Generating synthetic radiological images

Abstract

The present disclosure relates to the technical field of radiology, in particular to the support of radiologists in radiological examinations using methods of artificial intelligence. The present disclosure is concerned with training a machine learning model and using the trained model to predict a synthetic representation of a liver or a part of a liver of a human in a hepatobiliary phase following administration of a hepatobiliary contrast agent.

Description

[0001] BYP250099 FC

[0002] Generating synthetic radiological images TECHNICAL AREA

[0003] This disclosure relates to the technical field of radiology, in particular the support of radiologists in radiological examinations using artificial intelligence methods. The disclosure relates to training a machine learning model and using the trained model to predict a synthetic representation of a human liver or part of a liver in the hepatobiliary phase after administration of a hepatobiliary contrast agent.

[0004] INTRODUCTION

[0005] The temporal tracking of processes within a person's body using imaging techniques plays an important role in the diagnosis and / or therapy of diseases, among other things.

[0006] One example is the detection and differential diagnosis of focal liver lesions using dynamic contrast-enhancing magnetic resonance imaging (MRI) with a hepatobiliary contrast agent.

[0007] A hepatobiliary contrast agent such as Primovist® can be used to detect tumors in the liver. The blood supply to healthy liver tissue is primarily provided by the portal vein (vena portae), while the hepatic artery (arteria hepatica) supplies most primary tumors. Therefore, after an intravenous bolus injection of a contrast agent, a time delay can be observed between the signal increase in healthy liver parenchyma and in the tumor.

[0008] In addition to malignant tumors, benign lesions such as cysts, hemangiomas, and focal nodular hyperplasia (FNH) are frequently found in the liver. For appropriate treatment planning, these must be differentiated from malignant tumors. Primovist® can be used to detect benign and malignant focal liver lesions. It provides information about the nature of these lesions using T1-weighted MRI. Differentiation is achieved by utilizing the different blood supply to the liver and tumors, as well as the temporal course of contrast enhancement.

[0009] The contrast enhancement achieved with Primovist® during the onset phase reveals typical perfusion patterns that provide information for characterizing the lesions. Visualizing the vascularization helps to characterize the lesion types and determine the spatial relationship between the tumor and blood vessels.

[0010] In TI-weighted MRI scans, Primovist® leads to a marked signal enhancement in healthy liver parenchyma 10-20 minutes after injection (in the hepatobiliary phase), while lesions containing no or few hepatocytes, e.g. metastases or moderately to poorly differentiated hepatocellular carcinomas (HCCs), appear as darker areas.

[0011] Monitoring the distribution of the contrast agent over time offers a valuable tool for the detection and differential diagnosis of focal liver lesions; however, the examination takes a relatively long time. During this period, patient movement should be largely avoided to minimize motion artifacts in the MRI scans. This prolonged restriction of movement can be uncomfortable for the patient.

[0012] In the patent application WO2021 / 052896A1, it is proposed that one or more MRI scans during the hepatobiliary phase should not be generated manually, but rather calculated (predicted) based on MRI scans from one or more previous phases, in order to shorten the patient's time in the MRI scanner. The approach described in patent application WO2021 / 052896A1 involves training a machine learning model to predict an MRI scan of the same area at a later time based on MRI scans of the area before and / or immediately after the administration of a contrast agent.

[0013] There is a need to further improve the quality of predicted (synthetic) MRI scans in order to best support a radiologist and / or physician in diagnostics and / or therapy planning.

[0014] SUMMARY

[0015] This revelation is dedicated to this task and other tasks.

[0016] A first subject of the present disclosure is a computer-implemented method comprising the steps:

[0017] Providing an initial representation, wherein the initial representation represents a human liver or part of a liver in a portal venous phase following application of a hepatobiliary contrast agent in the frequency domain,

[0018] Providing a second representation, wherein the second representation represents the liver or part of the liver of a person in a transitional phase after application of the hepatobiliary contrast agent in the frequency space,

[0019] Generating a first difference representation based on the first representation and the second representation, where the first difference representation represents a difference between the first representation and the second representation.

[0020] Inputting the first difference representation into a trained machine learning model, receiving a second difference representation as output from the machine learning model,

[0021] Generating a third representation, wherein generating the third representation involves adding the second difference representation to the second representation or to the first representation,

[0022] Transforming the third representation into a third representation in spatial space, outputting and / or storing the third representation in spatial space and / or transmitting the third representation in spatial space to a separate computer system,

[0023] wherein the trained machine learning model was trained in a supervised learning process based on training data, wherein the training data for each reference person of a plurality of reference persons comprise (i) a first reference difference representation as input data and (ii) a second reference difference representation as target data, wherein the first reference difference representation represents a difference between a first reference representation and a second reference representation, and the second reference difference representation represents a difference between the second reference representation and a third reference representation or between the first reference representation and the third reference representation, wherein the first reference representation represents a liver or part of a liver of the reference person in a portal venous phase after application of the hepatobiliary contrast agent in the frequency domain.wherein the second reference representation represents the liver or part of the liver of the reference person in a transitional phase after application of the hepatobiliary contrast agent in the frequency domain, and the third reference representation represents the liver or part of the liver of the reference person in a hepatobiliary phase after application of the hepatobiliary contrast agent in the frequency domain. Another subject of the present disclosure is a computer system comprising

[0024] • a receiving unit,

[0025] • a control and computing unit and

[0026] • an output unit,

[0027] where the control and computing unit is configured,

[0028] to provide an initial representation or to cause the receiving unit to receive the initial representation, wherein the initial representation represents a liver or part of a liver of a human being in a portal venous phase after application of a hepatobiliary contrast agent in the frequency domain,

[0029] to provide a second representation or to cause the receiving unit to receive the second representation, wherein the second representation represents the liver or part of the liver of the human being in a transitional phase after application of the hepatobiliary contrast agent in the frequency domain,

[0030] to generate a first difference representation based on the first representation and the second representation, where the first difference representation represents a difference between the first representation and the second representation,

[0031] to input the first difference representation into a trained machine learning model, and to receive a second difference representation as output from the machine learning model,

[0032] to generate a third representation, wherein generating the third representation involves adding the second difference representation to the second representation or to the first representation,

[0033] to transform the third representation into a third representation in spatial space, to cause the output unit to output the third representation in spatial space, to store it in a data store and / or to transmit it to a separate computer system, wherein the trained machine learning model was trained in a supervised learning process based on training data, wherein the training data for each reference person of a plurality of reference persons comprise (i) a first reference difference representation as input data and (ii) a second reference difference representation as target data,wherein the first reference difference representation represents a difference between a first reference representation and a second reference representation, and the second reference difference representation represents a difference between the second reference representation and a third reference representation or between the first reference representation and the third reference representation, wherein the first reference representation represents a liver or part of a liver of the reference person in a portal venous phase after application of the hepatobiliary contrast agent in the frequency domain,where the second reference representation represents the liver or part of the liver of the reference person in a transition phase after administration of the hepatobiliary contrast agent in the frequency domain, and the third reference representation represents the liver or part of the liver of the reference person in a hepatobiliary phase after administration of the hepatobiliary contrast agent in the frequency domain.

[0034] Another subject of the present disclosure is a computer program product comprising a computer program that can be loaded into the working memory of a computer system and causes the computer system to perform the following steps: providing a first representation, wherein the first representation represents a liver or part of a liver of a human being in a portal venous phase after application of a hepatobiliary contrast agent in the frequency domain,

[0035] Providing a second representation, wherein the second representation represents the liver or part of the liver of a person in a transitional phase after application of the hepatobiliary contrast agent in the frequency space,

[0036] Generating a first difference representation based on the first representation and the second representation, where the first difference representation represents a difference between the first representation and the second representation.

[0037] Inputting the first difference representation into a trained machine learning model, receiving a second difference representation as output from the machine learning model,

[0038] Generating a third representation, wherein generating the third representation involves adding the second difference representation to the second representation or to the first representation,

[0039] Transforming the third representation into a third representation in spatial space, outputting and / or storing the third representation in spatial space and / or transmitting the third representation in spatial space to a separate computer system,

[0040] wherein the trained machine learning model was trained in a supervised learning process based on training data, wherein the training data for each reference person of a plurality of reference persons comprise (i) a first reference difference representation as input data and (ii) a second reference difference representation as target data, wherein the first reference difference representation represents a difference between a first reference representation and a second reference representation, and the second reference difference representation represents a difference between the second reference representation and a third reference representation or between the first reference representation and the third reference representation, wherein the first reference representation represents a liver or part of a liver of the reference person in a portal venous phase after application of the hepatobiliary contrast agent in the frequency domain.where the second reference representation represents the liver or part of the liver of the reference person in a transition phase after administration of the hepatobiliary contrast agent in the frequency domain, and the third reference representation represents the liver or part of the liver of the reference person in a hepatobiliary phase after administration of the hepatobiliary contrast agent in the frequency domain.

[0041] Another subject of the present disclosure is a kit comprising a hepatobiliary contrast agent and a computer program product comprising a computer program that can be loaded into the working memory of a computer system and causes the computer system to perform the following steps:

[0042] Providing an initial representation, wherein the initial representation represents a human liver or part of a liver in a portal venous phase following application of the hepatobiliary contrast agent in the frequency space,

[0043] Providing a second representation, wherein the second representation represents the liver or part of the liver of a person in a transitional phase after application of the hepatobiliary contrast agent in the frequency domain; generating a first difference representation based on the first representation and the second representation, wherein the first difference representation represents a difference between the first representation and the second representation.

[0044] Inputting the first difference representation into a trained machine learning model, receiving a second difference representation as output from the machine learning model,

[0045] Generating a third representation, wherein generating the third representation involves adding the second difference representation to the second representation or to the first representation,

[0046] Transforming the third representation into a third representation in spatial space, outputting and / or storing the third representation in spatial space and / or transmitting the third representation in spatial space to a separate computer system,

[0047] wherein the trained machine learning model was trained in a supervised learning process based on training data, wherein the training data for each reference person of a plurality of reference persons comprise (i) a first reference difference representation as input data and (ii) a second reference difference representation as target data, wherein the first reference difference representation represents a difference between a first reference representation and a second reference representation, and the second reference difference representation represents a difference between the second reference representation and a third reference representation or between the first reference representation and the third reference representation, wherein the first reference representation represents a liver or part of a liver of the reference person in a portal venous phase after application of the hepatobiliary contrast agent in the frequency domain,where the second reference representation represents the liver or part of the liver of the reference person in a transition phase after administration of the hepatobiliary contrast agent in the frequency domain, and the third reference representation represents the liver or part of the liver of the reference person in a hepatobiliary phase after administration of the hepatobiliary contrast agent in the frequency domain.

[0048] BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Fig. 1 shows an embodiment of the computer-implemented method of the present disclosure in the form of a flowchart.

[0050] Fig. 2 shows, by way of example and schematically, a method for training the machine learning model of the present disclosure.

[0051] Fig. 3 shows, by way of example and schematically, the use of the trained machine learning model for prediction.

[0052] Fig. 4 shows, by way of example and schematic representation, a computer system according to the present disclosure, which can be used to train the machine learning model and / or to use the trained machine learning model for prediction.

[0053] Fig. 5 shows, by way of example and schematic, another embodiment of the computer system of the present disclosure.

[0054] DETAILED DESCRIPTION

[0055] The subject matter of this disclosure is explained in more detail below, without distinguishing between the subject matter (method, computer system, computer program, kit). Rather, the following statements are intended to apply analogously to all subject matter (method, computer system, computer program, kit), regardless of the context in which they are made. If steps are specified in a sequence in this description or in the claims, this does not mean that the present disclosure is limited to the specified sequence. Rather, it is conceivable that the steps can also be carried out in a different sequence or even in parallel with one another, unless, for example, one step builds upon another, which requires that the building step be carried out subsequently (this will become clear in the specific case). The sequences mentioned are thus exemplary embodiments of the present disclosure.

[0056] The subject matter of this disclosure is further explained in some places with reference to drawings. These drawings depict specific embodiments with specific features and combinations of features, primarily for illustrative purposes; the drawings should not be understood as limiting this disclosure to the features and combinations of features shown in them. Furthermore, statements made in the description of the drawings with regard to features and combinations of features are intended to be generally applicable, that is, transferable to other embodiments and not limited to the embodiments shown. With the aid of this disclosure, representations of a human liver or part of a human liver can be predicted. Such a predicted representation is also referred to in this disclosure as a synthetic representation.The term "synthetic" means that the representation is not the (direct) result of a physical measurement on a real-world object (e.g., a person), but rather that the synthetic representation is generated using a machine learning model. A synonym for "synthetic" is "artificial." However, the synthetic representation is based on one or more measured representations; that is, the machine learning model can be configured and trained to generate a synthetic representation based on one or more measured representations (and / or optionally, additional data).

[0057] A representation of the liver or part of the liver is usually the result of a radiological examination. Radiology is the branch of medicine that deals with the application of primarily electromagnetic radiation and (including ultrasound diagnostics) mechanical waves for diagnostic, therapeutic, and / or scientific purposes. In addition to X-rays, other ionizing radiation such as gamma rays or electrons are also used. Since a key application is imaging, other imaging techniques such as sonography and magnetic resonance imaging (MRI) are also considered part of radiology, even though these techniques do not use ionizing radiation.The term "radiology" as used in this disclosure therefore includes in particular the following examination methods: computed tomography (CT), magnetic resonance imaging (MRI), sonography, positron emission tomography (PET).

[0058] In one embodiment of the present disclosure, the radiological examination is a magnetic resonance imaging or computed tomography examination. In one embodiment of the present disclosure, the radiological examination is a magnetic resonance imaging examination.

[0059] Magnetic resonance imaging, abbreviated MRI, is an imaging technique primarily used in medical diagnostics to visualize the structure and function of tissues and organs in the human or animal body. In MRI imaging, the magnetic moments of protons in a specimen are aligned within a background magnetic field, resulting in macroscopic magnetization along a longitudinal axis. This magnetization is then excited by the application of radiofrequency (RF) pulses. The return of the excited state to its resting position (relaxation), or the magnetization dynamics, is subsequently detected as relaxation signals using one or more RF receiver coils.

[0060] For spatial encoding, rapidly switched magnetic gradient fields are superimposed on the fundamental magnetic field. The acquired relaxation signals or the detected MRI data initially exist as raw data in the frequency domain (so-called k-space data) and can be transformed into spatial space (image space) by subsequent inverse Fourier transformation.

[0061] A representation within the meaning of the present disclosure may be an MRI scan, a CT scan, an ultrasound image, a PET scan or the like, or the representation may be generated from one or more MRI scans, CT scans, PET scans, ultrasound images or the like.

[0062] Contrast agents are frequently used in radiological examinations to enhance contrast. "Contrast agents" are substances or mixtures of substances that improve the visualization of bodily structures and functions during radiological examinations.

[0063] In computed tomography (CT), iodine-containing solutions are most commonly used as contrast agents. In magnetic resonance imaging (MRI), superparamagnetic substances (e.g., iron oxide nanoparticles, superparamagnetic iron-platinum particles (SIPPs)) or paramagnetic substances (e.g., gadolinium chelates, manganese chelates) are typically used as contrast agents. In ultrasound, fluids containing gas-filled microbubbles are usually administered intravenously. In positron emission tomography (PET), radiotracers are commonly used. Radiotracers contain a radioactive isotope and emit positrons during their decay. These positrons interact with electrons in the body, leading to the emission of gamma rays that can be detected by the PET scanner.

[0064] MRI contrast agents exert their effect by altering the relaxation times of the structures that absorb the contrast agent. Two groups of substances can be distinguished: para- and superparamagnetic agents. Both groups possess unpaired electrons that induce a magnetic field around the individual atoms or molecules. Superparamagnetic agents lead to a predominantly T2 shortening, while paramagnetic contrast agents primarily lead to a T1 shortening. The effect of these contrast agents is indirect, as the contrast agent itself does not generate a signal but only influences the signal intensity in its surroundings. An example of a superparamagnetic contrast agent is superparamagnetic iron oxide nanoparticles (SPIO). Examples of paramagnetic contrast agents include gadolinium chelates such as gadopentetate dimeglumine (trade name: Magnevist®, etc.).), Gadoteric acid (Dotarem®, Dotagita®, Cyclolux®), Gadodiamide (Omniscan®), Gadoteridol (ProHance®), Gadobutrol (Gadovist®) and Gadoxetic acid (Primovist® / Eovist®) .

[0065] In a first step, a first representation and a second representation are provided. The first and second representations can be transmitted, for example, by a CT scanner, an MRI scanner, an ultrasound device, or a PET scanner and received by the computer system of this disclosure. The first and second representations can be read from a data storage device. The data storage device can be a component of the computer system of this disclosure or connected to it, for example, via a network. The first and second representations can be generated based on corresponding spatial representations. The first representation represents a liver or part of a liver of a human being in the portal venous phase after administration of a hepatobiliary contrast agent in the frequency domain.

[0066] The second representation represents the liver or part of the liver of a person in a transitional phase after application of the hepatobiliary contrast agent in the frequency space.

[0067] A hepatobiliary contrast agent is typically characterized by its partial uptake by liver cells (hepatocytes), accumulation in functional tissue (parenchyma), and enhancement of contrast in healthy liver tissue before being excreted via bile into the feces. Examples of such gadoxetic acid-based contrast agents are described in US 6,039,931A; they are commercially available under brand names such as Primovist® and Eovist®. Another MRI contrast agent with lower hepatocyte uptake is gadobenate dimeglumine (Multihance®). Further hepatobiliary contrast agents are described, among other sources, in WO2022 / 194777.

[0068] Gadoxetate disodium (GD, Primovist®) is approved for use in liver MRI to detect and characterize lesions in patients with known or suspected focal liver disease. GD, with its lipophilic ethoxybenzyl moiety, exhibits a biphasic distribution: first, distribution in the intravascular and interstitial spaces after bolus injection, followed by selective uptake by hepatocytes. GD is excreted unchanged from the body in approximately equal amounts via the kidneys and the hepatobiliary pathway (50:50 dual excretion mechanism). Due to its selective accumulation in healthy liver tissue, GD is also referred to as a hepatobiliary contrast agent.

[0069] GD is approved at a dose of 0.1 ml / kg body weight (BW) (0.025 mmol / kg BW Gd). The recommended administration of GD involves an undiluted intravenous bolus injection at a flow rate of approximately 2 ml / second, followed by flushing the IV cannula with physiological saline. A standard protocol for liver imaging using GD consists of several planning and pre-contrast sequences. Following IV bolus injection of the contrast agent, dynamic images are typically acquired during the arterial (approximately 30 seconds post-injection, pi), portal venous (approximately 60 seconds pi), and transition phases (approximately 2–5 minutes pi). The transition phase typically already shows some increase in liver signal intensity due to the incipient uptake of the agent by hepatocytes.

[0070] In one embodiment, the hepatobiliary contrast agent is gadoxetate disodium.

[0071] In one embodiment of the present disclosure, the hepatobiliary contrast agent is an agent that contains a Gd 3+ -Complex of a compound of formula (I)

[0072]

[0073] (I) ,

[0074] includes, whereby

[0075] Ar a group selected from

[0076]

[0077] where # represents the link to X,

[0078] X represents a group consisting of

[0079] CH2, (CH2)2, (CH2)3, (CH2)4 and *-(CH2)2-O-CH2-# is selected

[0080] where * represents the link to Ar and # represents the link to the acetic acid residue,

[0081] R 1 , R 2 and R 3 independently represent a hydrogen atom or a group selected from C1-C3 alkyl, -CH2OH, -(CH2)2OH and -QEOCH3

[0082] R 4represents a group selected from C2-C4 alkoxy, (H3C-CH2)-O-(CH2)2-O-, (HsC-CH2)-O-(CH2)2-O-(CH2)2-O- and (H3C-CH2)-O-(CH2)2-O-(CH2)2-O-(CH2)2-O-

[0083] R 5 represents a hydrogen atom

[0084] and

[0085] R 6 represents a hydrogen atom

[0086] or a stereoisomer, tautomer, hydrate, solvate or salt thereof, or a mixture thereof.

[0087] In one embodiment of the present disclosure, the contrast agent is an agent that contains a Gd 3+ -Complex of a compound of formula (II)

[0088]

[0089] (II) ,

[0090] includes, whereby

[0091] Ar a group selected from

[0092] and

[0093]

[0094] represents

[0095] where # represents the link to X,

[0096] X represents a group selected from CH2, (CH2)2, (CH2)3, (CH2)4 and *-(CH2)2-O-CH2-#, where * represents the link to Ar and # represents the link to the acetic acid residue, R 7 represents a hydrogen atom or a group selected from Ci-Cs-alkyl, -CH2OH,-(CH2)2OH and -CH2OCH3;

[0097] R 8 a group selected from

[0098] C2-C4-alkoxy, (H3C-CH2O)-(CH2)2-O-, (H3C-CH2O)-(CH2)2-O-(CH2)2-O- and

[0099] (H3C-CH2O)-(CH2)2-O-(CH2)2-O-(CH2)2-O- represents;

[0100] R 9 and R 10 represent a hydrogen atom independently of each other;

[0101] or a stereoisomer, tautomer, hydrate, solvate or salt thereof, or a mixture thereof.

[0102] The term "C1-C3 alkyl" refers to a linear or branched, saturated, monovalent hydrocarbon group with 1, 2, or 3 carbon atoms, e.g., methyl, ethyl, n-propyl, and isopropyl. The term "C2-C4 alkyl" refers to a linear or branched, saturated, monovalent hydrocarbon group with 2, 3, or 4 carbon atoms.

[0103] The term "C2-C4-alkoxy" means a linear or branched, saturated, monovalent group of the formula (C2-C4-alkyl)-O-, in which the term "C2-C4-alkyl" is defined as above, e.g. a methoxy, ethoxy, n-propoxy or isopropoxy group.

[0104] In one embodiment of the present disclosure, the contrast agent is an agent comprising gadolinium 2,2',2"-(10-{l-carboxy-2-[2-(4-ethoxyphenyl)ethoxy]ethyl}-l,4,7,10-tetraazacy clododecan-1, 4, 7-triyl)triacetate (see e.g. WO2022 / 194777, Example 1).

[0105] In one embodiment of the present disclosure, the contrast agent is an agent comprising gadolinium 2,2',2"-{10-[l-carboxy-2-{4-[2-(2-ethoxyethoxy)ethoxy]phenyl}ethyl]-l,4,7,10-tetraazacyclododecane-l,4,7-triyl}triacetate (see, e.g., WO2022 / 194777, Example 2). In another embodiment of the present disclosure, the contrast agent is an agent comprising gadolinium 2,2',2"-{10-[(lR)-l-carboxy-2-{4-[2-(2-ethoxyethoxy)ethoxy]phenyl}ethyl]-l,4,7,10-tetraazacyclododecane-l,4,7-triyl}triacetate (see, e.g., WO2022 / 194777, Example 4). In one embodiment of the present disclosure, the contrast agent is an agent comprising gadolinium (2S,2'S,2"S)-2,2',2"-{ 10-[(lS)-l-carboxy-4-{4-[2-(2-ethoxyethoxy)ethoxy] phenyl}butyl] - 1 ,4,7, 10-tetraazacyclododecane- 1 ,4,7-triyl}tris(3-hydroxypropanoate) (see e.g. WO2022 / 194777, Example 15).

[0106] In one embodiment of the present disclosure, the contrast agent is an agent comprising gadolinium 2,2',2"-{10-[(lS)-4-(4-butoxyphenyl)-l-carboxybutyl]-l,4,7,10-tetraazacy clododecan-1, 4, 7-triyl}triacetate (see e.g. WO2022 / 194777, Example 31).

[0107] The first representation depicts a human liver or part of a liver in the portal venous phase after administration of a hepatobiliary contrast agent. The second representation depicts the human liver or part of the liver in a transitional phase after administration of the hepatobiliary contrast agent.

[0108] The hepatobiliary contrast agent is usually administered intravenously, for example, into a vein in the arm. After intravenous administration of the hepatobiliary contrast agent as a bolus into an arm vein, the contrast agent first reaches the liver via the arteries. These are shown with contrast enhancement in the corresponding radiological images. The phase in which the hepatic arteries appear with contrast enhancement in the radiological images is called the "arterial phase."

[0109] The contrast agent then reaches the liver via the hepatic veins. While the contrast in the hepatic arteries is already decreasing, it reaches its maximum in the hepatic veins. The phase in which the hepatic veins appear contrast-enhanced in radiological images is called the "portal venous phase." This phase can begin during the arterial phase and overlap with it.

[0110] The portal venous phase is followed by the transitional phase, in which the contrast in the hepatic arteries and hepatic veins continues to decrease. When a hepatobiliary contrast agent is used, the contrast in healthy liver cells gradually increases during the transitional phase.

[0111] The arterial phase, the portal venous phase and the transition phase are collectively referred to as the "dynamic phase".

[0112] Ten to twenty minutes after injection, a hepatobiliary contrast agent leads to a marked increase in signal intensity in healthy liver parenchyma. This phase is called the "hepatobiliary phase." The contrast agent is eliminated from the liver cells only slowly; accordingly, the hepatobiliary phase can last two hours or more.

[0113] The aforementioned phases are described in more detail 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 / jjhep.2019.05.005).

[0114] The first and second representations represent a human liver or part of a liver in the frequency domain. These representations can be derived, for example, from corresponding spatial representations via Fourier transformation, or they can be MRI data acquired as so-called k-space data.

[0115] In a spatial representation, the area under investigation (e.g., a liver or part of a human liver) is typically represented by a multitude of image elements (e.g., pixels or voxels), which may be arranged in a grid, with each image element representing a portion of the area under investigation. A widely used format in radiology for storing and processing spatial representations is the DICOM format. DICOM (Digital Imaging and Communications in Medicine) is an open standard for storing and exchanging information in medical image data management.

[0116] In a frequency-space representation, also referred to in this description as a frequency-space representation, the area under investigation is represented by a superposition of fundamental oscillations. For example, the area under investigation can be represented by a sum of sine and cosine functions with different amplitudes, frequencies, and phases. The amplitudes and phases can be plotted as a function of the frequencies, for example, in a two- or three-dimensional representation. Typically, the lowest frequency (origin) is placed at the center. The further one moves away from this center, the higher the frequencies. Each frequency can be assigned an amplitude, which represents the frequency in the frequency-space representation, and a phase, which indicates how far the respective oscillation is shifted relative to a sine or cosine oscillation.

[0117] A representation in spatial space can be transformed into a representation in frequency space, for example, by a Fourier transform. Conversely, a representation in frequency space can be transformed into a representation in spatial space, for example, by an inverse Fourier transform.

[0118] Details about spatial representations and frequency space representations and their respective conversions are described in numerous publications, see e.g.: https: / / see.stanford.edu / materials / lsoftaee261 / book-fall-07.pdf.

[0119] The first representation may be or include an MRI scan of a liver or part of a liver of a person after application of a hepatobiliary MRI contrast agent, or may be obtained from such a scan (e.g. by Fourier transformation based on a corresponding spatial representation).

[0120] The first representation may be or include a CT scan of a liver or part of a liver of a person after application of a hepatobiliary MRI contrast agent, or may be derived from such a scan (e.g. by Fourier transformation based on a corresponding spatial representation).

[0121] Likewise, the second representation can also be or include an MRI or CT scan of a liver or part of a liver of a person after application of a hepatobiliary MRI contrast agent, or be derived from such a scan (e.g. by Fourier transformation based on a corresponding spatial representation).

[0122] In a further step, a first difference representation is generated based on the first and second representations. This first difference representation represents a difference between the first and second representations. In other words, the first difference representation represents differences between the first and second representations. Generating the first difference representation can, for example, involve subtracting the first representation from the second representation.

[0123] In principle, it is also possible to capture differences between the first and second representations not by subtraction, but, for example, by division. For the sake of simplicity, the aspects of this disclosure are described using differences, without, however, limiting the disclosure to this form of execution. Those skilled in image processing are aware that differences between pictorial representations can also be determined and expressed in ways other than subtraction (e.g., by division). This is intended to ensure that other methods and representations that capture and represent differences between two representations also fall within the scope of this disclosure.

[0124] It is possible that the first and / or the second representation and / or the first difference representation are filtered using one or more filters. Examples of such filters are frequency-dependent weights as described, for example, in W02024100234A1 or EP24182725.2. However, other filters are also conceivable.

[0125] The first difference representation is fed into a trained machine learning model. The machine learning model is configured and trained to generate a second difference representation based on the first difference representation and model parameters (and optionally further input data).

[0126] The second difference representation is a difference representation predicted by the trained machine learning model based on the first difference representation (and optionally additional input data). The second difference representation is a synthetic representation. It represents a difference (or differences) between the second representation of the human liver or part of the liver and a third representation, or between the first representation of the human liver or part of the liver and a third representation (depending on what was used for training). The third representation is a representation of the human liver or part of the liver in a hepatobiliary phase after application of the hepatobiliary contrast agent in the frequency domain.The third representation is a hypothetical, theoretical, or assumed representation because, unlike the first and second representations, it was not measured. Instead, it is generated using the second difference representation, which was also not measured but predicted (as described below). A "machine learning model" can be understood as a computer-implemented data processing architecture. The model can receive input data and provide output data based on this input data and model parameters. Through training, the model can learn a relationship between the input data and the output data. During training, model parameters can be adjusted to produce a desired output for a given input.

[0127] When training such a model, it is presented with training data from which it can learn. The trained machine learning model is the result of the training process. In addition to input data, the training data includes the correct output data (target data) that the model is to generate based on the input data. During training, patterns are recognized that map the input data to the target data.

[0128] During the training process, the input data for the training data is fed into the model, and the model generates output data. This output data is then compared to the target data. Model parameters are modified to reduce the deviations between the output and target data to a (defined) minimum. To modify the model parameters with a view to reducing these deviations, an optimization method such as gradient descent can be used.

[0129] The deviations can be quantified using a loss function. Such a function can be used to calculate the loss for a given pair of output and target data. The goal of the training process can be to modify (adjust) the parameters of the machine learning model so that the loss is reduced to a (defined) minimum for all pairs in the training dataset.

[0130] If the output and target data are numbers, for example, the error function can be the absolute difference between these numbers. In this case, a large absolute error may mean that one or more model parameters need to be changed significantly.

[0131] For output data in the form of vectors, difference metrics between vectors such as the mean squared error, a cosine distance, a norm of the difference vector such as a Euclidean distance, a Chebyshev distance, an Lp norm of a difference vector, a weighted norm, or another type of difference metric of two vectors can be chosen as the error function.

[0132] For higher-dimensional outputs, such as two-dimensional, three-dimensional, or even higher-dimensional outputs, an element-wise difference metric can be used. Alternatively or additionally, the output data can be transformed before calculating an error value, for example, into a one-dimensional vector.

[0133] In the present case, the training data for training the machine learning model, for each reference person of a plurality of reference persons, comprise (i) a first reference difference representation as input data and (ii) a second reference difference representation as target data.

[0134] The term "much" usually means more than 10, e.g. more than 100 or more than 500.

[0135] The first reference-difference representation represents a difference (differences) between a first reference representation and a second reference representation. The second reference-difference representation represents (i) a difference (differences) between the second reference representation and a third reference representation, or (ii) a difference (differences) between the first reference representation and a third reference representation.

[0136] The first reference representation represents the liver or part of the liver of the reference subject in the portal venous phase after administration of the hepatobiliary contrast agent in the frequency domain. The second reference representation represents the liver or part of the liver of the reference subject in a transition phase after administration of the hepatobiliary contrast agent in the frequency domain. The third reference representation represents the liver or part of the liver of the reference subject in the hepatobiliary phase after administration of the hepatobiliary contrast agent in the frequency domain.

[0137] The term “reference” is used in several places in this disclosure to distinguish the data used to train and / or validate the machine learning model from the data used to generate a synthetic representation using the trained machine learning model. However, the term “reference” is not to be understood in any other restrictive sense; this distinction is made solely to avoid an objection of lack of clarity in patent granting proceedings.

[0138] The reference person is a human being. The first reference representation has the same properties as the first representation described in this revelation. The second reference representation has the same properties as the second representation described in this revelation. The first reference difference representation has the same properties as the first difference representation described in this revelation.

[0139] In addition to the aforementioned input data, further data can be used as input for training the machine learning model (and for post-training prediction). This can include data about the subject of the study (the human), such as age, sex, height, weight, diagnosed diseases, lifestyle habits (e.g., alcohol consumption), and / or other / additional data. It can also include further representations of the liver or a portion of the liver, such as a representation before the administration of the hepatobiliary contrast agent (native phase).

[0140] Training the machine learning model typically involves the following steps: for each reference person out of a multitude of reference persons:

[0141] Inputting the first reference difference representation (and optionally further input data) into the machine learning model,

[0142] Receiving a predicted second reference difference representation as output from the machine learning model,

[0143] Quantifying the deviation between the second reference difference representation and the predicted second reference difference representation,

[0144] Reducing the deviation by modifying model parameters of the machine learning model.

[0145] The modification (adjustment) of the model parameters is usually carried out in an optimization procedure such as a gradient descent method.

[0146] The training of the machine learning model can be terminated when a stop criterion is met. Such a stop criterion could be, for example: a predefined maximum number of training steps / cycles / epochs has been performed; deviations between output data and target data can no longer be reduced by modifying the model parameters; a predefined minimum of the error function is reached; and / or an extreme value (e.g., maximum or minimum) of another performance metric is reached.

[0147] When training a machine learning model, the training dataset is typically divided into several subsets to ensure robust evaluation and generalization. The most common sets include a training set, a validation set, and a test set. The training set is used to train the machine learning model. The validation set can be used to fine-tune hyperparameters and make decisions regarding the model architecture. It can also help evaluate the model's performance during training, allowing for the monitoring of overfitting and the implementation of adjustments to improve model performance. The test set is usually kept separate from the training and validation datasets and is used to evaluate the model's final performance after training is complete.There are various partitioning strategies for partitioning the data, e.g. random partitioning, k-fold cross-validation, and stratified partitioning.

[0148] The machine learning model can be, or include, an artificial neural network.

[0149] The artificial neural network can have an autoencoder architecture; for example, the artificial neural network can have an architecture like the U-Net (see, for example, O. Ronneberger et al. U-net: Convolutional networks for biomedical image segmentation, International Conference on Medical image computing and computer-assisted intervention, pages 234-241, Springer, 2015, https: / / doi.org / 10.1007 / 978-3-319-24574-4_28).

[0150] The artificial neural network can be a Generative Adversarial Network (GAN) or include one (see, e.g., M.-Y. Liu et al.: Generative Adversarial Networks for Image and Video Synthesis: Algorithms and Applications, arXiv:2008.02793; J. Henry et al.: Pix2Pix GAN for Image-to-Image Translation, DOI: 10.13140 / RG.2.2.32286.66887).

[0151] The artificial neural network can be or include a Regularized Generative Adversarial Network (see e.g. Q. Li et al. : RegGAN: An End-to-End Network for Building Footprint Generation with Boundary Regularization, Remote Sens. 2022, 14, 1835).

[0152] The artificial neural network can be a Conditional Adversarial Network or include one (see e.g. P. Isola et al. : Image-to-Image Translation with Conditional Adversarial Networks, arXiv: 1611.07004 [cs.CV]).

[0153] The artificial neural network can be a transformer network or include one (see e.g. D. Karimi et al. Convolution-Free Medical Image Segmentation using Transformers, arXiv:2102.13645 [eess.IV]).

[0154] Once the machine learning model is trained, it can be used for prediction. It can be used to generate a synthetic representation for a new person. The term "new" means that representations of the new person have not typically been used in the training.

[0155] As described above, the first difference representation (and optionally further input data) is fed into the trained machine learning model. Based on this first difference representation (and optionally the further input data), the trained machine learning model generates a (predicted, synthetic) second difference representation.

[0156] In a further step, a third representation is generated based on the second difference representation produced by the trained machine learning model and the existing second or first representation. Generating the third representation typically involves adding the second difference representation to the second representation or adding the second difference representation to the first representation.

[0157] The third representation is a synthetic representation, as it is based on the synthetic second difference representation. The third representation represents the liver or part of the liver of a person in a hepatobiliary phase after application of the hepatobiliary contrast agent in the frequency domain.

[0158] In a further step, a corresponding third spatial representation can be generated based on the third frequency-space representation. The third spatial representation can be generated, for example, by inverse Fourier transforming the third frequency-space representation. The third spatial representation is a synthetic representation because it is based on the synthetic second difference representation. The third spatial representation represents the liver, or the portion of the liver, of a person in the hepatobiliary phase after administration of the hepatobiliary contrast agent in spatial space.

[0159] The third spatial representation can be output (e.g. displayed on a monitor and / or printed on a printer), stored in a data storage device and / or transmitted to a separate computer system.

[0160] Generating a synthetic representation of the liver, or a portion thereof, in the hepatobiliary phase after administration of a hepatobiliary contrast agent has the advantage of limiting the potentially strenuous radiological examination to the portal venous and transitional phases. This means, for example, that the patient spends less time in the tomograph. The quality of the synthetic representation is higher compared to synthetic representations generated by other methods. In particular, the synthetic representation can exhibit a higher signal-to-noise ratio (SNR), improved spatial resolution, and / or higher contrast sharpness. A higher signal-to-noise ratio can improve the detectability of lesions and increase diagnostic accuracy.

[0161] Further embodiments are disclosed below. These embodiments are not necessarily subject matter that falls under patent protection. As is known to those skilled in the art in patent law, the scope of protection of a patent is defined by the patent claims. The description and the drawings are to be used to interpret the patent claims. The embodiments described below are part of the description and not of the patent claims. The following embodiments are intended to give the reader guidance on how various features described in this disclosure can be combined. They are therefore part of the present technical teaching and should not be confused with the subject matter of the patent claims.

[0162] Further embodiments of the present disclosure are:

[0163] Embodiment 1: Computer-implemented method comprising the steps of: providing a first representation, wherein the first representation represents a human liver or part of a liver in a portal venous phase after administration of a hepatobiliary contrast agent in the frequency domain; providing a second representation, wherein the second representation represents the human liver or part of the liver in a transition phase after administration of the hepatobiliary contrast agent in the frequency domain; generating a first difference representation based on the first representation and the second representation, wherein the first difference representation represents a difference between the first representation and the second representation; inputting the first difference representation into a trained machine learning model.Receiving a second difference representation as output from the trained machine learning model, generating a third representation, wherein generating the third representation includes adding the second difference representation to the second representation or to the first representation, transforming the third representation into a third representation in space, outputting and / or storing the third representation in space and / or transmitting the third representation in space to a separate computer system, wherein the trained machine learning model was trained in a supervised learning process based on training data, wherein the training data for each reference person of a plurality of reference persons comprise (i) a first reference difference representation as input data and (ii) a second reference difference representation as target data.wherein the first reference difference representation represents a difference between a first reference representation and a second reference representation, and the second reference difference representation represents a difference between the second reference representation and a third reference representation or between the first reference representation and the third reference representation, wherein the first reference representation represents a liver or part of a liver of the reference person in a portal venous phase after application of the hepatobiliary contrast agent in the frequency space,where the second reference representation represents the liver or part of the liver of the reference person in a transition phase after administration of the hepatobiliary contrast agent in the frequency space, and the third reference representation represents the liver or part of the liver of the reference person in a hepatobiliary phase after administration of the hepatobiliary contrast agent in the frequency space.

[0164] Embodiment 2: Method according to embodiment 1, wherein generating the first difference representation includes subtracting the first representation from the second representation.

[0165] Implementation form 3: Method according to embodiment 1 or 2, wherein generating the third representation comprises adding the second difference representation to the second representation, and wherein the second reference difference representation was generated on the basis of the second reference representation and the third reference representation, wherein generating the second reference difference representation comprised subtracting the second reference representation from the third reference representation.

[0166] Embodiment 4: Method according to one of embodiments 1 to 3, wherein the first representation and the second representation are the result of a radiological examination on the human being.

[0167] Embodiment 5: Method according to any one of embodiments 1 to 3, further comprising: generating the first representation from a first spatial representation, wherein the first spatial representation represents the liver or part of the liver of the human in the portal venous phase after application of the hepatobiliary contrast medium in spatial space; generating the second representation from a second spatial representation, wherein the second spatial representation represents the liver or part of the liver of the human in the transition phase after application of the hepatobiliary contrast medium in spatial space; wherein the first spatial representation and the second spatial representation are the result of a radiological examination of the human.

[0168] Embodiment 6: Method according to one of embodiments 1 to 4, wherein the first reference representation and the second reference representation of each reference person are the result of a radiological examination of the reference person.

[0169] Embodiment 7: Method according to embodiment 5, wherein the first reference representation in the frequency domain was generated based on a first reference representation in spatial space, wherein the first reference representation in spatial space represents the liver or part of the liver of the reference person in the portal venous phase after application of the hepatobiliary contrast agent in spatial space, wherein the second reference representation in the frequency domain was generated based on a second reference representation in spatial space, wherein the second reference representation in spatial space represents the liver or part of the liver of the reference person in the transition phase after application of the hepatobiliary contrast agent in spatial space, wherein the third reference representation in the frequency domain was generated based on a third reference representation in spatial space.wherein the third reference representation in local space represents the liver or part of the liver of the reference person in the hepatobiliary phase after application of the hepatobiliary contrast medium in local space, wherein the first reference representation in local space, the second reference representation in local space, and the third reference representation in local space of each reference person are the result of a radiological examination on the reference person.

[0170] Embodiment 8: Method according to one of embodiments 4 to 7, wherein the radiological examination is or comprises an MRI examination.

[0171] Implementation form 9: Procedure according to one of implementation forms 4 to 8, wherein the radiological examination is or includes a CT examination.

[0172] Implementation form 10: Procedure according to one of implementation forms 1 to 9, wherein the hepatobiliary contrast agent is or comprises an MRI contrast agent.

[0173] Implementation form 11: Procedure according to one of implementation forms 1 to 10, wherein the hepatobiliary contrast medium comprises gadoxetate disodium.

[0174] Design 12: Procedure according to one of design 1 to 10, wherein the hepatobiliary contrast agent comprises gadobenate dimeglumine.

[0175] Implementation form 13: Procedure according to one of implementation forms 1 to 12, wherein the trained machine learning model is or comprises an artificial neural network.

[0176] Implementation form 14: Method according to implementation form 13, wherein the artificial neural network has an autoencoder architecture.

[0177] Embodiment 15: Method according to embodiment 13 or 14, wherein the artificial neural network has a U-Net architecture.

[0178] Embodiment 16: Method according to any of embodiments 13 to 15, wherein the artificial neural network is or comprises a Generative Adversarial Network (GAN).

[0179] Embodiment 17: Method according to any of embodiments 13 to 16, wherein the artificial neural network is or comprises a Conditional Adversarial Network.

[0180] Embodiment 18: Method according to any of embodiments 13 to 17, wherein the artificial neural network is or comprises a transformer network.

[0181] Embodiment 19: A method according to any one of embodiments 1 to 18, wherein training the machine learning model comprises: for each reference person of the plurality of reference persons: inputting the first reference difference representation into the machine learning model, receiving a predicted second reference difference representation as output from the machine learning model, quantifying any deviation between the second reference difference representation and the predicted second reference difference representation, and reducing the deviation by modifying model parameters of the machine learning model. Embodiment 20: A method according to embodiment 19, wherein the deviation is quantified using an error function.

[0182] Embodiment 21: Method according to embodiment 19 or 20, wherein the modification of the model parameters is carried out in a gradient descent method.

[0183] Embodiment 22: Method according to one of embodiments 1 to 21, wherein the transformation of the third representation into a third representation in spatial space comprises an inverse Fourier transformation.

[0184] Embodiment 23: Method according to one of embodiments 1 to 22, wherein the first and / or the second representation and / or the first difference representation are filtered by means of one or more filters.

[0185] Embodiment 24: Method according to embodiment 23, wherein the filters comprise frequency-dependent weights. Embodiment 25: Method according to any one of embodiments 1 to 24, wherein the synthetic third representation has a higher signal-to-noise ratio than synthetic representations generated by other methods.

[0186] Implementation form 26: Method according to one of implementation forms 1 to 25, wherein the synthetic third representation has improved spatial resolution and / or higher contrast sharpness.

[0187] Implementation form 27: Method according to one of embodiments 1 to 26, wherein the method is executed as a cloud-based method in which the trained machine learning model is executed on one or more remote servers.

[0188] Embodiment 28: Method according to embodiment 27, wherein the first representation and the second representation are transmitted from a local computer system to the remote server via a network, and the third representation is transmitted back in local space from the remote server to the local computer system.

[0189] Embodiment 29: Method according to one of embodiments 1 to 28, wherein the method is performed as real-time processing, in which the synthetic third representation is generated during the ongoing radiological examination.

[0190] Embodiment 30: Method according to embodiment 29, wherein the trained machine learning model is executed on optimized hardware, the optimized hardware comprising one or more GPUs, TPUs and / or specialized AI accelerators.

[0191] Embodiment 31: Method according to one of embodiments 1 to 30, further comprising: performing an automatic segmentation of the liver and / or lesions based on the third representation.

[0192] Embodiment 32: Method according to one of embodiments 1 to 31, further comprising: automatic detection, classification and / or quantification of lesions based on the third representation using a further machine learning model.

[0193] Embodiment 33: Method according to embodiment 32, wherein the automatic detection comprises automatic marking of suspicious areas in the third representation.

[0194] Embodiment 34: Method according to embodiment 32 or 33, wherein the classification includes a classification of lesions as benign or malignant.

[0195] Embodiment 35: Method according to one of embodiments 32 to 34, wherein the quantification includes determining the size of lesions.

[0196] Embodiment 36: Method according to one of embodiments 1 to 35, further comprising: automatic quality control of the synthetic third representation, wherein a further machine learning model evaluates the quality and / or plausibility of the generated synthetic representation.

[0197] Embodiment 37: Method according to one of embodiments 1 to 36, wherein the training data includes data from more than 100 reference persons.

[0198] Embodiment 38: Method according to one of embodiments 1 to 37, wherein the training data includes data from more than 500 reference persons.

[0199] Embodiment 39: A method according to any one of embodiments 1 to 38, wherein the training data includes further data on the subject of the study, wherein the further data includes one or more of the following: age, sex, height, weight, diagnosed diseases, and lifestyle habits. Embodiment 40: A method according to any one of embodiments 1 to 39, wherein the training data further includes a representation of the liver or part of the liver before the application of the hepatobiliary contrast agent.

[0200] Implementation form 41: Computer system comprising a receiving unit, a control and computing unit and an output unit, wherein the control and computing unit is configured to execute the method according to one of embodiments 1 to 40.

[0201] Implementation form 42: Computer system according to embodiment 41, wherein the computer system is implemented as a laptop, notebook, netbook and / or tablet PC.

[0202] Embodiment 43: Computer system according to embodiment 41 or 42, wherein the computer system is a component of an MRI scanner, a CT scanner or an ultrasound diagnostic device.

[0203] Embodiment 44: Computer system according to one of embodiments 41 to 43, wherein the computer system is implemented as a cloud-based system in which the trained machine learning model is executed on one or more remote servers.

[0204] Embodiment 45: Computer system according to any one of embodiments 41 to 44, wherein the computer system comprises one or more GPUs, TPUs and / or specialized AI accelerators. Embodiment 46: Computer system according to any one of embodiments 41 to 45, wherein the computer system comprises a processing unit and a memory, the memory containing the trained machine learning model.

[0205] Embodiment 47: Computer system according to any of embodiments 41 to 46, wherein the computer system comprises one or more communication interfaces configured to send and / or receive information to and / or from an MRI scanner, a CT scanner, other computer systems, networks and / or data storage devices.

[0206] Embodiment 48: Computer program product comprising a computer program that can be loaded into a working memory of a computer system and causes the computer system to execute the method according to one of embodiments 1 to 40.

[0207] Embodiment 49: Computer program product according to embodiment 48, wherein the computer program product comprises a non-volatile data carrier on which the computer program is stored.

[0208] Embodiment 50: Kit comprising a hepatobiliary contrast agent and a computer program product according to embodiment 48 or 49.

[0209] Embodiment 51: Kit according to embodiment 50, wherein the hepatobiliary contrast agent comprises gadoxetate disodium.

[0210] Embodiment 52: Kit according to embodiment 50 or 51, wherein the kit further comprises means that allow a purchaser to obtain the computer program, wherein the means comprise a link and / or a code.

[0211] Embodiment 53: Method for training a machine learning model to predict a synthetic representation of a human liver or part of a liver in a hepatobiliary phase after administration of a hepatobiliary contrast agent, comprising: providing training data, wherein the training data for each reference person of a plurality of reference persons comprise (i) a first reference difference representation as input data and (ii) a second reference difference representation as target data, wherein the first reference difference representation represents a difference between a first reference representation and a second reference representation, and the second reference difference representation represents a difference between the second reference representation and a third reference representation or between the first reference representation and the third reference representation.wherein the first reference representation represents a liver or part of a liver of the reference subject in a portal venous phase after administration of the hepatobiliary contrast agent in the frequency domain, wherein the second reference representation represents the liver or part of the liver of the reference subject in a transition phase after administration of the hepatobiliary contrast agent in the frequency domain, and wherein the third reference representation represents the liver or part of the liver of the reference subject in a hepatobiliary phase after administration of the hepatobiliary contrast agent in the frequency domain, for each reference subject of the plurality of reference subjects: inputting the first reference difference representation into the machine learning model, receiving a predicted second reference difference representation as output from the machine learning model,Quantifying a deviation between the second reference difference representation and the predicted second reference difference representation, reducing the deviation by modifying model parameters of the machine learning model. Embodiment 54: Method according to embodiment 53, wherein the machine learning model is or comprises an artificial neural network.

[0212] Embodiment 55: Method according to embodiment 53 or 54, wherein the deviation is quantified using an error function.

[0213] Embodiment 56: Method according to one of embodiments 53 to 55, wherein the reduction of the deviation is carried out in a gradient descent method.

[0214] Embodiment 57: Method according to one of embodiments 53 to 56, wherein the training is terminated when a stop criterion is met, the stop criterion comprising one or more of a predetermined maximum number of training steps, a predetermined minimum of the error function and an extreme value of a performance value.

[0215] Embodiment 58: Trained machine learning model trained according to the method of one of embodiments 53 to 57.

[0216] Embodiment 59: Use of a trained machine learning model to predict a synthetic representation of a liver or part of a liver of a human in a hepatobiliary phase after application of a hepatobiliary contrast agent, wherein the trained machine learning model is configured to generate a second difference representation based on a first difference representation, wherein the first difference representation represents a difference between a first representation and a second representation, wherein the first representation represents a liver or part of a liver of a human in a portal venous phase after application of a hepatobiliary contrast agent in the frequency domain, and the second representation represents the liver or part of the liver of the human in a transition phase after application of the hepatobiliary contrast agent in the frequency domain.

[0217] Embodiment 60: Method according to one of embodiments 1 to 40, wherein the third representation in spatial space is displayed on a monitor.

[0218] Embodiment 61: Method according to one of embodiments 1 to 40 or 60, wherein the third representation in spatial space is printed on a printer.

[0219] Embodiment 62: Method according to one of embodiments 1 to 40, 60 or 61, wherein the third representation in spatial space is stored in DICOM format.

[0220] Embodiment 63: A method according to any one of embodiments 1 to 40, wherein the method is used to assist in the detection and / or differential diagnosis of focal liver lesions. Embodiment 64: A method according to embodiment 63, wherein the focal liver lesions comprise metastases and / or hepatocellular carcinomas.

[0221] Embodiment 65: Method according to one of embodiments 1 to 40, wherein the method is used to assist in therapy planning.-TI-

[0222] Embodiment 66: Method according to one of embodiments 1 to 40, wherein the portal venous phase begins approximately 60 seconds after the application of the hepatobiliary contrast medium.

[0223] Implementation form 67: Method according to one of embodiments 1 to 40, wherein the transition phase begins approximately 2 to 5 minutes after the application of the hepatobiliary contrast medium.

[0224] Embodiment 68: A method according to any one of embodiments 1 to 40, wherein the hepatobiliary phase begins approximately 10 to 20 minutes after administration of the hepatobiliary contrast medium. Embodiment 69: A method according to any one of embodiments 1 to 40, wherein the hepatobiliary contrast medium is administered intravenously as a bolus injection.

[0225] Embodiment 70: Method according to one of embodiments 1 to 40, wherein the first representation and the second representation are TI-weighted MRI scans or are based on Tl-weighted MRI scans.

[0226] The subject matter of this disclosure is explained in more detail below with reference to the drawings, without limiting the disclosure to the features and combinations of features shown in the drawings. Statements made in relation to the embodiments shown in the drawings are intended to be transferable to other embodiments to the extent technically possible. Features are intended to be combinable in any way, to the extent technically possible.

[0227] Fig. 1 shows an embodiment of the computer-implemented method of the present disclosure in the form of a flowchart.

[0228] The procedure (100) comprises the following steps:

[0229] (110) Providing a first representation, wherein the first representation represents a liver or part of a liver of a human being in a portal venous phase following administration of a hepatobiliary contrast agent in the frequency space,

[0230] (120) Providing a second representation, wherein the second representation represents the liver or part of the liver of a person in a transitional phase after application of the hepatobiliary contrast agent in the frequency space,

[0231] (130) Generating a first difference representation based on the first representation and the second representation, wherein the first difference representation represents a difference between the first representation and the second representation,

[0232] (140) Inputting the first difference representation into a trained machine learning model, (150) Receiving a second difference representation as output from the machine learning model,

[0233] (160) Generating a third representation, wherein generating the third representation includes adding the second difference representation to the second representation or to the first representation,

[0234] (170) Transforming the third representation into a third representation in spatial space, (180) Outputting and / or storing the third representation in spatial space and / or transmitting the third representation in spatial space to a separate computer system.

[0235] Figure 2 shows, by way of example and schematically, a method for training the machine learning model of the present disclosure. The machine learning model (MLM) is trained on the basis of training data. The training data comprises, for each reference person of a plurality of reference persons, (i) a first reference difference representation (RRD). 1 ) as input data and (ii) a second reference difference representation (RRD) 2 ) as target data.

[0236] In Fig. 2, only one data set of a reference person comprises a first reference difference representation (RRD). 1 ) and a second reference difference representation (RRD) 2 ) shown. The training data typically comprises a large number of such datasets.

[0237] During training, the first reference difference representation (RRD) is created. 1The data is fed into the machine learning model (MLM). The model is configured and trained based on the first reference difference representation (RRD). 1 ) and based on model parameters (MP) (and optionally based on further input data) a predicted second reference difference representation (RRD) 2 *) TO generate. The predicted second reference-difference representation (RRD) 2 *) is combined with the second reference difference representation (RRD). 2 ) compared. An error function (LF) is used to account for deviations between the predicted second reference difference representation (RRD). 2 *) and the second reference difference representation (RRD) 2 ) TO quantify. In an optimization procedure (e.g., a gradient descent method), the deviations are reduced by modifying the model parameters (MP).

[0238] The steps shown in Fig. 2 are performed for a variety of datasets. The training of the machine learning model (MLM) can be terminated when a stop criterion is met. Such a stop criterion could be, for example: a predefined maximum number of training steps / cycles / epochs has been performed; deviations between the predicted second reference difference representation and the second reference difference representation can no longer be reduced by modifying the model parameters; a predefined minimum of the error function is reached; and / or an extreme value (e.g., maximum or minimum) of another performance metric is reached.

[0239] Once the training is complete, the trained machine learning model can be used for prediction. This is shown schematically and by way of example in Fig. 3.

[0240] Figure 2 shows how the training data can be obtained, in addition to the training of the machine learning model (MLM). This, like the training, is demonstrated using a reference person.

[0241] The starting point shown in Fig. 2 consists of three reference representations, a first reference representation (RRi) 1 ), a second reference representation (RR2) 1 ) and a third reference representation (RR3) 1 ).

[0242] The first reference representation (RR1) 1 ) represents a liver or part of a liver of the reference subject in a portal venous phase after administration of the hepatobiliary contrast agent in local space. The second reference representation (RR2) 1 ) represents the liver or part of the liver of the reference subject in a transitional phase after administration of the hepatobiliary contrast agent in local space. The third reference representation (RR3) 1) represents the liver or part of the liver of the reference person in a hepatobiliary phase after application of the hepatobiliary contrast agent in local space.

[0243] From the first reference representation (RR1) 1 ) is transformed, for example, by a Fourier transform, into a first reference representation (RRi) F ) generated in the frequency domain. The first reference representation (RRÜ) represents the liver or part of the liver of the reference subject in the portal venous phase after administration of the hepatobiliary contrast agent in the frequency domain.

[0244] From the second reference representation (RR2) 1 ) is transformed, for example, by a Fourier transform, into a second reference representation (RR2). F ) in the frequency domain. The second reference representation (RR2) F) represents the liver or part of the liver of the reference person in the transition phase after administration of the hepatobiliary contrast agent in the frequency domain. From the third reference representation (RR3) 1 A third reference representation (RR;?) in the frequency domain is generated by a transformation, e.g., a Fourier transform. The third reference representation (RR3) F ) represents the liver or part of the liver of the reference person in the hepatobiliary phase after application of the hepatobiliary contrast agent in the frequency space.

[0245] Based on the first reference representation (RRI) F ) and the second reference representation (RR2) F ) becomes the first reference difference representation (RRD). 1 ) generated. The generation of the first reference-difference representation (RRD) 1 ) can involve subtracting the first reference representation (RRI). F ) from the second reference representation (RR2) F) include the creation of the first reference-difference representation (RRD). 1 ) can involve subtracting the second reference representation (RRi) F ) from the first reference representation (RR2) F ) include.

[0246] Based on the second reference representation (RR2) F ) and the third reference representation (RR3) F ) becomes the second reference difference representation (RRD). 2 ) generated. The generation of the second reference-difference representation (RRD) 2 ) can involve subtracting the second reference representation (RR2). F ) from the third reference representation (RR3) F ) include.

[0247] It is also possible to use the second reference difference representation (RRD). 2 ) based on the first reference representation (RRi) F ) and the third reference representation (RR3) F ) to generate. Generating the second reference-difference representation (RRD). 2) can involve subtracting the first reference representation (RR2). F ) from the third reference representation (RR3) F ) include.

[0248] Figure 2 illustrates how spatial representations are transformed into frequency-space representations. These spatial representations are typically the result of a radiological examination of the reference subject. Therefore, they are usually measurement-generated representations. It should be noted that spatial representations are not strictly necessary. Frequency-space representations can also be generated as a result of the radiological examination of the reference subject. For example, an MRI scan typically generates so-called k-space data, which represent a frequency-space area of ​​the subject.It is therefore possible, for example, that the frequency space representations are generated directly during the radiological examination of the reference person and used to generate the training data (first reference difference representation, second reference difference representation).

[0249] Fig. 3 shows, by way of example and schematically, the use of the trained machine learning model (MLM') for prediction. The model (MLM) 1 ) of machine learning may have been trained as described in relation to Fig. 2.

[0250] In a first step, a first representation (RI) is created. F ) and a second representation (R2) F ) provided.

[0251] The first representation (R1) F ) represents a liver or part of a liver of a human in the portal venous phase after administration of a hepatobiliary contrast agent in the frequency domain. The second representation (R2) F) represents the liver or part of the liver of a person in a transitional phase after application of the hepatobiliary contrast agent in the frequency domain.

[0252] The first representation (Ri F ) in the frequency domain can be based on a first representation (Ri 1 ) are generated in the local space. The first local space representation (Ri 1 ) represents the liver or part of the liver in humans during the portal venous phase after administration of hepatobiliary contrast agent in local space. The first representation (Ri F ) in the frequency domain can, for example, be transformed from the first representation (Ri) by Fourier transformation 1 ) are generated in the local area.

[0253] The second representation (R2) F ) in the frequency domain can be based on a second representation (R2) 1 ) are generated in the local space. The second local space representation (R2) 1) represents the liver or part of the liver of a person in the transition phase after administration of the hepatobiliary contrast agent in local space. The second representation (R2) F ) in the frequency domain can, for example, be obtained from the second representation (R2) by Fourier transformation 1 ) are generated in spatial space. Based on the first representation (RI). F ) in the frequency domain and the second representation (R2) F ) in the frequency domain a first difference representation (RD) is created 1 ) generated. The first difference representation (RD) 1 ) represents a difference (differences) between the first representation (RI) F ) and the second representation (R2) F ). Generating the first difference representation (RD) 1 ) can, for example, involve subtracting the first representation (RI) F ) from the second representation (R2 F ) include the generation of the first difference representation (RD).1 ) can, for example, involve subtracting the second representation (R2). F ) from the first representation (Ri F ) include. Here, the subtraction chosen is the same one used when training the machine learning model. The first difference representation (RD) 1 ) is applied to the trained model (MLM) 1 ) of machine learning. The machine learning model (MLM') may, for example, have been trained as described in relation to Fig. 2.

[0254] The trained machine learning model (MLM') generates based on the first difference representation (RD). 1 ) (and optionally based on further input data) a predicted second difference representation (RD) 2 *).

[0255] In a further step, a third representation (R3) can be created. 1 ) are generated. The generation of the third representation (R3) 1) can be an addition of the second representation (R2) F ) to the predicted second difference representation (RD) 2 *) include. The generation of the third representation (R3) 1 ) can be an addition of the first representation (RI) F ) to the predicted second difference representation (RD) 2 *) include. Here, the addition method chosen is the same one used when training the machine learning model.

[0256] The third representation (Rs F ) represents the liver or part of the liver of a person in a hepatobiliary phase after administration of the hepatobiliary contrast agent in the frequency domain. Based on the third representation (R3) 1 ) in the frequency domain, a third representation (R3) can be obtained through a transformation, e.g. an inverse Fourier transform. 1 ) are generated in the local space. The third local space representation (R3) 1) represents the liver or part of the liver of a person in the hepatobiliary phase after application of the hepatobiliary contrast agent in local space.

[0257] The third spatial representation (R3) 1 ) can be output, i.e., displayed on a monitor and / or printed on a printer and / or stored in a data storage device and / or transmitted to a separate computer system.

[0258] Figure 3 illustrates how spatial representations are transformed into frequency-space representations. Spatial representations are typically the result of a radiological examination of a human being. Therefore, they are usually measurement-generated representations. It should be noted that spatial representations are not strictly necessary. Frequency-space representations can also be generated as a result of the radiological examination. For example, an MRI scan typically produces so-called k-space data, which are a representation of an area of ​​investigation within the frequency domain.It is therefore possible, for example, that the frequency space representations are generated directly during the radiological examination of the human being and are used to generate the input (first difference representation) for the trained machine learning model.

[0259] Fig. 4 shows, by way of example and schematic representation, a computer system according to the present disclosure, which can be used to train the machine learning model and / or to use the trained machine learning model for prediction.

[0260] A "computer system" is a system for electronic data processing that processes data using programmable instructions. Such a system typically comprises a "computer," the unit containing a processor for performing logical operations, as well as peripherals.

[0261] In computer technology, "peripherals" refers to all devices connected to a computer that are used to control the computer and / or as input and output devices. Examples include monitors (screens), printers, scanners, mice, keyboards, drives, cameras, microphones, speakers, etc. Internal ports and expansion cards are also considered peripherals in computer technology.

[0262] The computer system (10) shown in Fig. 4 comprises a receiving unit (11), a control and computing unit (12) and an output unit (13).

[0263] The control and computing unit (12) serves to control the computer system (10), to coordinate the data flows between the units of the computer system (10) and to perform calculations.

[0264] The control and computing unit (12) is configured:

[0265] to provide an initial representation or to cause the receiving unit to receive the initial representation, wherein the initial representation represents a liver or part of a liver of a human being in a portal venous phase after application of a hepatobiliary contrast agent in the frequency domain,

[0266] to provide a second representation or to cause the receiving unit to receive the second representation, wherein the second representation represents the liver or part of the liver of the human being in a transitional phase after application of the hepatobiliary contrast agent in the frequency domain,

[0267] to generate a first difference representation based on the first representation and the second representation, where the first difference representation represents a difference between the first representation and the second representation,

[0268] to input the first difference representation into a trained machine learning model, and to receive a second difference representation as output from the machine learning model,

[0269] to generate a third representation, wherein generating the third representation involves adding the second difference representation to the second representation or to the first representation,

[0270] to transform the third representation into a third representation in spatial space, to cause the output unit to output the third representation in spatial space, to store it in a data storage device and / or to transmit it to a separate computer system.

[0271] Fig. 5 shows, by way of example and schematic, another embodiment of the computer system of the present disclosure.

[0272] The computer system (1) comprises a processing unit (21) connected to a memory (22). The processing unit (21) and the memory (22) form a control and arithmetic unit, as shown in Fig. 4.

[0273] The processing unit (21) may comprise one or more processors alone or in combination with one or more memories. The processing unit (21) may be ordinary computer hardware capable of processing information such as digital images, computer programs, and / or other digital information. The processing unit (21) typically consists of an array of electronic circuits, some of which may be implemented as an integrated circuit or as several interconnected integrated circuits (an integrated circuit is sometimes referred to as a "chip"). The processing unit (21) may be configured to execute computer programs, which may be stored in a working memory of the processing unit (21) or in the memory (22) of the same or another computer system.

[0274] The memory (22) can be ordinary computer hardware capable of storing information such as digital image recordings (e.g., representations of the study area), data, computer programs, and / or other digital information, either temporarily and / or permanently. The memory (22) can be volatile and / or non-volatile and can be permanently installed or removable. Examples of suitable memory include RAM (Random Access Memory), ROM (Read-Only Memory), a hard disk, flash memory, a removable computer disk, an optical disc, a magnetic tape, or a combination of the above. Optical discs can include read-only compact discs (CD-ROM), read / write compact discs (CD-R / W), DVDs, Blu-ray discs, and similar devices.In addition to the memory (22), the processing unit (21) can also be connected to one or more interfaces (11, 12, 31, 32, 33) to display, transmit, and / or receive information. The interfaces can include one or more communication interfaces (32, 33) and / or one or more user interfaces (11, 12, 31). The one or more communication interfaces can be configured to send and / or receive information, for example, to and / or from an MRI scanner, a CT scanner, an ultrasound camera, other computer systems, networks, data storage devices, or the like. The one or more communication interfaces can be configured to transmit and / or receive information via physical (wired) and / or wireless communication links.The one or more communication interfaces may include one or more interfaces for connecting to a network, e.g., using technologies such as cellular, Wi-Fi, satellite, cable, DSL, fiber optic, and / or the like. In some examples, the one or more communication interfaces may include one or more near-field communication interfaces configured to connect devices using near-field communication technologies such as NFC, RFID, Bluetooth, Bluetooth LE, ZigBee, infrared (e.g., IrDA), or similar technologies.

[0275] The user interfaces may include a display (31). A display (31) may be configured to show information to a user. Suitable examples include a liquid crystal display (LCD), a light-emitting diode (LED) display, a plasma display (PDP), or the like. The user input interface(s) (11, 12) may be wired or wireless and may be configured to receive information from a user into the computer system (1), for example, for processing, storage, and / or display. Suitable examples of user input interfaces include a microphone, an image or video recording device (e.g., a camera), a keyboard or keypad, a joystick, a touch-sensitive surface (separate from or integrated into a touchscreen), or the like.In some examples, the user interfaces may include automatic identification and data capture (AIDC) technology for machine-readable information. This could include barcodes, radio frequency identification (RFID), magnetic stripes, optical character recognition (OCR), integrated circuit cards (ICC), and similar technologies. The user interfaces may also include one or more interfaces for communication with peripheral devices such as printers and the like.

[0276] One or more computer programs (40) can be stored in memory (22) and executed by the processing unit (21), which is programmed to perform the functions described in this description. The retrieval, loading, and execution of instructions from the computer program (40) can be performed sequentially, with one instruction being retrieved, loaded, and executed at a time. However, the retrieval, loading, and / or execution can also be performed in parallel.

[0277] The memory (22) can also contain the machine learning model.

[0278] The computer system of the present disclosure can be implemented as a laptop, notebook, netbook, and / or tablet PC; it can also be a component of an MRI scanner, a CT scanner, or an ultrasound diagnostic device. In one embodiment, the computer system can also be implemented as a cloud-based system in which the trained machine learning model is executed on one or more remote servers (cloud servers). In this case, the first and second representations can be transmitted from a local computer system (e.g., an MRI scanner or a workstation in a clinic) to the cloud server via a network (e.g., the internet). The cloud server can execute the method according to the present disclosure and transmit the third representation back to the local computer system in local space.Such a cloud-based implementation makes it possible to offload computationally intensive operations to powerful servers and reduce the computing requirements on local devices.

[0279] In a further embodiment, the method according to the present disclosure can be performed as real-time processing. "Real-time processing" means that the synthetic third representation is generated during the ongoing radiological examination, i.e., while the patient is still in the tomograph. Real-time processing can be achieved by running the trained machine learning model on optimized hardware (e.g., GPUs, TPUs, or specialized AI accelerators). The advantage of real-time processing is that the radiologist can view and evaluate the synthetic representation of the hepatobiliary phase immediately after the acquisition of the portal venous and transition phases, without having to wait for the actual hepatobiliary phase. This can significantly improve the efficiency of the diagnostic workflow.

[0280] The subject matter of this disclosure also includes a computer program product. Such a computer program product comprises a non-volatile data carrier such as a CD, a DVD, a USB flash drive, or another medium for storing data. A computer program is stored on the data carrier. The computer program can be loaded into the main memory of a computer system (in particular, into the main memory of a computer system of this disclosure) and cause the computer system to perform the following steps:

[0281] Providing an initial representation, wherein the initial representation represents a human liver or part of a liver in a portal venous phase following application of a hepatobiliary contrast agent in the frequency domain,

[0282] Providing a second representation, wherein the second representation represents the liver or part of the liver of a person in a transitional phase after application of the hepatobiliary contrast agent in the frequency space,

[0283] Generating a first difference representation based on the first representation and the second representation, where the first difference representation represents a difference between the first representation and the second representation.

[0284] Inputting the first difference representation into a trained machine learning model, receiving a second difference representation as output from the machine learning model,

[0285] Generating a third representation, wherein generating the third representation involves adding the second difference representation to the second representation or to the first representation,

[0286] Transforming the third representation into a third representation in spatial space, outputting and / or storing the third representation in spatial space and / or transmitting the third representation in spatial space to a separate computer system,

[0287] wherein the trained machine learning model was trained in a supervised learning process based on training data, wherein the training data for each reference person of a plurality of reference persons comprise (i) a first reference difference representation as input data and (ii) a second reference difference representation as target data, wherein the first reference difference representation represents a difference between a first reference representation and a second reference representation, and the second reference difference representation represents a difference between the second reference representation and a third reference representation or between the first reference representation and the third reference representation, wherein the first reference representation represents a liver or part of a liver of the reference person in a portal venous phase after application of the hepatobiliary contrast agent in the frequency domain.where the second reference representation represents the liver or part of the liver of the reference person in a transition phase after administration of the hepatobiliary contrast agent in the frequency domain, and the third reference representation represents the liver or part of the liver of the reference person in a hepatobiliary phase after administration of the hepatobiliary contrast agent in the frequency domain.

[0288] The computer program product can also be marketed in combination (in a kit) with the contrast agent. Such a kit is also referred to as a kit. A kit includes the contrast agent and the computer program product. It is also possible for such a kit to include the contrast agent and means that allow a buyer to obtain the computer program, for example, by downloading it from an internet site. These means may include a link, i.e., an address of the internet site where the computer program can be obtained, for example, from which the computer program can be downloaded to an internet-connected computer system. These means may include a code (e.g., an alphanumeric string, a QR code, a DataMatrix code, a barcode, or another optically and / or electronically readable code) that grants the buyer access to the computer program.Such a link and / or code can, for example, be printed on the packaging of the contrast agent and / or on an accompanying leaflet for the contrast agent. A kit is therefore a combination product comprising a contrast agent and a computer program (e.g., in the form of access to the computer program or in the form of executable program code on a data carrier), which is offered for sale together.

[0289] In one embodiment of the present disclosure, the method can be combined with other artificial intelligence methods. For example, based on the third representation (the synthetic representation of the hepatobiliary phase), automatic segmentation of the liver and / or lesions can be performed. Another machine learning model can be configured and trained to automatically detect, classify, and / or quantify lesions based on the third representation. Such combined AI methods can assist radiologists in diagnosis by, for example, automatically marking suspicious areas, determining the size of lesions, and / or suggesting a classification (e.g., benign vs. malignant).The combination with other AI methods can also include automatic quality control of the synthetic representation, where another machine learning model evaluates the quality and / or plausibility of the generated synthetic representation.

Claims

Patent claims 1. Computer-implemented method comprising the following steps: Providing an initial representation (RI) F ), where the first representation (RI F ) represents a liver or part of a liver of a human being in a portal venous phase after application of a hepatobiliary contrast agent in the frequency domain, Providing a second representation (R2) F ), where the second representation (R2) F ) the liver or part of the liver of a person in a transitional phase after application of the hepatobiliary contrast agent in the frequency domain, Generating a first difference representation (RD) 1 ) based on the first representation (Ri F ) and the second representation (R2) F ), where the first difference representation (RD) 1 ) a difference between the first representation (RI) F ) and the second representation (R2)F ) represents, Entering the first difference representation (RD) 1 ) into a trained machine learning model (MLM'), Receiving a second difference representation (RD) 2 *) as output from the trained model (MLM) 1 ) of machine learning, Generating a third representation (Rs F ), where the generation of the third representation (RS F ) an addition of the second difference representation (RD) 2 *) TO the second representation (R2 F ) or includes the first representation (RR), Transforming the third representation (Rs F ) into a third representation (R3) 1 ) in spatial space, output and / or storage of the third representation (R3) 1 ) in the local space and / or transmission of the third representation (R3) 1 ) in the local area to a separate computer system, where the trained model (MLM) 1) of machine learning was trained in a supervised learning process based on training data, wherein the training data for each reference person of a plurality of reference persons (i) is an initial reference difference representation (RRD). 1 ) as input data and (ii) a second reference difference representation (RRD) 2 ) as target data, where the first reference difference representation (RRD) 1 ) a difference of a first reference representation (RRi) F ) and a second reference representation (RR2) F ) represents and the second reference difference representation (RRD) 2 ) a difference of the second reference representation (RR2) F ) and a third reference representation (RR3) F ) or the first reference representation (RRI) F ) and the third reference representation (RR3) F) represents, wherein the first reference representation (RRÜ) represents a liver or part of a liver of the reference subject in a portal venous phase after administration of the hepatobiliary contrast agent in the frequency space, wherein the second reference representation (RR2) F ) the liver or part of the liver of the reference person in a transition phase after application of the hepatobiliary contrast agent in the frequency space and the third reference representation (RR3) F ) represents the liver or part of the liver of the reference person in a hepatobiliary phase after application of the hepatobiliary contrast agent in the frequency space.

2. The method according to claim 1, wherein the generation of the first difference representation (RD) 1 ) a subtraction of the first representation (RÜ) from the second representation (R2) F ) includes.

3. Method according to one of claims 1 or 2, wherein the generation of the third representation (R3) F ) an addition of the second difference representation (RD) 2 *) TO the second representation (R2 F ) includes, and where the second reference difference representation (RRD) 2 ) based on the second reference representation (RR2) F ) and the third reference representation (RR3) F ) was generated, whereby the generation of the second reference difference representation (RRD) 2 ) a subtraction of the second reference representation (RR2) F ) from the third reference representation (RR, 1 ) included.

4. Method according to any one of claims 1 to 3, wherein the first representation (RI) F ) and the second representation (R2) F ) are the result of a radiological examination of the human being.

5. Method according to any one of claims 1 to 3, further comprising: - Generating the first representation (Ri F ) from an initial representation of the local space (Ri 1 ), where the first spatial representation (Ri 1 ) the liver or part of the liver of humans in the portal venous phase after application of the hepatobiliary contrast agent in local space, - Generating the second representation (R2) F ) from a second spatial representation (1U 1 ). where the second spatial representation (R2) represents the liver or part of the liver of the human in the transition phase after application of the hepatobiliary contrast agent in spatial space, where the first spatial representation (Ri 1 ) and the second spatial representation (R2) are the result of a radiological examination of the human.

6. Method according to any one of claims 1 to 4, wherein the first reference representation (RR1) and the second reference representation (RR2) F) of each reference person are the result of a radiological examination of the reference person.

7. Method according to claim 5, wherein the first reference representation (RRÜ) in the frequency domain is based on a first reference representation (RRi). 1 ) was generated in local space, with the first reference representation (RRi) 1 ) in local space represents the liver or part of the liver of the reference person in the portal venous phase after administration of the hepatobiliary contrast agent in local space, wherein the second reference representation (RR2) F ) in the frequency domain based on a second reference representation (RR2) I ) was generated in the local space, with the second reference representation (RR2) I ) in local space represents the liver or part of the liver of the reference person in the transition phase after application of the hepatobiliary contrast agent in local space, where the third reference representation (RR3) F) in the frequency domain based on a third reference representation (RR3) 1 ) was generated in the local space, with the third reference representation (RR3) 1 ) in local space represents the liver or part of the liver of the reference person in the hepatobiliary phase after application of the hepatobiliary contrast agent in local space, where the first reference representation (RRi) 1 ) in the local space and the second reference representation (RR? ). 1 ) in the local space and the third reference representation (RR3) 1 ) in the local area of ​​each reference person are the result of a radiological examination of the reference person.

8. Method according to any one of claims 4 to 7, wherein the radiological examination is or comprises an MRI examination.

9. A method according to any one of claims 4 to 8, wherein the radiological examination is or comprises a CT scan.

10. A method according to any one of claims 1 to 9, wherein the hepatobiliary contrast agent is or comprises an MRI contrast agent.

11. Method according to any one of claims 1 to 10, wherein the hepatobiliary contrast agent comprises one or more of the following substances: Gadoxetate disodium Gd 3+ -Complex of a compound of formula (I) (I) , where Ar a group selected from where # represents the link to X, X represents a group consisting of CH2, (CH2)2, (CIE),. (CH3)4 and *-(CH2)2-O-CH2-# is selected where * represents the link to Ar and # represents the link to the acetic acid residue, R 1 , R 2 and R 3independently represent a hydrogen atom or a group selected from Ci-Cs-alkyl, -CH2OH, -(CH2)2OH and -CH2OCH3 R 4 represents a group selected from C2-C4 alkoxy, (H3C-CH2)-O-(CH2)2-O-, (H3C-CH2)-O-(CH2)2-O- (CH2)2-O- and (H3C-CH2)-O-(CH2)2-O-(CH2)2-O-(CH2)2-O- R 5 represents a hydrogen atom and R 6 represents a hydrogen atom or a stereoisomer, tautomer, hydrate, solvate or salt thereof, or a mixture thereof, Gd 3+ -Complex of a compound of formula (II) (II) , where Ar a group selected from R 8 where # represents the link to X, X represents a group selected from CH2, (CH2)2, (CH2)3, (CH2)4 and *-(CH2)2-O-CH2-#, where * represents the link to Ar and # represents the link to the acetic acid residue, R 7 a hydrogen atom or group selected from Ci-Cs-alkyl, -CH2OH, represents -(CH2)2OH and -CH2OCH3; R 8 a group selected from C2-C4-alkoxy, (H3C-CH2O)-(CH2)2-O-, (H3C-CH2O)-(CH2)2-O-(CH2)2-O- and (H3C-CH2O)-(CH2)2-O-(CH2)2-O-(CH2)2-O- represents; R 9 and R 10 represent a hydrogen atom independently of each other; or a stereoisomer, tautomer, hydrate, solvate or salt thereof, or a mixture thereof, gadolinium 2,2',2"-(10-{ l-carboxy-2-[2-(4-ethoxyphenyl)ethoxy]ethyl}-l,4,7,10-tetraazacyclododecane- 1 ,4,7-triyl)triacetate, Gadolinium 2,2',2"-{10-[l-carboxy-2-{4-[2-(2-ethoxyethoxy)ethoxy]phenyl}ethyl]-1,4,7, 10-tetraazacyclododecane-1,4,7-triyl}triacetate, Gadolinium 2,2',2"-{10-[(lR)-1-carboxy-2-{4-[2-(2-ethoxyethoxy)ethoxy]phenyl}ethyl]-1,4,7, 10-tetraazacyclododecane-1,4,7-triyl}triacetate, Gadolinium (2S,2'S,2"S)-2,2',2"-{ 10-[(1S)-1-carboxy-4-{4-[2-(2-ethoxyethoxy)ethoxy]phenyl}butyl] - 1 ,4,7, 10-tetraazacyclododecane- 1 ,4,7-triyl}tris(3-hydroxypropanoate), Gadolinium 2,2',2"-{ 10-[(1S)-4-(4-butoxyphenyl)-1-carboxybutyl] -1,4, 7,10-tetraazacyclododecane- 1 ,4,7-triyl}triacetate.

12. Method according to any one of claims 1 to 11, wherein the trained machine learning model (MLM') is or comprises an artificial neural network.

13. Method according to any one of claims 1 to 12, wherein training the machine learning model (MLM) comprises: for each reference person out of the multitude of reference persons: Entering the first reference difference representation (RRD) 1 ) into the machine learning model (MLM), Receiving a predicted second reference difference representation (RR) D 2 *) as output from the machine learning model (MLM), Quantifying a deviation between the second reference difference representation (RRD) 2 ) and the predicted second reference difference representation (RRD) 2 *), Reducing the deviation by modifying model parameters (MP) of the machine learning model (MLM).

14. Computer system (10) comprising • a receiving unit (11), • a control and calculation unit (12) and • one output unit (13), wherein the control and computing unit (12) is configured to execute the method according to any one of claims 1 to 13.

15. Computer program product comprising a computer program that can be loaded into the main memory of a computer system and causes the computer system to execute the method according to any one of claims 1 to 13.

16. Kit comprising a hepatobiliary contrast agent and a computer program product according to claim 15.

Images