gadobutrol (Gadovist®), among other gadolinium chelates.

[0041] Using the present invention, it is possible to predict one or more composite radiographic images of an examination area. The prediction is performed using a prediction model.

[0042] The predictive model is a computer-aided model configured to predict one or more second representations of the examination region in frequency space based on multiple first representations of the examination region in frequency space. At least some of the multiple first representations represent the examination region during a first time span after the administration of the contrast agent. At least one second representation represents the examination region during a second time span.

[0043] The word "plural" means at least two numbers. The first plural expression used in predictions is usually 10 or less.

[0044] The second time span can come before or after the first time span. It is also possible that the time spans at least partially overlap, or that one time span falls within the other.

[0045] Figures 1(a), 1(b), 1(c), and 1(d) are for illustrative purposes only. Figures 1(a), 1(b), 1(c), and 1(d) each show a timeline. Defined points in time are marked on the timeline. Time t0 characterizes the point in time when the contrast agent is administered to the subject. The dashed frame indicates the time span during which the subject undergoes radiographic examination, i.e., the time the subject spends in, for example, a magnetic resonance imaging system or computed tomography system.

[0046] Figure 1(a) schematically shows a typical profile of a radiographic examination. The subject of examination is introduced into the tomography machine. Time t -1 In this case, the object being examined is located inside the tomography device. -1 At this point in time (t -1 At time point t0, the contrast agent has not yet been administered to the object being examined; that is, its representation is a contrast-free (native) representation. At time point t0, the contrast agent is administered to the object being examined, which is located inside the tomography scanner. At times t1, t2, and t3, further representations of the object being examined are generated. Subsequently, the object is removed from the tomography scanner. The object is then examined over a relatively long time span T. t It was located within the tomography apparatus. During this time, four representations of the examination area were generated based on the measurements.

[0047] Figure 1(b) schematically shows the profile of the radiographic examination according to the present invention. The object to be examined is introduced into the tomography apparatus. Time t -1 In this case, the object being examined is located inside the tomography device. -1 At this point in time (t -1In (0), the contrast agent has not yet been administered to the examination subject, that is, its representation is a (native) representation without the contrast agent. At time point t0, the contrast agent is administered to the examination subject located within the tomographic imaging device. At time points t1 and t2, further representations of the examination subject are generated. Thereafter, the examination subject leaves the tomographic imaging device. The representation generated based on the measurement at time point t3 in the case of FIG. 1(a) is predicted (calculated) in the case of the profile shown in FIG. 1(b). The examination subject is located within the tomographic imaging device during the time span T a and the time span T a is shorter than the time span T t in FIG. 1(a). Therefore, in the case of FIG. 1(b), the radiation examination does not last as long as in the case of FIG. 1(a) and is more comfortable for the examination subject; what is generated in both cases is a representation of the examination area representing the examination area at the time points t -1 , t1, t2 and t3. The representations generated based on the measurements at time points t -1 , t1 and t2 are the first representations of the examination area of the examination subject in the context of the present invention, where at least some represent the examination area during the first time span after administration of the contrast agent, that is, the representations at time points t1, t2 and t3; the representation at time point t -1 represents the examination area during the time span before administration of the contrast agent. The representation at time point t3 is predicted by the prediction model according to the present invention based on the representations t -1 , t1 and t2. This represents the examination area during the second time span, which comes after the first time span in the case of FIG. 1(b).

[0048] Figure 1(c) schematically shows a further profile of the radiological examination according to the present invention. The contrast agent is administered to the subject at time t0. At time t0, the subject is not yet located inside the tomography scanner. The subject is introduced into the tomography scanner only after the contrast agent has been administered. At times t1, t2, and t3 (in the first time span after contrast agent administration), representations of the examination area of ​​the subject are generated. These represent the examination area in the first time span after contrast agent administration. After the generation of the representation at time t3, the subject can leave the tomography scanner and the radiological examination is complete. From the representations at times t1, t2, and t3, at time t -1 The representation at time t can be predicted. -1 The expression in represents the inspection area during the second time span, and the second time span precedes the first time span. The object of inspection is the time span T b During the time span T, the device is located inside the tomography apparatus. b This is the time span T in Figure 1(a). t Shorter than . Therefore, in the case of Figure 1(c), the radiation examination does not last as long as in the case of Figure 1(a), and is more comfortable for the person being examined, but what is produced in both cases is t -1 This is a representation of the inspection area, showing the inspection area at time points t1, t2, and t3.

[0049] Figure 1(d) schematically shows a further profile of the radiographic examination according to the present invention. This embodiment is intended to clarify that the present invention is not limited to individual administrations of contrast agents. For example, two or more administrations are possible. In the case of individual administrations, it is not necessary to administer the same contrast agent, and instead, different contrast agents may be administered. The (first) contrast agent is administered to the object under examination at time t0. At time (t0), the object under examination is not yet located in the tomography apparatus. The object under examination is introduced into the tomography apparatus only after the contrast agent has been administered. At time t2, a first representation of the examination area of ​​the object under examination is generated. The representation at time t2 represents the examination area in a first time span after the administration of the (first) contrast agent. At time t3, the (second) contrast agent is administered. At time (t3), the object under examination is located in the tomography apparatus. After the administration of the (second) contrast agent, two further representations of the examination area are generated, one at time t4 and the other at time t5. The representations generated at times t2, t4, and t5 represent the examination area during the first time span after the administration of the contrast agent. These representations are at time t -1 It can be used to predict the representation of the test region at time t1 and / or the representation of the test region at time t1. -1 The representation of the inspection region at t1 represents the inspection region during a second time span, the second time span preceding the first time span.

[0050] The profiles shown in Figures 1(b), 1(c), and 1(d) can also be combined with further profiles / variations.

[0051] In a preferred embodiment of the present invention, the first time span begins before or with the administration of the contrast agent. It is advantageous if one or more representations of the examination area showing the examination area without contrast (native image) are generated, because the radiologist can already obtain important information about the health status of the subject from such images. For example, the radiologist can identify bleeding in the native MRI image.

[0052] In order for the prediction model according to the present invention to be able to make the predictions described herein, it must be properly configured beforehand.

[0053] Here, the term “prediction” means that at least one representation of the examination region representing the examination region during a second time span in frequency space is calculated using a plurality of first representations of the examination region in frequency space, where at least some of the plurality of first representations represent the examination region during a first time span after the administration of the contrast agent.

[0054] The predictive model is preferably constructed (built, trained) using a self-learning algorithm in a supervised machine learning process. The training data is used for training. The training data includes multiple representations of the test domain for each test subject of a multitude of test subjects. The test domain is usually the same for all test subjects (e.g., a part of the human body, an organ, or a part of an organ). Representations in the training dataset are also referred to herein as reference representations. The term “multitude” preferably means more than 10, and more preferably more than 100.

[0055] For each examination subject, the training data includes i) a plurality of first reference representations of the examination region in frequency space, at least some of which represent the examination region during a first time span after administration of the contrast agent, and ii) one or more second reference representations of the examination region in frequency space representing the examination region during a second time span.

[0056] The predictive model is trained to predict (calculate) one or more second reference representations from multiple first reference representations for each test subject.

[0057] Self-learning algorithms generate statistical models based on training data during machine learning. This means that the algorithm "recognizes" patterns and regularities in the training data, rather than simply memorizing examples. Therefore, predictive models can also evaluate unknown data. Validation data can be used to test the quality of evaluation of unknown data.

[0058] The predictive model can be trained by supervised learning, where pairs of datasets (first and second representations) are presented to the algorithm, and the algorithm then learns the relationship between the first and second representations.

[0059] Self-learning systems trained through supervised learning are widely described in prior art (see, for example, C. Perez: Machine Learning Techniques: Supervised Learning and Classification, Amazon Digital Services LLC-Kdp Print Us, 2019, ISBN 1096996545, 9781096996545).

[0060] Preferably, the predictive model is an artificial neural network or comprises such a network.

[0061] An artificial neural network comprises at least three layers of processing elements: a first layer having input neurons (nodes), an Nth layer having at least one output neuron (node), and N-2 hidden layers, where N is a natural number greater than 2.

[0062] Input neurons receive the first representation. Output neurons output one or more second representations for multiple first representations.

[0063] The processing elements in the layer between the input neuron and the output neuron are connected to each other in a predetermined pattern having predetermined connection weights.

[0064] Preferably, the artificial neural network is a so-called convolutional neural network (CNN).

[0065] Convolutional neural networks can process input data in the form of a matrix. A CNN essentially consists of filters (convolutional layers) and aggregation layers (pooling layers), which are repeated alternately, and finally, it consists of one or more "normal" fully connected neurons (dense / fully connected layers).

[0066] Neural network training can be performed, for example, by the backpropagation method. The objective of the network here is to achieve the greatest reliability in mapping given input data to given output data. The mapping quality is described by the loss function. The objective is to minimize the loss function. In the case of the backpropagation method, the artificial neural network is learned by changing the connection weights.

[0067] In a trained state, the connection weights between processing elements include information about the relationship between a first representation and one or more second representations, which can be used to predict one or more second representations that indicate the examination region during a second time span for a number of new first representations (e.g., new objects of examination), at least some of which indicate the examination region during the first time span after the administration of the contrast agent.

[0068] Cross-validation can be used to split the data into training and validation datasets. The training dataset is used for backpropagation training of the network weights. The validation dataset is used to check the accuracy of predictions that the trained network can make when applied to unknown data.

[0069] Further details regarding the construction and training of artificial neural networks can be found in prior art (see, for example, S. Khan et al.: A Guide to Convolutional Neural Networks for Computer Vision, Morgan & Claypool Publishers 2018, ISBN 1681730227, 9781681730226, WO2018 / 183044A1, WO2018 / 200493, WO2019 / 074938A1, WO2019 / 204406A1, WO2019 / 241659A1).

[0070] Preferably, the predictive model is a Generative Adversarial Network (GAN) (see, for example, http: / / 3dgan.csail.mit.edu / ).

[0071] In addition to the representation, further information about the object being examined, the area being examined, the examination conditions, and / or the radiographic examination can also be used for training, validation, and prediction.

[0072] Examples of information about the subject of the examination include sex, age, weight, height, medical history, nature, duration, and amount of medications already taken, blood pressure, central venous pressure, respiratory rate, serum albumin, total bilirubin, blood glucose, iron content, and respiratory capacity. This information can be read, for example, from a database or electronic patient file.

[0073] Examples of information related to the examination area include: existing condition, surgery, partial resection, liver transplant, iron liver, fatty liver, etc.

[0074] As already explained, the representation of the test domain used for training, validation, and prediction is the representation of the test domain in frequency space (also called spatial frequency space, Fourier space, frequency domain, or Fourier representation).

[0075] In magnetic resonance imaging, raw data is typically generated as so-called k-space data by the measurement method described above. This k-space data is a description of the test domain in frequency space, and that is, such k-space data can be used for training, validation, and prediction. If a representation in real space exists, such a representation in real space can be converted to a representation in frequency space by a Fourier transform; conversely, a representation in frequency space can be converted to a representation in real space by, for example, an inverse Fourier transform.

[0076] Therefore, if the radiographic image of the inspection area exists in the form of a two-dimensional image in real space, this representation of the inspection area can be transformed into a two-dimensional representation of the inspection area in frequency space by a 2D Fourier transform.

[0077] The 3D image (volume plot) of the inspection area can be treated as a stack of 2D images. Furthermore, it is conceivable to convert the 3D image into a 3D representation of the inspection area in frequency space using a 3D Fourier transform.

[0078] To convert a real-space representation to a frequency-space representation, it is possible to use transformations other than the Fourier transform. The three main properties that such transformations must satisfy are as follows: a) The existence of a clear inverse transformation (a clear connection between the real-space description and the frequency-space description). b) Locality of contrast information c) Robustness with respect to defective image registration.

[0079] Details of the conversion from one depiction to another are described in numerous textbooks and publications (see, for example, W. Burger, MJBurge: Digital Image Processing: An Algorithmic Introduction Using Java, Texts in Computer Science, 2nd edition, Springer-Verlag, 2016, ISBN: 9781447166849; W. Birkfellner: Applied Medical Image Processing, Second Edition: A Basic Course, Verlag Taylor & Francis, 2014, ISBN: 9781466555570; R. Bracewell: Fourier Analysis and Imaging, Verlag Springer Science & Business Media, 2004, ISBN: 9780306481871).

[0080] Figure 2 (Fig. 2) illustrates and schematically shows the generation of the test domain representation in real space and frequency space.

[0081] Figure 2 shows the timeline. At three different time points t1, t2, and t3, representations of the examination area are generated based on measurements. The examination area is the human lung. At time point t1, a first representation is generated. This could be a representation of the examination area (lung) in real space (O1) or a representation of the examination area (lung) in frequency space (F1). The representation of the examination area in real space (O1) can be transformed into a representation of the examination area in frequency space (F1) by the Fourier transform FT. The representation of the examination area in frequency space (F1) can be transformed back into a representation of the examination area in real space (O1) by the inverse Fourier transform iFT. The two representations (O1) and (F1) contain the same information about the examination area, differing only in their representation. At time point t2, a further representation is generated. This could be a representation of the examination area (lung) in real space (O2) or a representation of the examination area (lung) in frequency space (F2). The representation of the examination region in real space (O2) can be transformed into a representation of the examination region in frequency space (F2) by the Fourier transform (FT). The representation of the examination region in frequency space (F2) can be transformed back into a representation of the examination region in real space (O2) by the inverse Fourier transform (iFT). The two representations (O2) and (F2) contain the same information about the examination region, differing only in their representation. At time t3, a further representation is generated. This could be a representation of the examination region (lung) in real space (O3) or a representation of the examination region (lung) in frequency space (F3). The representation of the examination region in real space (O3) can be transformed into a representation of the examination region in frequency space (F3) by the Fourier transform (FT). The representation of the examination region in frequency space (F3) can be transformed back into a representation of the examination region in real space (O3) by the inverse Fourier transform (iFT). The two representations (O3) and (F3) contain the same information about the examination region, differing only in their representation.

[0082] The representations of the examination area in real space (O1), (O2), and (O3) are familiar to humans and can be immediately grasped by them. Representations (O1), (O2), and (O3) show how the contrast agent dynamically spreads within the vein. The same information is contained in representations (F1), (F2), and (F3), although they are more difficult for humans to grasp.

[0083] Figure 3 (Fig. 3) schematically and illustratively illustrates how the representations of the test domains (F1), (F2), and (F3) in frequency space, as generated in Figure 2, can be used to train a predictive model (PM). The representations (F1), (F2), and (F3) form a set of training data to be tested. Training is performed using a large number of training datasets for a large number of tests.

[0084] Representations (F1) and (F2) are first reference representations of multiple (two in this case) examination domains in frequency space, at least some of which represent the examination domain during a first time span after contrast agent administration. Representation (F3) is a second reference representation of the examination domain in frequency space, representing the examination domain during a second time span. In Figure 3, the prediction model is trained to predict the representation (F3) of the examination domain in frequency space from the representations (F1) and (F2) of the examination domain in frequency space. Representations (F1) and (F2) are input to the prediction model (PM), and the prediction model computes representation (F3*) from representations (F1) and (F2). An asterisk (*) indicates that representation (F3*) is a predicted representation. The computed representation (F3*) is compared to representation (F3). Deviations may be used in backpropagation to train the prediction model to reduce deviations to a defined minimum. If a predictive model has been trained on a large number of training datasets for a large number of test subjects, and the predictions reach a defined accuracy, then the trained predictive model can be used for prediction. This is illustrated in an illustrative and schematic manner in Figure 4.

[0085] Figure 4 (Fig. 4) shows the predictive model (PM) trained in Figure 3. The predictive model is used to generate one or more second representations of the examination area in frequency space based on a plurality of first representations of the examination area in frequency space, at least some of which represent the examination area in a first time span after administration of contrast agent, and the one or more second representations represent the examination area in a second time span.

[0086] In this example, two first representations of the test domain in frequency space (F ~ 1) and (F ~ 2) is input to the prediction model (PM), and the prediction model (PM) is the second representation (F ~ Generate (calculate) 3*). A tilde (~) indicates that the representation is a new representation under test, and usually, the representation does not contain any representations used in the training method for training the predictive model. An asterisk (*) indicates that the representation (F ~ 3*) indicates that this is the predicted representation. Representation of the test domain in frequency space (F ~ 3*) is, for example, the representation of the inspection region in real space by the inverse Fourier transform iFT (O ~ It can be converted to 3*).

[0087] Using a representation of the inspection domain in frequency space is advantageous over using a representation of the inspection domain in real space (also known as image space). When using a representation of the inspection domain in frequency space, contrast information, which is important for training and prediction, can be separated from detailed information (fine structure). This allows for focus on the information to be learned by the predictive model during training, and also on the information to be predicted by the predictive model: contrast information, during prediction.

[0088] In real-space representations of an inspection region, contrast information is typically distributed throughout the representation (each pixel / voxel inherently possesses contrast information), whereas in frequency-space representations of an inspection region, contrast information is encoded within and around the center of the frequency space. In other words, low frequencies in the frequency-space representation are involved in contrast, while high frequencies contain information about fine structure.

[0089] Therefore, it is possible to separate contrast information, limit training and prediction to contrast information, and reintroduce information about fine structures after training / prediction.

[0090] In a preferred embodiment, the method according to the present invention is as follows: - A step of receiving a plurality of first representations of the examination region to be examined in frequency space, wherein at least a portion of the first representations represent the first examination region during a first time span after administration of contrast agent, - A step of identifying a region in the first representation, where the identified region includes the center of the frequency space, - A step of reducing the first representation to a specified region, where multiple reduced first representations are obtained, - A step of supplying multiple reduced first representations to the predictive model, - A step of receiving one or more second representations of the test region in frequency space from the predictive model, where one or more second representations represent the test region during a second time span. - A step of supplementing one or more second representations with one or more regions of a received first representation that are outside the specified region, where one or more supplemented second representations are obtained, - A step of converting one or more supplemented second representations into one or more representations of the examination domain in real space. - A step of outputting and / or storing one or more representations of the inspection region in real space. Includes.

[0091] Identifying a region in a first representation can be achieved, for example, by a user of the computer system according to the present invention inputting one or more parameters into the computer system according to the present invention and / or selecting from a list that defines the shape and / or size of the region. However, identification may also be performed automatically, for example, by a computer system according to the present invention that is appropriately configured to select a predetermined region within a representation of an inspection region.

[0092] The identified region is usually smaller than the frequency space filled by the first representation, but in either case, it includes the center of the frequency space.

[0093] A region of frequency space containing the center of frequency space (also called the origin or zero) contains contrast information relevant to the method according to the present invention. If the identified region is smaller than the frequency space filled by the first representation, the result is lower computational complexity for subsequent predictions (this also applies in particular to training the prediction model). Thus, the choice of region size can have a direct impact on computational complexity.

[0094] In principle, it is also possible to identify a region corresponding to the entire frequency space satisfied by the first representation; in such a case, there is no reduction of the frequency space into sub-regions, and the computational complexity is maximized.

[0095] Therefore, by identifying a region around the center of the frequency space, a user of the computer system according to the present invention can decide for themselves whether they want a complete representation of the inspection region in frequency space to form the basis for training and prediction, or whether they want to reduce computational complexity. Here, the user can directly influence the computational complexity required through the size of the identified region.

[0096] The specified region typically has the same dimensions as the frequency space: in the case of a 2D representation in 2D frequency space, the specified region is typically an area, and in the case of a 3D representation in 3D frequency space, the specified region is typically a volume.

[0097] The specified region can, in principle, have any shape, and therefore, for example, it can be circular and / or angular, concave and / or convex. Preferably, the region is a cuboid or cube if it is in 3D frequency space in Cartesian coordinates, and a rectangle or square if it is in 2D frequency space in Cartesian coordinates. However, it can also be spherical or circular, or have other shapes.

[0098] Preferably, the geometric centroid of the identified region coincides with the center of the frequency space.

[0099] The representations used for training, validation, and prediction are reduced to a specified area. The term "reduction" here means that any portion of the representation that does not fall within the specified range is either cut off (discarded) or masked. In the case of masking, areas outside the specified region are covered by the mask, leaving only the specified region uncovered. When masked, the color value of the corresponding pixel / voxel can be set to, for example, zero (black).

[0100] The expression obtained in this manner is also referred to as the reduced expression in this specification.

[0101] The first representation obtained after reduction (the reduced first representation) is supplied to the predictive model: the predictive model is pre-trained in a training method for learning the dynamic effect of the amount of contrast agent on the representation of the examination domain in frequency space. Training similarly preferably utilizes the reduced representations (the reduced first representation and the reduced second representation).

[0102] Thus, the predictive model learns how the contrast agent has a dynamic effect on the representation of the examination area and can apply this learned “knowledge” to predict one or more (reduced) second representations based on the (reduced) first representation.

[0103] One or more predicted second representations represent the examination region in frequency space during the second time span.

[0104] One or more predicted second representations are calculated and output by the prediction model based on the (reduced) first representation.

[0105] If at least one predicted second representation is generated based on the reduced first representation, it is preferable to re-add previously discarded portions (cut out or masked) so as to avoid losing as much information about the fine structure in the final artificially generated image.

[0106] To reuse previously discarded (cut out or masked) portions, at least one predicted second representation may be superimposed on at least one received first representation such that at least one predicted second representation replaces the corresponding superimposed frequency domain of at least one initially received first representation. Preferably, the predicted second representation replaces the corresponding frequency domain of the initially received first representation representing the examination area without contrast.

[0107] The superimposed frequency domain substitution corresponds to the supplementation by one or more regions of the frequency space of the received first representation that were omitted when the first representation was reduced to a specified region.

[0108] In other words: the frequency space of at least one predicted second representation in the examination domain is filled by the domain of at least one first received first representation, thereby at least one first received first representation being larger than the predicted second representation.

[0109] Therefore, by using a representation of the inspection domain in frequency space, it is possible to separate contrast information from detailed information, limit training and prediction to contrast information, and then reintroduce detailed information after training and / or prediction. As already explained, this procedure reduces the computational complexity between training, validation, and prediction.

[0110] However, working in frequency space has yet another advantage over working in real space: in frequency space, the colregistration of individual representations is not as critical as in real space. "Colregistration" (also called "image registration" in prior art) is a crucial process in digital image processing that works to harmonize two or more images of the same scene, or at least similar scenes, with each other in the best possible way. One of the images is defined as the reference image, and the other is called the target image. A compensatory transformation is calculated to optimally match the target image with the reference image. The images to be aligned are different from each other because they were acquired from different locations, at different times, and / or using different sensors.

[0111] In the present invention, each of the multiple first representations of the examination area is, firstly, generated at different time points; and secondly, differs in terms of the content and diffusion of the contrast agent in the examination area.

[0112] Therefore, using a representation of the inspection domain in frequency space has advantages over using a representation of the inspection domain in real space, and training, validation, and prediction methods are more forgiving with respect to errors in image alignment. In other words, if the representations in frequency space are not precisely superimposed, this has less impact than if the representations in real space are not superimposed with pixel / voxel accuracy. This is due to the properties of the Fourier transform: as already explained, the contrast information of a Fourier-transformed image is always mapped near the origin (center) of Fourier space. Turns or rotations in image space (real space) result in image information (e.g., visible structure) being localized to different regions of the transformed image. However, in Fourier space, these transforms do not change the region in which the contrast information relevant to this invention is encoded.

[0113] Figure 5 (Fig. 5) illustrates and schematically shows the steps in training a predictive model according to a preferred embodiment of the present invention.

[0114] What is received are two first representations (F1) and (F2) of the object under examination in frequency space, and a second representation (F3) of the object under examination in frequency space. In representations (F1), (F2), and (F3), the same region A is identified. Region A contains the center of frequency space, in this case having a square, the geometric centroid of the square coincides with the center of frequency space. Representations (F1), (F2), and (F3) are reduced to the identified region A, respectively: three reduced representations (F1 red ), (F2 red ) and (F3 red The reduced representation is used for training. The predictive model uses the reduced representation (F1 red ) and (F2) reduced representation (F3 red Trained to predict ). Reduced representation (F1 red ) and (F2 red ) is fed to the predictive model (PM), and the predictive model uses a reduced representation (F3 red A scaled-down representation that is as close as possible to (F3*red ) calculate.

[0115] Figure 6 illustrates and schematically shows how the predictive model trained in Figure 5 can be used for prediction.

[0116] In this example, two first representations of the test domain in frequency space (F ~ 1) and (F ~ 2) is received and reduced to the specified region A. The result is two reduced first representations (F ~ 1 red ) and (F ~ 2 red ) is the reduced first representation (F ~ 1 red ) and (F ~ 2 red ) is fed to the trained predictive model (PM). The trained predictive model (PM) uses the reduced first representation (F ~ 1 red ) and (F ~ 2 red ) a reduced second representation (F ~ 3 red Calculate *). Reduced second representation (F ~ 3 red *) is, in a further stage, the received first representation (F ~ 1) The first received representation (F) that is discarded when reducing ~ 1) Region (F ~ 1 DI ) is supplemented by (the region outside the identified region A). As explained, the first received representation (F ~ 1) In place of, or in addition to, the received second expression (F ~ 2) A reduced third representation (F ~ 3 red* It is also possible to add it to ).

[0117] Supplementary expression (F ~ 3 red *)+(F ~ 1 DI) represents the inspection area in real space (O ~ 3*) can be generated by the inverse Fourier transform.

[0118] It should be noted that other methods, such as iterative reconstruction, can also be used to convert frequency-space descriptions to real-space descriptions.

[0119] The method according to the present invention can be performed using a computer system. The present invention further provides a computer system configured to perform the method according to the present invention (for example, using a computer program according to the present invention).

[0120] Figure 7 (Fig. 7) schematically and illustratively shows one embodiment of the computer system according to the present invention. The computer system (10) comprises a receiving unit (11), a control and calculation unit (12), and an output unit (13).

[0121] A "computer system" is a system for electronic data processing that processes data according to programmable calculation rules. Such a system typically comprises a control and calculation unit, often also called a "computer," which includes a processor for performing logical operations, memory for loading computer programs, and peripheral devices.

[0122] In computer technology, "peripherals" refer to all devices connected to a computer and used for controlling the computer and / or as input and output devices. Examples include monitors (screens), printers, scanners, mice, keyboards, joysticks, drives, cameras, microphones, and speakers. Built-in ports and expansion cards are also considered peripherals of computer technology.

[0123] Today's computer systems are generally subdivided into desktop PCs, portable PCs, laptops, notebooks, netbooks, and tablet PCs, as well as so-called handheld devices (e.g., smartphones); all such systems can be used for the implementation of the present invention.

[0124] Input to the computer system (for example, for user control) is achieved via input means such as a keyboard, mouse, microphone, or touch-sensitive display. Output is achieved via an output unit (13), which can be a monitor (screen), printer, and / or data storage medium.

[0125] The computer system (10) according to the present invention is configured to predict one or more second representations of the examination area in frequency space from a plurality of first representations of the examination area in frequency space representing the examination area during a first time span after administration of a contrast agent, wherein the one or more second representations represent the examination area during a second time span.

[0126] The control and calculation unit (12) functions to control the receiving unit (11) and the output unit (13), to coordinate data and signal flows between the various units, to process representations of the examination area, and to generate artificial radiographic images. Multiple control and calculation units may be present.

[0127] The receiving unit (11) functions to receive representations of the examination area. Representations may be transmitted, for example, from a magnetic resonance imaging system, transmitted from a computed tomography system, or read from a data storage medium. The magnetic resonance imaging system or computed tomography system can be a component of the computer system according to the present invention. However, the computer system according to the present invention may also be a component of the magnetic resonance imaging system or computed tomography system. Representations may be transmitted via a network connection or a direct connection. Representations may be transmitted via wireless communication (such as WLAN, Bluetooth, or mobile communication) and / or via cable. Multiple receiving units may be present. The data storage medium may also be a component of the computer system according to the present invention, or may be connected to it, for example, via a network. Multiple data storage media may be present.

[0128] The representation and, if applicable, further data (e.g., information about the object being examined, image acquisition parameters, etc.) are received by the receiving unit and transmitted to the control and computing units.

[0129] The control and computing units are configured to generate artificial radiographic images based on the received data.

[0130] The artificial radiation image can be displayed (for example, on a monitor), printed (for example, via a printer), or stored in a data storage medium via the output unit (13). Multiple output units may be present.

[0131] Figure 8 illustrates a preferred embodiment of the method according to the present invention for training a predictive model in the form of a flowchart.

[0132] Method (100) is as follows: (110) Step of receiving training data, wherein for each of a number of examination subjects, the training data includes: i) a number of first reference representations of the examination area in frequency space, at least a portion of which represent the examination area during a first time span after administration of contrast agent; and ii) one or more second reference representations of the examination area in frequency space representing the examination area during a second time span. (120) For each object under test: A step of supplying a prediction model with multiple first reference representations, wherein the prediction model is trained to generate one or more second reference representations based on the multiple first reference representations, wherein the training includes minimizing a loss function, wherein the loss function quantifies the deviation of the generated second reference representation from one or more received second reference representations. (130) Steps to output and / or store the trained predictive model and / or supply the trained predictive model to a method for predicting one or more representations of a new test domain under test. Includes.

[0133] Figure 9 illustrates a more preferred embodiment of the method according to the present invention for training a predictive model in the form of a flowchart.

[0134] Method (200) is as follows: (210) Step of receiving training data, wherein for each of a number of examination subjects, the training data includes: i) a number of first reference representations of the examination area in frequency space, at least a portion of which represent the examination area during a first time span after administration of contrast agent; and ii) one or more second reference representations of the examination area in frequency space representing the examination area during a second time span. (220) A step of identifying a region in the first reference representation, wherein the identified representation includes the center of the frequency space, (230) A step of reducing the reference representation to a specified region, wherein multiple reduced first reference representations and one or more reduced second reference representations are obtained for each object being examined. (240) For each object under test: A step of supplying a prediction model with multiple reduced first reference representations, wherein the prediction model is trained to generate one or more reduced second reference representations based on the multiple reduced first reference representations, wherein the training includes minimizing a loss function, wherein the loss function quantifies the deviation of the generated reduced second reference representations from one or more reduced second reference representations of the training data. (250) Steps to output and / or store the trained predictive model and / or supply the trained predictive model to a method for predicting one or more representations of a new test domain under test. Includes.

[0135] Figure 10 illustrates a preferred embodiment of the method according to the present invention for predicting one or more representations in the form of a flowchart.

[0136] Method (300) is as follows: (310) A step of providing a predictive model, where the predictive model has been trained according to method (100) described above, (320) A step of receiving a plurality of first representations of the examination area to be examined in frequency space, wherein at least a portion of the first representations represent the examination area during a first time span after administration of contrast agent, (330) A step of supplying multiple first representations to a predictive model, (340) A step of receiving one or more predicted representations of the examination region in frequency space from the prediction model, where one or more predicted representations represent the examination region during a second time span, (350) A step of converting one or more predicted representations into one or more representations of the examination domain in real space, (360) Step of outputting and / or storing representations of one or more inspection regions in real space. Includes.

[0137] Figure 11 illustrates a more preferred embodiment of the method according to the present invention for predicting one or more representations in the form of a flowchart.

[0138] Method (400) is as follows: (410) A step of providing a predictive model, where the predictive model has been trained according to the method (200) described above, (420) A step of receiving a plurality of first representations of the examination area to be examined in frequency space, wherein at least a portion of the first representations represent the examination area during a first time span after administration of contrast agent, (430) A step of identifying a region in the first representation, wherein the identified region includes the center of the frequency space, (440) A step of reducing the first representation to a specified region, where a plurality of reduced first representations are obtained, (450) A step of supplying multiple reduced first representations to a predictive model, (460) Step of receiving one or more second representations of the test region in frequency space from the prediction model, where one or more second representations represent the test region during a second time span, (470) A step of supplementing one or more second representations with one or more regions of a received first representation that are outside the identified region, wherein one or more supplemented second representations are obtained, (480) A step of converting one or more supplemented second representations into one or more representations in real space, (490) A step of outputting and / or storing one or more representations of the inspection area in real space. Includes. [Examples]

[0139] The following are some examples of how the present invention can be used to generate artificial radiation images.

[0140] Example 1 In one embodiment, the present invention is used to simulate an intravascular contrast agent (also known as a blood pool (contrast) agent).

[0141] When generating radiographic images with relatively long acquisition / scanning times, for example, in the case of acquiring images of the chest and abdomen under free breathing to depict the vascular system (e.g., diagnosis of pulmonary embolism under free breathing in MRI), the extracellular contrast agent is eliminated from the vascular system relatively rapidly, which means that the contrast decreases rapidly. However, it is advantageous if the contrast can be maintained for a longer period of time.

[0142] To solve this problem, multiple first representations of the examination region in frequency space are generated / received in a first step, where at least some of the first representations represent the examination region after administration of the contrast agent.

[0143] The contrast agent administered may be an extracellular contrast agent and / or an intracellular contrast agent.

[0144] The contrast agent is introduced into the blood vessels being examined, such as the brachial vein, preferably using a dosage based on body weight. From there, it travels along the circulatory system with the blood.

[0145] The "circulatory system" is the blood-covered pathway within the body of humans and most animals. This is the blood flow system formed by the heart and the network of blood vessels (cardiovascular system, vascular system).

[0146] Extracellular contrast agents circulate within the bloodstream for a certain period, depending on the object being examined, the contrast agent, and the dosage, while being continuously eliminated from the bloodstream via the kidneys.

[0147] While the contrast agent spreads and / or circulates within the vascular system under examination, at least one first representation of the vascular system or a portion thereof is captured. Multiple first representations representing different phases of contrast agent diffusion in the vascular system or a portion thereof (e.g., distribution phase, arterial phase, venous phase, etc.) can be captured. Capturing multiple images allows for later distinction of vascular types.

[0148] The measured representations indicate the vascular system or a portion thereof with enhanced contrast against surrounding tissue. Preferably, at least one representation indicates an artery with enhanced contrast (arterial phase), while at least one further representation indicates a vein with enhanced contrast (venous phase).

[0149] Artificial representations are generated based on measured representations. Preferably, the artificial representations show the same examination area as the measured representations. If multiple measured representations of an examination area are captured at different time points after contrast agent administration, later representations will show vessels with progressively decreasing contrast against surrounding tissue, as the contrast agent is gradually eliminated from the vessels. In contrast, artificial representations show vessels with consistently high contrast against surrounding tissue.

[0150] This is achieved by using measured representations supplied to a predictive model according to the present invention, which is pre-trained to predict multiple representations that show constant vascular contrast enhancement over time, based on measured representations that show vascular contrast enhancement that changes over time.

[0151] Reference data used for training and validating such predictive models typically includes measured representations of the examination area after administration of extracellular or intracellular contrast agents. Reference data may further include representations of the examination area after administration of blood pool contrast agents. Such reference data can be verified, for example, in clinical studies. An example of an intravascular contrast agent that can be used in such clinical trials is fermoxyl. Fermoxyl is a colloidal iron-carbohydrate complex approved for parenteral treatment of iron deficiency in chronic kidney disease where oral therapy is not feasible. Fermoxyl is administered as an intravenous injection. Fermoxyl is marketed as a liquid for intravenous injection under the trade names Rienso® or Ferahme®. Iron-carbohydrate complexes exhibit superparamagnetic properties and can therefore be used for contrast enhancement in MRI scans (off-label) (see, for example, LPSmits et al.: Evaluation of ultrasmall superparamagnetic iron-oxide (USPIO) enhanced MRI with ferumoxytol to quantify arterial wall inflammation, Atherosclerosis 2017, 263:211-218).

[0152] Similarly, it is conceivable to use the results after administration of the intravascular contrast agent Ablavar (registered trademark) as training data.

[0153] Similarly, it is conceivable to synthesize a reference representation showing blood vessels in the examination area with a certain contrast enhancement over time, for example, by a segmentation method based on the first representation. Segmentation methods are widely described in the literature. The following publications can be cited as examples: F. Conversano et al.: Hepatic Vessel Segmentation for 3D Planning of Liver Surgery, Acad Radiol 2011, 18:461-470; S. Moccia et al.: Blood vessel segmentation algorithms - Review of methods, datasets and evaluation metrics, Computer Methods and Programs in Biomedicine 158(2018)71-91; M. Marcan et al.: Segmentation of hepatic vessels from MRI images for planning of electroporation-based treatments in the liver, Radiol Oncol 2014;48(3):267-281; ​​TAHope et al.: Improvement of Gadoxetate Arterial Phase Capture With a High Spatio-Temporal Resolution Multiphase Three-Dimensional SPGR-Dixon Sequence, Journal of Magnetic Resonance Imaging 38:938-945(2013); WO2009 / 135923A1, US6754376B1, WO2014 / 162273A1, WO2017 / 139110A1, WO2007 / 053676A2, EP2750102A1).

[0154] Based on the first provided representation, the trained predictive model then generates a second representation showing a constant contrast enhancement of blood vessels over time.

[0155] Example 2 In a preferred embodiment, the present invention is used to generate (predict) one or more artificial MRI images in dynamic contrast-enhanced magnetic resonance imaging.

[0156] In the following text, the term “image” is used. “Image” is a representation in the context of this invention. An image can be a representation in real space or a representation in frequency space. For training and predicting predictive models, representations in frequency space, i.e., k-space data, are always used. However, if the representations in real space are generated based on measurements, they can be transformed into representations in frequency space, for example, by a Fourier transform, before being introduced into training and / or prediction.

[0157] The examination area is introduced into a fundamental magnetic field. The examination area is subjected to the MRI method, and in this process, multiple MRI images showing the examination area are generated during a first time span. These MRI images generated based on measurements during the first time span are also referred to herein as the first MRI images.

[0158] The term "multiple" means that at least two (first) MRI images, preferably at least three (first) MRI images, and very preferably at least four (first) MRI images are generated.

[0159] The contrast agent, which spreads across the examination area, is administered to the subject of examination. The contrast agent is preferably administered intravenously as a bolus (for example, into a vein in the arm) using a dosage based on body weight.

[0160] Preferably, the contrast agent is a hepatobiliary contrast agent such as Gd-EOB-DTPA or Gd-BOPTA. In a particularly preferred embodiment, the contrast agent is a substance or mixture of substances having gadoxetic acid or a salt of gadoxetic acid as a contrast-enhancing active substance. Disodium salt of gadoxetic acid (Gd-EOB-DTPA disodium) is particularly preferred.

[0161] The first time span preferably includes a contrast agent distributed within the examination area. Preferably, the first time span includes the arterial phase and / or portal phase and / or late phase in dynamic contrast-enhanced magnetic resonance imaging of the liver or a portion of the liver being examined. The phases are defined and described in the following publications, for example: J.Magn.Reson.Imaging, 2012, 35(3):492-511, doi:10.1002 / jmri.22833; Clujul Medical, 2015, Vol.88 no.4:438-448, DOI:10.15386 / cjmed-414; Journal of Hepatology, 2019, Vol.71:534-542, http: / / dx.doi.org / 10.1016 / j.jhep.2019.05.005).

[0162] Figure 12 schematically shows the time-course profile of contrast agent concentration in the hepatic artery (A), hepatic vein (V), and healthy hepatocytes (P) after administration of hepatobiliary contrast agent into a human brachial vein. Concentration is shown as signal intensity I in the described region (hepatic artery, hepatic vein, hepatocytes) in magnetic resonance measurements as a function of time t. Upon intravenous bolus injection, the contrast agent concentration first rises in the hepatic artery (A) (dashed curve). The concentration passes a maximum value and then decreases. The concentration in the hepatic vein (V) rises more slowly than in the hepatic artery, reaching its maximum value (dotted curve). The contrast agent concentration in healthy hepatocytes (P) rises slowly (continuous curve), reaching its maximum value only at a very later point in time (the maximum value is not shown in Figure 12). Several characteristic time points can be defined: at time point TP0, the contrast agent is administered intravenously as a bolus. At time point TP1, the contrast agent concentration (signal intensity) in the hepatic artery reaches its maximum value. At time point TP2, the signal intensity curves for the hepatic artery and hepatic vein intersect. At time point TP3, the contrast agent concentration (signal intensity) in the hepatic vein passes its maximum value. At time point TP4, the signal intensity curves for the hepatic artery and hepatocytes intersect. At time point T5, the concentrations in the hepatic artery and hepatic vein have decreased to a level that no longer produces measurable contrast enhancement.

[0163] In a preferred embodiment, the first time span is (i) Indicates the examination area that does not contain contrast agent, (ii) The examination area during the arterial phase is shown, where the contrast agent spreads through the arteries into the examination area. (iii) The examination area during the portal vein phase, where the contrast agent enters the examination area via the portal vein, and (iv) The examination area during the late phase is shown, where the concentration of contrast agent in arteries and veins decreases, and the concentration of contrast agent in hepatocytes increases. The system is selected to generate MRI images of the liver or a portion of the liver being examined.

[0164] Preferably, the first time span begins with a time span of 1 minute to 1 second before the administration of the contrast agent, or with the administration of the contrast agent, and lasts for a time span of 2 to 15 minutes, preferably 2 to 13 minutes, and more preferably 3 to 10 minutes, after the administration of the contrast agent. Because the contrast agent undergoes very slow renal and biliary excretion, the second time span may last for up to 2 hours or more after the administration of the contrast agent.

[0165] In a preferred embodiment, the first time span includes at least time points TP0, TP1, TP2, TP3, and TP4.

[0166] In a preferred embodiment, at least MRI images are generated (based on measurements) for all of the following phases: in the time span before TP0, in the time span from TP0 to TP1, in the time span from TP1 to TP2, in the time span from TP2 to TP3, and in the time span from TP3 to TP4.

[0167] It is possible that one or more MRI images are generated (based on measurements) in the time spans before TP0, TP0-TP1, TP1-TP2, TP2-TP3, and TP3-TP4.

[0168] Based on the (first) MRI images generated (based on measurements) during the first time span, a second MRI image or a number of second MRI images showing the examination area during the second time span are predicted. The MRI images predicted for the second time span are also referred to herein as second MRI images.

[0169] In a preferred embodiment of the present invention, the second time span follows the first time span.

[0170] The second time span is preferably a time span within the hepatobiliary phase; preferably a time span that begins at least 10 minutes after administration of the contrast agent, and more preferably at least 20 minutes after administration of the contrast agent.

[0171] At least one second representation representing the examination area during a second time span is predicted using a predictive model according to the present invention. The predictive model is pre-trained to predict one or more MRI images representing the examination area during a second time span based on a plurality of first MRI images representing the examination area during a first time span.

[0172] The embodiment described here is also schematically shown in Figure 1(b).

[0173] Example 3 In a more preferred embodiment of the present invention, the present invention is used to distinguish liver lesions from blood vessels. In T1-weighted MRI images, Primovist® leads to a clear signal enhancement in the parenchyma tissue of healthy liver 10–20 minutes after injection (in the hepatobiliary phase), while lesions that do not contain hepatocytes or contain only a small number of hepatocytes, such as metastases or moderately to poorly differentiated hepatocellular carcinoma (HCC), appear as darker areas. However, in the hepatobiliary phase, blood vessels also appear as dark areas, meaning that in MRI images generated between the hepatobiliary phases, liver lesions and blood vessels cannot be distinguished based solely on contrast.

[0174] The present invention can be used to generate artificial MRI images of the liver or a portion of the liver to be examined, in which the contrast between blood vessels and liver cells within the liver is artificially minimized to make liver lesions easier to identify.

[0175] "Image" is a representation in the context of this invention. An image can be a representation in real space or a representation in frequency space. Representations in frequency space are always used for training and predicting predictive models. However, representations in real space can be generated based on measurements and then converted to representations in frequency space, for example, by a Fourier transform, before being introduced into training and / or prediction.

[0176] The multiple first representations include at least one representation of an examination area in which blood vessels are identifiable, and are preferably depicted with contrast enhancement by a contrast agent (vascular representation).

[0177] When paramagnetic contrast agents are used, blood vessels in such representations are characterized by high signal intensity due to contrast enhancement (high signal rendering). Therefore, (continuous) structures in such representations with signal intensity within an empirically observable range can be attributed to blood vessels. This means that using such representations provides information about where blood vessels are depicted in real-space representations, or which structures can be attributed to blood vessels (arteries and / or veins) in real-space representations.

[0178] Multiple first representations further include at least one representation of a test region in which healthy hepatocytes are depicted with contrast enhancement (hepatocyte representation), for example, a representation of a test region acquired during the hepatobiliary phase.

[0179] Information from at least one vascular representation via blood vessels is combined with information from at least one hepatocyte representation. This involves (artificially) generating (computing) at least one representation in which the contrast difference between structures that may be attributable to blood vessels and structures that may be attributable to healthy hepatocytes is leveled.

[0180] Here, the term "leveling" means "harmonization" or "minimization." The purpose of leveling is to eliminate the boundary between blood vessels and healthy liver cells in the artificially generated representation, making the blood vessels and healthy liver cells appear as a uniform tissue, while making liver lesions structurally prominent due to the resulting contrast.

[0181] Typically, the artificial representation of 1 (number=1) is predicted based on the vascular representation of 1 (number=1) and the hepatocyte representation of 1 (number=1).

[0182] In addition to at least one vascular representation and at least one hepatocyte representation, at least one native representation may also be used to predict at least one artificial representation.

[0183] In one embodiment, the generation of an artificial representation is as follows: - A step of supplying a predictive model with at least one vascular representation and at least one hepatocyte representation, wherein the predictive model is trained by supervised learning to generate at least one artificial representation from at least one reference vascular representation and at least one reference hepatocyte representation based on a reference representation, wherein the difference in contrast between structures that may be attributable to blood vessels and structures that may be attributable to healthy hepatocytes is leveled in at least one artificial representation. - A step of receiving at least one artificial representation as output from the predictive model. Includes.

[0184] Example 4 In a further embodiment, the present invention is used to generate native MRI images of the liver. Here, one or more artificial MRI images are generated of the liver or part of the liver being examined, showing the liver or part of the liver without contrast enhancement induced by a contrast agent. All artificial MRI images are created based on MRI images obtained with contrast enhancement induced by a contrast agent.

[0185] "Image" is a representation in the context of this invention. An image can be a representation in real space or a representation in frequency space. Representations in frequency space are always used for training and predicting predictive models. However, representations in real space can be generated based on measurements and then converted to representations in frequency space by a Fourier transform, for example, before being introduced into training and / or prediction.

[0186] The examination area is introduced into a basic magnetic field. A contrast agent that spreads across the examination area is administered to the subject of examination. The contrast agent is preferably administered intravenously (e.g., into the brachial vein) as a bolus using weight-based dosing. Preferably, the contrast agent is a hepatobiliary contrast agent such as Gd-EOB-DTPA or Gd-BOPTA. In a particularly preferred embodiment, the contrast agent is a substance or mixture of substances having gadoxetic acid or a salt of gadoxetic acid as a contrast-enhancing active substance. Disodium gadoxetic acid (Gd-EOB-DTPA disodium) is particularly preferred.

[0187] Multiple first representations of the examination area are generated, representing the examination area in a first time span after the administration of the contrast agent. Preferably, the multiple first representations are T1-weighted depictions.

[0188] Preferably, the plurality of first representations include at least one representation of the examination area representing the examination area during the dynamic phase, e.g., at least one representation representing the examination area during the arterial phase, during the venous phase, and / or during the late phase (see, for example, the relevant description in Figure 12 and Example 2). When a paramagnetic contrast agent is used, blood vessels in such representations are characterized by high signal intensity due to contrast enhancement (high signal depiction).

[0189] Preferably, the multiple first representations further include at least one representation of an examination area representing the examination area during the hepatobiliary phase. During the hepatobiliary phase, healthy liver tissue (parenchyma) is depicted with enhanced contrast.

[0190] Dynamic and hepatobiliary phase MRI examinations are performed over relatively long time spans. Over these time spans, patient movement should be avoided to minimize motion artifacts in the radiographic images. Prolonged restriction of movement can be uncomfortable for the patient. Therefore, shortened MRI procedures are currently being established in which the contrast agent is already administered to the subject over a time span prior to MRI image acquisition (i.e., 10-20 minutes) to allow direct acquisition of MRI images within the hepatobiliary phase. Then, after administration of a second dose of the contrast agent, dynamic phase MRI images are acquired in the same MRI process. Compared to conventional MRI processes, the dwell time of the patient or subject in the MRI is significantly shorter as a result. Therefore, according to the present invention, it is preferable to record at least one representation of the liver or portion of the liver in the hepatobiliary phase after the (first) administration of the first contrast agent to the subject, and to record at least one further representation of the same liver or portion of the liver in the dynamic phase after administration of a second contrast agent or a second administration of the first contrast agent to the same subject. The first contrast agent is a hepatobiliary paramagnetic contrast agent. The second contrast agent may be an extracellular paramagnetic contrast agent.

[0191] Next, a first representation of the examination region is supplied to a predictive model according to the present invention. The predictive model is pre-trained to predict, based on the received first representation, one or more second representations showing the liver or portion of the liver under examination without contrast enhancement caused by a contrast agent. The predictive model is preferably created using a self-learning algorithm in a supervised machine learning process. The training data used for training includes multiple representations of the examination region between the dynamic and hepatobiliary phases of a large number of livers or portions of the liver under examination. Furthermore, the training data also includes representations of the examination region generated in the absence of contrast enhancement, i.e., without administration of a contrast agent.

[0192] The embodiment described here is also schematically shown in Figure 1(c).

[0193] Example 5 In a more preferred embodiment, the present invention is used to reduce the examination time for patients in dynamic contrast-enhanced magnetic resonance imaging of the liver.

[0194] Here, the contrast agent is administered in the form of two boluses. The first administration is performed when the object of examination is not yet positioned within the MRI scanner. In the case of the first administration, the first contrast agent is administered. The first contrast agent is preferably administered intravenously (e.g., into the brachial vein) as a bolus using body weight-based dosing. The first contrast agent is preferably a hepatobiliary contrast agent such as Gd-EOB-DTPA or Gd-BOPTA. In a particularly preferred embodiment, the first contrast agent is a substance or mixture of substances having gadoxetic acid or a salt of gadoxetic acid as a contrast-enhancing active substance. Disodium salt of gadoxetic acid (Gd-EOB-DTPA disodium) is particularly preferred.

[0195] After the administration of the first contrast agent, the subject of examination is introduced into the MRI scanner, and a certain time span can be waited before the first MRI image is generated at the first point in time.

[0196] "Image" is a representation in the context of this invention. An image can be a representation in real space or a representation in frequency space. Representations in frequency space are always used for training and predicting predictive models. However, representations in real space can be generated based on measurements and then converted to representations in frequency space by a Fourier transform, for example, before being introduced into training and / or prediction.

[0197] The time span between the first administration and the generation of the first MRI image is preferably in the range of 5 minutes to 1 hour, more preferably in the range of 10 minutes to 30 minutes, and most preferably in the range of 8 minutes to 25 minutes.

[0198] The first MRI image represents the liver or a portion of the liver being examined during the hepatobiliary phase after the administration of the first contrast agent. Healthy hepatocytes are depicted with enhanced contrast in the first MRI image as a result of the administration of the first contrast agent.

[0199] The hepatobiliary phase, in which the first MRI image is generated, is also referred to herein as the first hepatobiliary phase. The first contrast agent reaches healthy hepatocytes, and in the case of a paramagnetic contrast agent, it results in enhanced contrast and signaling of healthy hepatocytes. No MRI images are generated during the arterial phase, portal phase, and late phase that occur after the administration of the first contrast agent. The arterial phase, portal phase, and late phase that occur after the administration of the first contrast agent are also referred to herein as the first arterial phase, the first portal phase, and the first late phase.

[0200] It is possible that multiple MRI images are generated during the first hepatobiliary phase.

[0201] After the generation of one or more first MRI images during the first hepatobiliary phase, a second contrast agent is administered. The second contrast agent is a second contrast agent. The second contrast agent may be the same as the first contrast agent; however, the second contrast agent may be a different contrast agent, preferably an extracellular contrast agent. Similarly, the second contrast agent is preferably administered intravenously (e.g., into the brachial vein) as a bolus using body weight-based dosing.

[0202] The administration of the first contrast agent is also referred to herein as the first administration; the administration of the second contrast agent is also referred to herein as the second administration. If the first contrast agent and the second contrast agent are the same, the procedure performed in this manner is the first administration of the hepatobiliary contrast agent, and at a later point in time, the second administration of the hepatobiliary contrast agent. If the first contrast agent and the second contrast agent are different, the procedure performed is the first administration of the first contrast agent, the first contrast agent being the hepatobiliary contrast agent, and at a later point in time, the second administration of the second (different) contrast agent.

[0203] At the time of the second administration (or administration of the second contrast agent), the object of examination is preferably already positioned within the MRI scanner. After the administration of the second contrast agent, the system again passes through the arterial phase, portal phase, and late phase. These arterial phase, portal phase, and late phase are also referred to herein as the second arterial phase, the second portal phase, and the second late phase. During the second arterial phase and / or the second portal phase and / or the second late phase, an MRI image or a series of MRI images are generated. These MRI images are referred to as the second, third, fourth, and so on, in order of acquisition.

[0204] In a preferred embodiment, a second MRI image is generated during the second arterial phase, a third MRI image is generated during the second portal venous phase, and a fourth MRI image is generated during the second late phase. Such a second MRI image particularly shows arteries with contrast enhancement; such a third MRI image particularly shows veins with contrast enhancement.

[0205] Furthermore, it is possible that two or more MRI images may be generated between the aforementioned phases.

[0206] It is possible to calculate an artificial MRI image from MRI images generated during one or more phases after the administration of the first and second contrast agents.

[0207] The purpose of generating an artificial MRI image from the measured MRI image is to enhance the contrast between healthy liver tissue and other areas. When a paramagnetic contrast agent for the hepatobiliary tract is used as the first contrast agent, the signal intensity of healthy liver tissue during the second arterial phase, second portal venous phase, and second late phase is still elevated as a result of the administration of the first contrast agent. Similarly, the second contrast agent, which spreads in the second phase as described above, will result in a high signal in the tissue it spreads to. This means that there is only low contrast in the MRI image between healthy liver tissue and the rest of the tissue, and the rest of the tissue is enhanced in contrast by the (second) contrast agent. To enhance this contrast, what is generated using a predictive model is at least one artificial MRI image showing how the examination area looks in the dynamic phase after the administration of the first contrast agent, or how the examination area looks if only the second contrast agent is administered: blood vessels are depicted with enhanced contrast as a result of the administration of the second contrast agent, but healthy hepatocytes are not depicted with enhanced contrast as a result of the administration of the first contrast agent. In other words, the resulting image is an artificial MRI image that looks like a second MRI image, but the difference is that the contrast enhancement of healthy liver cells caused by the administration of the first contrast agent is subtracted (removed) from the second MRI image.

[0208] The embodiment described here is also schematically shown in Figure 1(d). [Brief explanation of the drawing]

[0209] [Figure 1(a)] Figure 1(a) schematically shows a typical profile of a radiographic examination. [Figure 1(b)] Figure 1(b) schematically shows the profile of the radiation inspection according to the present invention. [Figure 1(c)] Figure 1(c) schematically shows a further profile of the radiographic inspection according to the present invention. [Figure 1(d)] Figure 1(d) schematically shows a further profile of the radiographic inspection according to the present invention. [Figure 2]Figure 2 (Fig. 2) illustrates and schematically shows the generation of the test domain representation in real space and frequency space. [Figure 3] Figure 3 (Fig. 3) schematically and illustratively shows how the representations of the test domains (F1), (F2), and (F3) in frequency space, as generated in Figure 2, can be used to train a predictive model (PM). [Figure 4] Figure 4 (Fig. 4) shows the predictive model (PM) trained in Figure 3. [Figure 5] Figure 5 (Fig. 5) illustrates and schematically shows the steps in training a predictive model according to a preferred embodiment of the present invention. [Figure 6] Figure 6 illustrates and schematically shows how the predictive model trained in Figure 5 can be used for prediction. [Figure 7] Figure 7 (Fig. 7) schematically and illustratively shows one embodiment of the computer system according to the present invention. [Figure 8] Figure 8 illustrates a preferred embodiment of the method according to the present invention for training a predictive model in the form of a flowchart. [Figure 9] Figure 9 illustrates a more preferred embodiment of the method according to the present invention for training a predictive model in the form of a flowchart. [Figure 10] Figure 10 illustrates a preferred embodiment of the method according to the present invention for predicting one or more representations in the form of a flowchart. [Figure 11] Figure 11 illustrates a more preferred embodiment of the method according to the present invention for predicting one or more representations in the form of a flowchart. [Figure 12] Figure 12 schematically shows the time-course profiles of contrast agent concentrations in the hepatic artery (A), hepatic vein (V), and healthy hepatocytes (P) after administration of hepatobiliary contrast agent into a human brachial vein.

Claims

1. A computer implementation method, - A step of receiving a plurality of first representations of a test area to be examined in frequency space, wherein at least some of the first representations represent the test area during a first time span after administration of a contrast agent, - A step of supplying the plurality of first representations to a predictive model, wherein the predictive model is trained on the first reference representations to generate one or more second reference representations from first reference representations of the examination areas of a plurality of examination subjects, wherein at least some of the first reference representations represent the examination areas during a first time span after administration of a contrast agent in frequency space, and the one or more second reference representations represent the examination areas during a second time span in frequency space, - A step of receiving one or more predicted representations of the test region in frequency space from the prediction model, wherein the one or more predicted representations represent the test region during a second time span, - A step of converting one or more predicted representations into one or more representations of the inspection domain in real space, - A step of outputting one or more representations of the inspection area in real space, - A step of receiving a plurality of first representations of a test area to be examined in frequency space, wherein at least some of the first representations represent the test area during a first time span after administration of a contrast agent, - A step of identifying a region in the first representation, wherein the identified region includes the center of the frequency space, - A step of reducing the first representation to the specified region, - A step of supplying a plurality of the reduced first representations to a predictive model, - A step of receiving one or more second representations of the test region in frequency space from the prediction model, wherein the one or more second representations represent the test region during a second time span, - A step of supplementing one or more second representations with one or more regions of the received first representation located outside the identified region, - A step of converting the one or more supplemented second representations into one or more representations of the inspection area in real space, - A computer implementation method comprising the step of outputting one or more representations of the inspection area in real space.

2. The method according to claim 1, wherein the plurality of first expressions are - Includes at least one representation of the examination area in frequency space representing the examination area prior to the administration of the contrast agent, and - Includes at least one representation of the examination area in frequency space representing the examination area in the first time span after the administration of the contrast agent, The method wherein one or more predicted representations represent the inspection area in the second time span, and the second time span follows the first time span.

3. The method according to claim 1, wherein the plurality of first representations include at least two representations of the examination area in frequency space representing the examination area in a first time span after the administration of the contrast agent, and the one or more predicted representations represent the examination area in a second time span, the second time span preceding the first time span.

4. The method according to claim 1, wherein the plurality of first expressions are - Includes at least one representation of the examination area in frequency space representing the examination area in the first time span after the administration of the first contrast agent, - The method comprising at least one representation in frequency space of the examination area representing the examination area in the first time span after the administration of the second contrast agent, wherein the first contrast agent and the second contrast agent are identical or different, the second contrast agent is administered after the first contrast agent, and the one or more predicted representations represent the examination area in the second time span, the second time span precedes the first time span.

5. The method according to claim 1, wherein the one or more predicted representations represent the inspection area in a second time span accompanied by a constant contrast enhancement over time.

6. The method according to any one of claims 1 to 5, wherein the prediction model comprises an artificial neural network.

7. The method according to any one of claims 1 to 6, wherein the first representation of the inspection region in frequency space is k-space data of magnetic resonance imaging.

8. The method according to any one of claims 1 to 6, wherein a plurality of radiation images of the inspection area in real space are received in a first step, and these received radiation images are converted by Fourier transform to the first representation of the inspection area in frequency space.

9. The method according to any one of claims 1 to 8, wherein the one or more supplemented second representations are transformed by an inverse Fourier transform into one or more representations of the inspection region in real space.

10. A system comprising: a receiving unit, a control and calculation unit, and an output unit, wherein the control and calculation unit, - The receiving unit is configured to prompt itself to receive a plurality of first representations of the examination area of ​​the object to be examined in frequency space, at least some of the first representations representing the examination area during a first time span after the administration of the contrast agent, - The plurality of first representations are configured to supply a predictive model, which is trained on the first reference representations to generate one or more second reference representations from the first reference representations of the examination region of a plurality of examination subjects, wherein at least some of the first reference representations represent the examination region during a first time span after administration of contrast agent in frequency space, and the one or more second reference representations represent the examination region during a second time span in frequency space. - The system is configured to receive one or more predicted representations of the test region in frequency space from the prediction model, wherein the one or more predicted representations represent the test region during a second time span. - The system is configured to convert one or more predicted representations into one or more representations of the inspection area in real space, - The output unit is configured to prompt the output unit to output one or more representations of the inspection area in real space, - It is configured to receive a plurality of first representations of the examination region of the object under examination in frequency space, and at least some of the first representations represent the examination region during a first time span after administration of contrast agent, - Configured to identify a region in the first representation, the identified region includes the center of the frequency space, - The first representation is configured to be reduced to the specified region, - Configured to supply a plurality of the reduced first representations to a predictive model, - The system is configured to receive one or more second representations of the test region in frequency space from the prediction model, wherein the one or more second representations represent the test region during a second time span. - The one or more second representations are configured to be supplemented by one or more regions of the received first representation located outside the identified region, - The system is configured to convert the one or more supplemented second representations into one or more representations of the inspection area in real space. - The system configured to output one or more representations of the inspection area in real space.

11. A computer program product comprising a computer program that can be loaded into the memory of a computer system, wherein the computer program product: - A step of receiving a plurality of first representations of a test area to be examined in frequency space, wherein at least some of the first representations represent the test area during a first time span after administration of a contrast agent, - A step of supplying the plurality of first representations to a predictive model, wherein the predictive model is trained on the first reference representations to generate one or more second reference representations from first reference representations of the examination areas of a plurality of examination subjects, wherein at least some of the first reference representations represent the examination areas during a first time span after administration of a contrast agent in frequency space, and the one or more second reference representations represent the examination areas during a second time span in frequency space, - A step of receiving one or more predicted representations of the test region in frequency space from the prediction model, wherein the one or more predicted representations represent the test region during a second time span, - A step of converting one or more predicted representations into one or more representations of the inspection area in real space, - A step of outputting one or more representations of the inspection area in real space, - A step of receiving a plurality of first representations of a test area to be examined in frequency space, wherein at least some of the first representations represent the test area during a first time span after administration of a contrast agent, - A step of identifying a region in the first representation, wherein the identified region includes the center of the frequency space, - A step of reducing the first representation to the specified region, - A step of supplying a plurality of the reduced first representations to a predictive model, - A step of receiving one or more second representations of the test region in frequency space from the prediction model, wherein the one or more second representations represent the test region during a second time span, - A step of supplementing one or more second representations with one or more regions of the received first representation located outside the identified region, - A step of converting the one or more supplemented second representations into one or more representations of the inspection area in real space, - The product, which prompts the computer system to perform the steps of outputting one or more representations of the inspection area in real space.

12. A kit comprising a contrast agent and the computer program product according to claim 11.

Images