Machine learning in contrast radiology
Patent Information
- Application Number
- JP2023553036
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2021-04-07
- Filing Date
- 2021-11-29
- Publication Date
- 2024-12-04
- Estimated Expiration
- 2041-11-29
AI Technical Summary
Existing methods for generating artificial radiographic images using artificial neural networks require co-registration of radiological images, leading to artifacts and high computational complexity, especially when stitching subregions, which can result in misdiagnosis and require significant computational resources.
The method involves generating and processing representations of examination regions in frequency space using machine learning models, allowing for the prediction of radiographic images with contrast enhancement without actual contrast agent administration, reducing computational power and error tolerance, and minimizing artifacts.
This approach reduces the need for high computational power, minimizes artifacts, and allows for accurate generation of artificial radiographic images with enhanced contrast, thereby improving diagnostic accuracy and efficiency.
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to the technical field of generating artificial contrast radiological images using machine learning methods. [Background technology]
[0002] WO2019 / 074938A1 discloses a method for reducing the amount of contrast agent in the production of radiological images using an artificial neural network.
[0003] In a first step, a training data set is generated. For a plurality of individuals, the training data set includes for each individual: i) a native radiographic image (zero contrast image), ii) a radiographic image after administration of a low amount of contrast agent (low contrast image), and iii) a radiographic image after administration of a standard amount of contrast agent (full contrast image). The term "multiplicity" preferably means more than 10, even more preferably more than 100.
[0004] In a second step, an artificial neural network is trained to predict for each person in the training dataset an artificial radiographic image showing the acquisition area after administration of a standard amount of contrast agent based on the native image and the image after administration of a low amount of contrast agent. The measured radiographic image after administration of the standard amount of contrast agent is used in each case as a reference (ground truth) in training.
[0005] In a third step, the trained artificial neural network can be used to predict for a new individual, based on the native image and the radiological image after administration of a low amount of contrast agent, an artificial radiological image showing the acquisition area as it would appear if a standard amount of contrast agent were administered.
[0006] The method disclosed in WO2019 / 074938A1 has drawbacks.
[0007] For example, coregistration of radiographic images is required to match individual radiographic images so that the pixels / voxels correspond to each other, i.e., one pixel / voxel of a radiographic image from one person shows the same examination area as a pixel / voxel of a different radiographic image from said person. If the radiographic images do not match, what appears in the artificially generated radiographic image are artifacts that mask and / or distort and / or simulate small anatomical structures within the acquisition area.
[0008] Moreover, the method disclosed in WO2019 / 074938A1 always uses the complete radiographic image for prediction. In particular, when the training of the artificial neural network and the prediction of the artificial radiographic image involves the use of not only the radiographic images described above under numbers i) and ii), but also further radiographic images, for example radiographic images after administration of various amounts of contrast agent, the computational complexity for the generation of the artificial radiographic image can rapidly become very large. It is conceivable that the computation of the artificial radiographic image takes a lot of time and / or that special and / or expensive hardware is required to be able to perform the computation (within a reasonable time range). It is conceivable that the radiographic image is reduced (has to be reduced) into partial regions (patches), said partial regions being (has to be processed) separately from each other so as not to overload the computer memory with excessively large radiographic images. However, such an approach may result in artifacts at the interface when the partial regions processed separately from each other are recombined to form the complete radiographic image (stitching artifacts). Subsequent removal of such stitching artifacts implies additional complexity and risk of errors in the synthetically generated radiological images, which may be misinterpreted by the radiologist (risk of misdiagnosis). [Prior art documents] [Patent documents]
[0009] [Patent Document 1] International Publication No. 2019 / 074938 Brochure Summary of the Invention [Problem to be solved by the invention]
[0010] The object of the present invention is therefore to go beyond the described prior art and to create a solution for generating artificial radiological images which is resistant to errors in coregistration and / or the prediction requires less computing power and / or the computational complexity can be adapted to given hardware and / or available time and / or the risk of artifacts (in particular stitching artifacts) can be reduced.
[0011] This object is achieved by the subject matter of the independent claims. Preferred embodiments of the invention can be found in the dependent claims, the description and the drawings. [Means for solving the problem]
[0012] In a first aspect, the present invention provides a method for producing a pharmaceutical composition comprising the steps of: - receiving a first representation of an examination region of an examination object in frequency space, where the first representation represents the examination region without a contrast agent or after administration of a first amount of a contrast agent; - receiving a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of a contrast agent; - feeding at least a portion of the first representation and at least a portion of the second representation to a machine learning model; - receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of the contrast agent; - generating a representation of the examination area in the real space representation based on the third representation; - outputting and / or storing a representation of the examination area in a real space representation; The present invention provides a computer-implemented method comprising:
[0013] The present invention further comprises the following: Receiving unit, A control and computing unit, and Output unit Equipped with - wherein the control and computing unit is configured to prompt the receiving unit to receive at least two representations, a first representation and a second representation, of an examination area of an examination object, where the first representation represents the examination area without contrast agent or after administration of a first amount of contrast agent and where the second representation represents the examination area after administration of a second amount of contrast agent; - where the control and computation unit is configured to provide at least a portion of the first representation and at least a portion of the second representation to a machine learning model; - wherein the control and computation unit is configured to receive from the machine learning model a third representation of the examination region, where the third representation represents the examination region in frequency space after administration of a third amount of the contrast agent; - where the control and computation unit is configured to generate a representation of the search area in real space based on the third representation; wherein the control and calculation unit is configured to prompt the output unit to output and / or store a representation of the examination area in real space. A computer system is provided.
[0014] The present invention further provides a computer program capable of being loaded into the memory of a computer system, the computer program comprising: - receiving a first representation of an examination region of an examination object in frequency space, where the first representation represents the examination region without a contrast agent or after administration of a first amount of a contrast agent; - receiving a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of a contrast agent; - feeding at least a portion of the first representation and at least a portion of the second representation to a machine learning model; - receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of the contrast agent; - generating a representation of the examination area in the real space representation based on the third representation; - outputting and / or storing a representation of the examination area in a real space representation; The computer system is prompted to execute the
[0015] The present invention further provides the use of a contrast agent in a radiological method comprising: - generating a first representation of an examination region of the examined object in frequency space, where the first representation represents the examination region without contrast agent or after administration of a first amount of contrast agent; - generating a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of contrast agent, where the second amount is different from the first amount; - feeding at least a portion of the first representation and at least a portion of the second representation to a machine learning model; - receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of the contrast agent, where the third amount is different from the first amount and the second amount; - generating a representation of the examination area in the real space representation based on the third representation; - outputting and / or storing a representation of the examination area in a real space representation; Includes.
[0016] Further provided is a contrast agent for use in a radiological method, the method comprising: - generating a first representation of an examination region of the examined object in frequency space, where the first representation represents the examination region without contrast agent or after administration of a first amount of contrast agent; - generating a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of contrast agent; - feeding at least a portion of the first representation and at least a portion of the second representation to a machine learning model; - receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of contrast agent, where the third amount of contrast agent is different from the first amount and the second amount; - generating a representation of the examination area in the real space representation based on the third representation; - outputting and / or storing a representation of the examination area in a real space representation; Includes.
[0017] Further provided is a kit comprising an imaging agent and a computer program according to the invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0018] The invention will be described in more detail below without distinguishing between the subject matter of the invention (method, computer system, computer program, use, imaging agent for use, kit). On the contrary, the following description is intended to apply equally to all subject matter of the invention, regardless of the context (method, computer system, computer program, use, imaging agent for use, kit) in which it occurs.
[0019] When steps are described in a sequence in this specification or claims, this does not necessarily mean that the invention is limited to the sequence described. Instead, steps may be performed in a different sequence or in parallel with each other, unless a step is built on another step, which does not necessarily require that a step built on a previous step be performed consecutively (as will become clear in each individual case). Thus, the sequence described is a preferred embodiment of the invention.
[0020] According to the present invention, an artificial radiation image of an examination area of an examination object can be generated.
[0021] A "test subject" is usually a living organism, preferably a mammal, very particularly preferably a human being.
[0022] An "examination area" is a part of an object under examination, for example an organ or part of an organ.
[0023] The examination area, also called image volume (field of view, FOV), is the volume that is imaged in particular in a radiological image. The examination area is typically defined, for example by a radiologist, on an overview image (localizer). Of course, the examination area can alternatively or additionally be defined automatically, for example based on a selected protocol.
[0024] The examination area is subjected to a radiological examination.
[0025] "Radiology" is the branch of medicine that deals with the application of electromagnetic and mechanical waves (including, for example, ultrasound diagnosis) for diagnostic, therapeutic and / or scientific purposes. Besides X-rays, other ionizing radiations such as gamma rays or electrons are also used. Since the basic intended purpose is imaging, radiology also includes other imaging methods such as ultrasound and magnetic resonance imaging (nuclear magnetic resonance imaging), even if these methods do not use ionizing radiation. Thus, the term "radiology" in the context of the present invention particularly encompasses the following examination methods: computed tomography, magnetic resonance imaging, ultrasound.
[0026] In a preferred embodiment of the invention, the radiological examination is a magnetic resonance imaging examination.
[0027] Magnetic resonance imaging, or MRI for short, is an imaging technique used especially in medical diagnosis to depict the structure and function of tissues and organs inside the human or animal body.
[0028] In MRI, the magnetic moments of protons in the subject are aligned in a basic magnetic field, resulting in a macroscopic magnetization along the longitudinal direction. This is then deflected from its rest position by irradiation with a radio frequency (HF) pulse (excitation). The return of the excited state to its rest position (relaxation), i.e., the magnetization dynamics, is then detected as a relaxation signal by one or more HF receiving coils.
[0029] For spatial encoding, rapidly switched gradient fields are superimposed on the basic magnetic field. The captured relaxation signals, or detected MRI data, initially exist as raw data in frequency space and can be transformed into real space (image space) by a subsequent inverse Fourier transformation.
[0030] For native MRI, tissue contrast is created by different relaxation times (T1 and T2) and proton densities. T1 relaxation represents the transition of longitudinal magnetization to its equilibrium state, where T1 is the time it takes to reach 63.21% of the equilibrium magnetization before resonant excitation. This is also called the longitudinal relaxation time or spin-lattice relaxation time. T2 relaxation similarly represents the transition of transverse magnetization to its equilibrium state.
[0031] In radiological examinations, contrast agents are commonly used to enhance contrast.
[0032] An "imaging agent" is a substance or mixture of substances that improves the depiction of body structures and functions in imaging procedures such as diagnostic x-rays, magnetic resonance imaging, and ultrasound.
[0033] In computed tomography, iodine-containing solutions are usually used as contrast agents.In magnetic resonance imaging (MRI), superparamagnetic substances (e.g., iron oxide nanoparticles, superparamagnetic iron platinum particles (SIPP)) or paramagnetic substances (e.g., gadolinium chelate, manganese chelate) are usually used as contrast agents.In ultrasound examination, liquids containing gas-filled microbubbles are usually administered intravenously. Examples of contrast agents can be found in the literature (e.g., ASL Jascinth et al.: Contrast Agents in computed tomography: A Review, Journal of Applied Dental and Medical Sciences, 2016, Vol. 2, Issue 2, 143-149; H. Lusic et al.: X-ray-Computed Tomography Contrast Agents, Chem. Rev. 2013, 113, 3, 1641-1666; https: / / www.radiology.wisc.edu / wp-content / uploads / 2017 / 10 / contrast-agents-tutorial.pdf; MRNough et al.: Radiographic and magnetic resonances contrast agents: Essentials and tips for safe practices, World J Radiol. 2017 Sep28; 9(9):339-349; LCAbonyi et al.: Intravascular Contrast Media in Radiography: Historical Development & Review of Risk Factors for Adverse Reactions.
[0034] MRI contrast agents exert their effect by modifying the relaxation time of the structures that incorporate them. Two groups of substances can be distinguished: paramagnetic and superparamagnetic. Both groups of substances have unpaired electrons that induce a magnetic field around individual atoms or molecules. Superparamagnetic contrast agents mainly cause a shortening of T2, whereas paramagnetic contrast agents mainly cause a shortening of T1. The effect of said contrast agents is indirect, since the agent itself does not emit a signal, but instead merely affects the intensity of the signal in its vicinity. An example of a superparamagnetic contrast agent is iron oxide nanoparticles (SPIO, superparamagnetic iron oxide). Examples of paramagnetic contrast agents are gadolinium chelates such as gadopentetate dimeglumine (trade name: Magnevist (registered trademark), etc.), gadoteric acid (Dotarem (registered trademark), Dotagita (registered trademark), Cyclolux (registered trademark)), gadodiamide (Omniscan (registered trademark)), gadoteridol (ProHance (registered trademark)), and gadobutrol (Gadovist (registered trademark)).
[0035] Using the present invention it is possible to generate artificial radiation images showing an examination area of an object in such a way that it appears as if a specific amount of contrast agent had been administered to the object / examination area, without the need for said specific amount of contrast agent to actually be administered.
[0036] For this purpose, a machine learning model (also referred to herein as a predictive model) is used that learns based on training data how different amounts of contrast agent affect the contrast of a radiological image of an examination region. The trained model can then be used to predict a radiological image having a contrast enhancement that would occur after administration of a specific amount of contrast agent, without the need to administer said specific amount of contrast agent.
[0037] Thus, the present invention can be used, for example, to reduce the amount of contrast agent administered, without having to sacrifice the benefits of a high dose of contrast agent (i.e., high contrast enhancement), as described in WO2019 / 074938A1.
[0038] However, as described in WO2019 / 074938A1, the radiation image is not predicted based on the image. The image used in WO2019 / 074938A1 for the prediction of the artificial radiation image is a representation (depiction) of the examination area in real space (also called image space).
[0039] According to the present invention, a machine learning model is trained using a representation of the inspection regions of multiple inspection objects in frequency space, and predictions are also made based on the representation of the inspection regions in frequency space (also called spatial-frequency space or Fourier space or frequency domain or Fourier representation).
[0040] In magnetic resonance imaging, raw data usually arise from the above-mentioned measurement methods as so-called k-space data, which is a representation of the examination area in frequency space. Such k-space data can be used for training machine learning models, for model validation and for prediction, using the trained model according to the invention.
[0041] However, if the representation is in real space, it can be converted (transformed) into a representation in frequency space, for example by a Fourier transform.
[0042] Thus, if a radiation image of the examination area exists in the form of a two- or three-dimensional image in real space, this representation of the examination area can be converted into a two- or three-dimensional representation of the examination area in frequency space, for example by means of a 2D or 3D Fourier transform.
[0043] Conversely, a representation in frequency space can be converted (transformed) into a representation in real space by an inverse Fourier transform.
[0044] FIG. 1 (Fig.1) shows a schematic and exemplary relationship between a representation of an examination area in real space and a representation of an examination area in frequency space. A timeline is shown in FIG. 1. At three different time points t1, t2 and t3, a representation of an examination area is generated based on the examination. The examination area is a human lung. At time point t1, a first representation is generated. This can be a representation O1 of the examination area (lung) in real space or a representation F1 of the examination area (lung) in frequency space. The representation O1 of the examination area in real space can be transformed by a Fourier transform (FT) into a representation F1 of the examination area in frequency space. The representation F1 of the examination area in frequency space can be transformed by an inverse Fourier transform (iFT) into a representation O1 of the examination area in real space.
[0045] Representations O1 and F1 may for example be representations of the examination region without contrast agent or after administration of a first amount of contrast agent.
[0046] The two representations O1 and F1 contain the same information about the examination area, simply depicted differently.
[0047] At time t2, a further representation is generated. This can be a representation O2 of the examination area (lungs) in real space or a representation F2 of the examination area (lungs) in frequency space. The representation O2 of the examination area in real space can be transformed by a Fourier transform (FT) into a representation F2 of the examination area in frequency space. The representation F2 of the examination area in frequency space can be transformed by an inverse Fourier transform (iFT) into a representation O2 of the examination area in real space.
[0048] Representations O2 and F2 may for example be representations of the examination region after administration of a second amount of contrast agent.
[0049] The two representations O2 and F2 contain the same information about the examination area, simply depicted differently.
[0050] At time t3, a further representation is generated. This can be a representation O3 of the examination area (lungs) in real space or a representation F3 of the examination area (lungs) in frequency space. The representation O3 of the examination area in real space can be transformed by a Fourier transform (FT) into a representation F3 of the examination area in frequency space. The representation F3 of the examination area in frequency space can be transformed by an inverse Fourier transform (iFT) into a representation O3 of the examination area in real space.
[0051] Representations O3 and F3 may for example be representations of the examination region after administration of a third amount of contrast agent.
[0052] The two representations O3 and F3 contain the same information about the examination area, simply depicted differently.
[0053] The representations O1, O2 and O3 of the examination area in real space are familiar representations for humans, and such representations of real space can be immediately grasped by humans. The representations O1, O2 and O3 show how different amounts of contrast agent can affect the appearance of the examination area in an MRI examination. In this case, the amount of contrast agent increases from O1 through O2 to O3. The same information is also contained in the representations F1, F2 and F3, but said information is more difficult for humans to grasp from the representations F1, F2 and F3.
[0054] It is possible to consider using transforms other than the Fourier transform to convert the real space representation into a frequency space representation. There are three main properties that must be satisfied by such a transform: a) The existence of a clear inverse transform (a clear link between the real space representation and the frequency space representation) b) Locality of contrast information c) Robustness with respect to defective image registration It is.
[0055] The details of the transformation from one representation to another are described in numerous textbooks and publications (e.g. W. Burger, MJ Burger: Digital Image Processing: An Algorithmic Introduction Using Java, Texts in Computer Science, 2nd edition, Springer-Verlag, 2016, ISBN: 9781447166849; W. Birkfellner: Applied Medical Image Processing, Second Edition: A Basic Course, Verlag Taylor & Francis, 2014, ISBN: 9781466555570; R. Bracewell: Fourier Analysis and Imaging, Verlag Springer Science & Business Media, 2004, ISBN: 9780306481871).
[0056] In order for the machine learning model (predictive model) according to the present invention to be able to make the predictions described herein, it must be properly configured (trained) in advance.
[0057] Here, the term "prediction" means that at least one representation of the examination area representative of the examination area after administration of a specific amount of contrast agent in frequency space is calculated based on at least two representations of the examination area in frequency space, where the at least two representations represent the examination area after administration of different amounts of contrast agent and / or after administration of different contrast agents.
[0058] In other words, the at least one first representation of the examination area of the examination object and the at least one second representation of the examination area of the examination object are used to predict the at least one third representation of the examination area of the examination object, where all the representations, the at least one first representation, the at least one second representation and the at least one third representation are representations of the examination area in frequency space.
[0059] At least one first representation represents the examination region after administration of a first amount of a first contrast agent, said first amount also being possible to be zero (no administration of contrast agent).
[0060] At least one second representation represents the examination region after administration of a second amount of the first contrast agent or a second amount of the second contrast agent. If a first contrast agent is used to generate the second representation, the second amount is usually not equal to the first amount. If a second contrast agent is used, the second amount may or may not be equal to the first amount.
[0061] At least one third representation represents the examination region after administration of the third amount of the first contrast agent or the third amount of the second contrast agent or the third amount of the third contrast agent. If the same contrast agent is used to generate the first, second and third representations, the third amount is usually not equal to the second amount and not equal to the first amount; the following preferably applies to the first amount M1, the second amount M2 and the third amount M3: the first amount M1 is equal to or greater than zero, the second amount M2 is greater than the first amount M1 and the third amount M3 is greater than the second amount M2 (0≦M1 <M2<M3)。
[0062] If the production of the third representation involves the use of a different contrast agent compared to the production of the first and / or second representation, the amount of the different contrast agent may or may not be equal to the amount of contrast agent for the production of the first and / or second representation.
[0063] Thus, the predictive model can be trained to learn the effect of different amounts of contrast agent on the representation of the examination region in frequency space, but it can also be trained to learn the effect of different contrast agents on the representation of the examination region in frequency space.
[0064] The predictive model is preferably constructed (constructed, trained) using a self-learning algorithm in a supervised machine learning process. Training data is used for learning. The training data includes a plurality of representations of examination areas for each examination object among a plurality of examination objects. The examination areas are typically the same for all examination objects (e.g., body parts, organs or organ parts). The representations of the training dataset are also referred to herein as reference representations.
[0065] For each examination object, the training data comprises i) at least one first reference representation of the examination area in frequency space, representing the examination area without contrast agent or after administration of a first amount of contrast agent, ii) at least one second reference representation of the examination area in frequency space, representing the examination area after administration of a second amount of contrast agent, and iii) at least one third reference representation of the examination area in frequency space, representing the examination area after administration of a third amount of contrast agent. As mentioned above, the following applies: the first, second and third amounts of contrast agent and / or the particular contrast agents used are different from each other.
[0066] The predictive model is trained to predict (calculate) for each test subject at least one third referring expression based on the at least one first referring expression and the at least one second referring expression.
[0067] During machine learning, a self-learning algorithm generates a statistical model based on training data. This means that the examples are not simply memorized, but the algorithm "recognizes" patterns and regularities in the training data. Thus, the predictive model can also evaluate unknown data. Validation data can be used to test the quality of the evaluation of unknown data.
[0068] The predictive model can be trained by supervised learning, i.e., three data sets (first, second, and third representations) are presented sequentially to the algorithm, which then learns the relationships between these data sets.
[0069] Self-learning systems trained by supervised learning have been widely described in the prior art (see, for example, C. Perez: Machine Learning Techniques: Supervised Learning and Classification, Amazon Digital Services LLC-Kdp Print Us, 2019, ISBN1096996545, 9781096996545).
[0070] Preferably, the predictive model is or comprises an artificial neural network.
[0071] An artificial neural network comprises at least three layers of processing elements: a first layer having input neurons (nodes), an Nth layer having at least one output neuron (node), and N-2 inner layers, where N is a natural number greater than 2.
[0072] The input neuron is operative to receive the first and second representations, and the output neuron is operative to output at least one third representation for the at least one first representation and the at least one second representation.
[0073] The processing elements of the layer between the input neurons and the output neurons are connected to each other in a predetermined pattern with predetermined connection weights.
[0074] Preferably, the artificial neural network is a so-called Convolutional Neural Network (abbreviated as CNN).
[0075] Convolutional neural networks are able to process input data in the form of matrices. CNNs essentially consist of alternating filters (convolutional layers) and aggregation layers (pooling layers), and finally, one or more layers of "normal" fully connected neurons (dense / fully connected layers).
[0076] The training of the neural network can be carried out, for example, by backpropagation. The objective here for the network is maximum reliability in mapping given input data to given output data. The mapping quality is described by a loss function. The objective is to minimize the loss function. In the case of backpropagation, the artificial neural network is taught by modifying the connection weights.
[0077] In the trained state, the connection weights between the processing elements include information about a relationship between, first, the at least one first representation and the at least one second representation, and, second, the at least one third representation, said information being usable to predict the at least one third representation based on the at least one first and the at least one second representation.
[0078] Cross-validation methods can be used to split the data into training and validation data sets. The training data set is used for backpropagation training of the network weights. The validation data set is used to check the accuracy of predictions that the trained network can make when applied to unknown data.
[0079] Further details regarding the construction and training of artificial neural networks can be gleaned from the prior art (see, for example, S. Khan et al.: A Guide to Convolutional Neural Networks for Computer Vision, Morgan & Claypool Publishers 2018, ISBN 1681730227, 9781681730226, WO2018 / 183044A1, WO2018 / 200493, WO2019 / 074938A1, WO2019 / 204406A1, WO2019 / 241659A1).
[0080] Preferably, the predictive model is a generative adversarial network (GAN) (see, for example, http: / / 3dgan.csail.mit.edu / ).
[0081] In addition to the representation, further information about the inspection object, inspection area, inspection conditions and / or radiological inspection method may also be used for training, validation and prediction.
[0082] Examples of information about the test subject are sex, age, weight, height, medical history, nature and duration and amount of medications already taken, blood pressure, central venous pressure, respiratory rate, serum albumin, total bilirubin, blood glucose, iron content, respiratory capacity, etc. Said information about the test subject can also be read, for example, from a database or an electronic patient file.
[0083] Examples of information regarding the examination area are existing conditions, surgery, partial resection, liver transplant, iron liver, fatty liver, etc.
[0084] Figure 2 (Fig. 2) shows, in a schematic and exemplary manner, how the representations (F1), (F2) and (F3) of the test region in frequency space as generated in Figure 1 can be used to train a predictive model (PM). The representations (F1), (F2) and (F3) form a set of training data of the test subject. Training is performed using multiple training data sets of multiple test subjects.
[0085] Representations (F1) and (F2) are first and second representations of the examination region in frequency space, said representations representing the examination region after administration of different amounts of contrast agent (in one example it is also possible that no contrast agent is administered).
[0086] Representation (F3) is a third representation of the examination region in frequency space, representing the examination region after administration of a third amount of contrast agent.
[0087] In FIG. 2, a prediction model is trained to predict a representation of an examination region in frequency space (F3) from representations of the examination region in frequency space (F1) and (F2).
[0088] The representations (F1) and (F2) are input to a predictive model (PM), which calculates a representation (F3*) from the representations (F1) and (F2). The asterisk (*) indicates that the representation (F3*) is a predicted representation. The calculated representation (F3*) is compared with the representation (F3). The deviation between the calculated representation (F3*) and the measured representation (F3) can be used in a backpropagation method to train a predictive model to reduce the deviation to a defined minimum value. If the predictive model is trained based on multiple training datasets of multiple test subjects, and the prediction reaches a defined accuracy, the trained predictive model can be used for prediction.
[0089] FIG. 3 exemplarily and diagrammatically illustrates a preferred embodiment of the method according to the invention for training a machine learning model in the form of a flow chart.
[0090] The method (100) comprises the steps of: (110) receiving a training dataset, where the training dataset includes, for each test object of a plurality of test objects, i) a first reference representation of an examination area of the test object in frequency space, ii) a second reference representation of the examination area of the test object in frequency space, and iii) a third reference representation of the examination area of the test object in frequency space, where the first reference representation represents the examination area after administration of a first amount of a contrast agent, where the first amount may also be zero, where the second reference representation represents the examination area after administration of a second amount of a contrast agent, and where the third reference representation represents the examination area after administration of the third amount of a contrast agent; (120) for each examination object: feeding at least a portion of the first reference representation and at least a portion of the second reference representation to a machine learning model, where the machine learning model is trained to generate a representation of the examination area in frequency space after administration of a third amount of a contrast agent based on the first reference representation and the second reference representation, where the training includes minimizing a loss function, where the loss function quantifies deviation of the generated representation of the examination area from the third reference representation; (130) outputting and / or storing the trained model and / or feeding the trained model to a method for predicting a representation of the test region of a new test object; Includes.
[0091] The use of the trained prediction model for prediction is exemplarily and diagrammatically shown in Fig. 4 (Fig. 4), which shows the prediction model (PM) trained in Fig. 2. The prediction model is used to predict at least one third representation of the examination area of the examination object in frequency space based on at least one first representation of the examination area of the examination object in frequency space and at least one second representation of the examination area of the examination object in frequency space, said representations representing the examination area after administration of different amounts of contrast agent.
[0092] In this example, a first representation of the examination area in frequency space (F ~ 1) and the second representation (F ~ 2) is input to a predictive model, which then produces a third representation (F ~ 3). The tilde (~) indicates that the expression is a new expression to be tested, and typically, no expression was used in the training method to train the predictive model. The asterisk (*) indicates that the expression (F ~ 3*) is the predicted representation. The representation of the test region in frequency space (F ~ 3*) is obtained by, for example, the inverse Fourier transform iFT, which is a representation of the inspection area in real space (O ~ 3*)).
[0093] FIG. 5 exemplarily and diagrammatically illustrates in the form of a flow chart a preferred embodiment of the method according to the invention for predicting a representation of a test area of a test object.
[0094] The method (200) includes the steps of: (210) receiving a first representation of an examination region of an examination object in frequency space, where the first representation represents the examination region without a contrast agent or after administration of a first amount of a contrast agent; (220) receiving a second representation of an examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of a contrast agent; (230) providing the first representation and the second representation to a machine learning model; (240) receiving a third representation of the examination region in frequency space from the machine learning model, where the third representation represents the examination region after administration of a third amount of the contrast agent; (250) generating a representation of the examination area in a real space representation based on the third representation; (260) Outputting and / or storing a representation of the examination area in a real space representation. Includes.
[0095] As already explained, the representation of the test domain used for training, validation and prediction is a representation of the test domain in frequency space (also called spatial-frequency space or Fourier space or frequency domain or Fourier representation).
[0096] In magnetic resonance imaging, raw data usually arises as so-called k-space data by the above-mentioned measurement method. Said k-space data is a representation of the examination area in frequency space, i.e. such k-space data can be used for training, validation and prediction. If a representation in real space exists, such a representation in real space can be converted (transformed) into a representation in frequency space, for example by Fourier transform, and conversely, the representation in frequency space can be converted (transformed) into a representation in real space, for example by inverse Fourier transform.
[0097] Thus, if a radiological image of the examination area exists in the form of a two-dimensional image in real space, this representation of the examination area can be transformed by a 2D Fourier transform into a two-dimensional representation of the examination area in frequency space.
[0098] A 3D image (volume rendering) of the examination area can be treated as a stack of 2D images. It is also possible to convert the 3D image into a 3D representation of the examination area in frequency space by a 3D Fourier transform.
[0099] The use of a representation of the examination area in frequency space according to the invention has various advantages over the use of a representation of the examination area in real space.
[0100] For example, coregistration of individual representations is less important in frequency space than in real space. "Co-registration" (also called "image registration" in the prior art) is an important process in digital image processing, which serves to match two or more images of the same, or at least similar, scene with each other in the best possible way. One of the images is defined as the reference image, the others are called object images. To optimally match these with the reference image, a compensatory transformation is calculated. The images to be aligned differ from each other because they were acquired from different positions, at different times, and / or with different sensors.
[0101] In the present case, the individual representations are generated at different times and, secondly, they differ with regard to the content of contrast agent in the examination region and / or the contrast agent used.
[0102] Thus, the use of a representation of the examination area in frequency space has the advantage that the training, validation and prediction methods are more resistant to errors in image registration compared to the use of a representation of the examination area in real space. In other words, if the representation in frequency space is not exactly superimposed, this has less impact than if the representation in real space is not superimposed with pixel / voxel accuracy. This is due to the properties of the Fourier transform, which, as already explained, always maps the contrast information of a Fourier transformed image to the vicinity of the origin of the Fourier space. Turns or rotations in image space (real space) result in image information (e.g. visible structures) being localized in different regions of the transformed image. However, in Fourier space, these transformations do not change the regions in which the contrast information relevant to the present invention is encoded.
[0103] A further advantage of using a representation in frequency space is that contrast information, which is important for training and prediction, can be separated from detailed information (fine structure). Thus, in the training case, one can focus on the information that is learned by the predictive model, and in the prediction case, one can also focus on the contrast information, which is the information that is predicted by the predictive model.
[0104] While the contrast information in a representation of the examination area in real space is typically distributed throughout the representation (each pixel / voxel essentially has information about the contrast), the contrast information in a representation of the examination area in frequency space is encoded in and around the center of the frequency space. In other words, the low frequencies in the representation in frequency space are responsible for the contrast, while the high frequencies contain information about the fine structure.
[0105] It is therefore possible to isolate the contrast information, restrict training and prediction to the contrast information, and reintroduce the information about the fine structure after training / prediction.
[0106] To this end, regions may be specified in the representation of the test region that is used for training, validation and prediction.
[0107] Identification of the region may for example be achieved by a user of the computer system according to the invention by inputting one or more parameters into the computer system according to the invention and / or by selecting from a list defining the shape and / or size of the region, however it is also conceivable that identification may be performed automatically, for example by a computer system according to the invention suitably configured to select a predefined region within a representation of the examination region.
[0108] The identified region is typically smaller than the frequency space filled by each representation, but in each case includes the center of the frequency space.
[0109] The region of the representation that includes the center of the frequency space (also called the origin or zero point) contains the contrast information relevant to the method according to the invention. If the identified region is smaller than the frequency space filled by the respective representation, the result is a lower computational complexity for training, validation and subsequent prediction. Thus, the choice of the size of the region can have a direct impact on the computational complexity.
[0110] In principle, it would also be possible to identify regions corresponding to the entire frequency space filled by each representation, in which case there would be no reduction of the frequency space to a smaller region and the computational complexity would be at its maximum.
[0111] Thus, by specifying a region around the center of the frequency space, a user of the computer system of the present invention can decide whether he or she wants a complete representation of the test region in frequency space to form the basis for training, validation, and prediction (which would imply maximum computational complexity), or whether he or she wants to reduce the computational complexity by specifying a smaller region of the frequency space that would be filled by the representation, where the user can directly influence the computational complexity required through the size of the region specified.
[0112] The identified region typically has the same dimensions as the frequency space; in the case of a 2D representation in a 2D frequency space, the identified region typically is an area, and in the case of a 3D representation in a 3D frequency space, the identified region typically is a volume.
[0113] The identified region can in principle have any shape, and thus can be, for example, circular and / or angular, concave and / or convex. Preferably, the region is rectangular or cubic in the case of a 3D frequency space in a Cartesian coordinate system, and rectangular or square in the case of a 2D frequency space in a Cartesian coordinate system. However, it is also conceivable that the identified region is spherical or circular or has other shapes.
[0114] Preferably, the geometric centre of the particular region coincides with the centre of the frequency space.
[0115] The representations used for training, validation and prediction are reduced to the identified regions. The term "reduced" here means that all parts of the representation that are not in the identified range are cut off (discarded) or covered by masking. In the case of masking, the regions outside the identified regions are covered with a mask, so that only the identified regions remain uncovered, and in the case of masking, the color values of the corresponding pixels / voxels can be set to zero (black), for example.
[0116] The representation thus obtained is also referred to herein as the reduced representation.
[0117] The reduced representation can be used for training, validation and prediction.
[0118] Thus, if what is predicted for the object of inspection is a third representation of the inspection area based on the first and second representations, the reduced first representation and the reduced second representation can be fed to a prediction model, which then generates the reduced third representation. In a further step, the detailed information cut or covered during the reduction can be reintroduced into the prediction representation. Thus, for example, it is possible to introduce into the reduced third representation the part of the first representation that is outside the identified area. It is also conceivable that the part of the second representation that is outside the identified area is introduced into the reduced third representation. It is also conceivable that both the part of the first representation that is outside the identified area and the part of the second representation that is outside the identified area are introduced into the reduced third representation. In other words, the third representation is supplemented by the part of the first and / or second representation that was cut / covered with a mask during the reduction. The result is the supplemented third representation. From the supplemented third representation, it is possible to obtain a real space representation by transformation (for example, an inverse Fourier transform).
[0119] The signal intensities obtained for the various position coordinates can be converted in a further step into greyscale or colour values in order to make the digital images available in a common image format (eg DICOM).
[0120] The representation of the examination area in the real space representation may be displayed on a screen, outputted on a printer, and / or stored on a data storage medium.
[0121] The described approach is exemplarily explained in more detail below on the basis of Figures 6 (Fig. 6) and 7 (Fig. 7).
[0122] FIG. 6 exemplarily illustrates diagrammatically the steps in training a predictive model according to a preferred embodiment of the present invention.
[0123] As in FIG. 2, a first representation (F1), a second representation (F2) and a third representation (F3) of an examination area of an examination object in frequency space are received. In the representations (F1), (F2) and (F3), respectively, an area A of the same size and shape is identified. The area A in each case contains the center of the frequency space and in this case has the shape of a square whose geometric center coincides with the center of the frequency space. The representations (F1), (F2) and (F3) are each reduced to the identified area A, and three reduced representations (F1 red ), (F2 red ), and (F3 red ) The reduced representation is used for training. The predictive model is then trained on the reduced representation (F1 red ) and (F2 red ) to the reduced representation (F3 red ) is trained to predict the reduced representation (F1 red ) and (F2 red ) is fed to the predictive model (PM), which then uses the reduced representation (F3 red ) as close as possible to the reduced representation (F3* red ) to calculate
[0124] FIG. 7 exemplarily and diagrammatically illustrates how the predictive model trained in FIG. 6 can be used for prediction.
[0125] In this example, a first representation of the examination area in frequency space (F ~ 1) and the second representation (F ~ 2) are received and each is reduced to a specific region A. The result is two reduced representations (F ~ 1 red ) and (F ~ 2 red ). The reduced first representation (F ~ 1 red ) and the reduced second representation (F~ 2 red ) is fed to a trained predictive model (PM), which then receives the reduced representation (F ~ 1 red ) and (F ~ 2 red ) to the third representation (F ~ 3 red* ) in a further step. ~ 3 red* ) is the first representation received (F ~ 1) the received first representation (F ~ 1) domain. In other words, the reduced third representation (F ~ 3 red* ) are added to the first representation (F ~ 1). As described, the received first representation (F ~ Alternatively or in addition to part 1), a second representation (F ~ 2) into a third representation (F ~ 3 red* ) can also be added.
[0126] The supplemented expression (F ~ 3 red* )+(F ~ 1 DI ) to obtain the expression of the inspection area in real space (O ~ 3 red* ) can be generated by inverse Fourier transform.
[0127] It should be noted that other methods can also be used to convert the frequency space representation into a real space representation, such as, for example, iterative reconstruction methods.
[0128] FIG. 8 illustrates, in the form of a flow chart, a preferred embodiment of the method according to the present invention for training a machine learning model.
[0129] The method (300) is as follows: (310) receiving a training dataset, where the training dataset includes, for each test object of a plurality of test objects, i) a first reference representation of an examination area of the test object in frequency space, ii) a second reference representation of the examination area of the test object in frequency space, and iii) a third reference representation of the examination area of the test object in frequency space, where the first reference representation represents the examination area after administration of a first amount of a contrast agent, where the first amount may also be zero, where the second reference representation represents the examination area after administration of a second amount of a contrast agent, and where the third reference representation represents the examination area after administration of a third amount of a contrast agent; (320) identifying a region within the referring expression, where the identified region includes a center of frequency space; (330) reducing the representation to the identified region, where a reduced first referring representation, a reduced second referring representation, and a reduced third referring representation are obtained for each test object; (340) For each examination object: providing a machine learning model with the reduced first reference representation and the reduced second reference representation, where the machine learning model is trained to generate a reduced representation of the examination area in frequency space after administration of a third amount of contrast agent based on the reduced first reference representation and the reduced second reference representation, where the training includes minimizing a loss function, where the loss function quantifies deviation of the generated reduced representation of the examination area from the reduced third reference representation; (350) outputting and / or storing the trained model and / or feeding the trained model to a method for predicting a representation of the test region of a new test object. Includes.
[0130] FIG. 9 illustrates, in the form of a flow chart, a schematic representation of a preferred embodiment of a method according to the invention for predicting a representation of a care area of a care object.
[0131] The method (400) is as follows: (410) receiving a first representation of an examination region of an examination object in frequency space, where the first representation represents the examination region without a contrast agent or after administration of a first amount of a contrast agent; (420) receiving a second representation of an examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of a contrast agent; (430) identifying a region in the first representation and the second representation, where the identified region includes a center of frequency space; (440) reducing the first representation and the second representation to the identified region, where a reduced first representation and a reduced second representation are obtained; (450) providing the reduced first representation and the reduced second representation to a machine learning model; (460) receiving a third representation of the examination region from the machine learning model in frequency space, where the third representation represents the examination region after administration of a third amount of the contrast agent; (470) supplementing a third representation with parts of the received first and / or the received second representation that are not present in the identified region; (480) converting the captured third representation, obtained after the third representation is captured, into a representation of the examination area in the real space representation; (490) Outputting and / or storing a representation of the examination area in a real space representation. Includes.
[0132] FIG. 10 shows, in the form of a flow chart, a further preferred embodiment of a method according to the invention for predicting a representation of a care area of a care object.
[0133] The method (500) includes the steps of: (510) providing a trained machine learning model, where the model has been trained according to the above-described method (100) or (300) for learning the effect of the amount of contrast agent on a representation of the examination region in frequency space; (520) receiving a first representation of an examination region of the examination object in frequency space, where the first representation represents the examination region after administration of a first amount of a contrast agent, where the first amount may be zero; and receiving a second representation of the examination region of the examination object in frequency space, where the second representation represents the examination region after administration of a second amount of the contrast agent; (530) Optionally, identifying a region in the first representation and the second representation, where the identified region includes a center of frequency space; (540) Optionally, reducing the first representation and the second representation to the identified region, where a reduced first representation and a reduced second representation are obtained; (550) a received first representation and a received second representation or, if present, the reduced first representation and the reduced second representation feeding the data to a machine learning model; (560) receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of the contrast agent; (570) Optionally, supplementing a third representation with parts of the received first and / or received second representation that are not present in the identified region, where a supplemented third representation is obtained; (580) Third Representation Or, if present, a supplemented third representation generating a representation of the examination area in a real space representation from (590) outputting a representation of the inspection area in a real space representation; Includes.
[0134] Further embodiments of the present invention are as follows: 1. A computer-implemented method comprising: - receiving a first representation of an examination region of an examination object in frequency space, where the first representation represents the examination region without a contrast agent or after administration of a first amount of a contrast agent; - receiving a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of a contrast agent; - identifying a region of frequency space in the first representation and / or the second representation, where the identified region comprises a centre of the frequency space; - reducing the first representation and / or the second representation to the identified area; - feeding the first and second representations obtained after the reduction to a machine learning model, the machine learning model being trained on a training data set for learning the effect of the amount of contrast agent on the representation of the examination region in frequency space, - receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of the contrast agent; - if the frequency space occupied by the third representation is smaller than the frequency space occupied by the received first representation and / or the received second representation: filling the frequency space of the third representation with parts of the received first representation and / or the received second representation that are larger than the third representation; - converting the third representation into a representation of the examination area in a real space representation; - outputting a representation of the examination area in a real space representation; The method comprising:
[0135] 2. A computer-implemented method comprising: - receiving a first representation of an examination region of an examination object in frequency space, where the first representation represents the examination region without a contrast agent or after administration of a first amount of a contrast agent; - receiving a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of a contrast agent; - identifying an area in the first representation and / or the second representation, where the identified area comprises a center of frequency space; - reducing the first representation and / or the second representation to the identified area; - feeding the first and second representations obtained after reduction to a machine learning model, the machine learning model being trained on a training data set for learning the effect of the amount of contrast agent on the representation of the examination region in frequency space, - receiving a third representation of the examination region in frequency space from the machine learning model, where the third representation represents the examination region after administration of a third amount of the contrast agent; - supplementing the third representation with those parts of the first received and / or the second received representation that are not present in the identified region; - converting the captured third representation into a representation of the examination area in a real space representation; - outputting a representation of the examination area in a real space representation; The method comprising:
[0136] - receiving a first representation of an examination region of an examination object in frequency space, where the first representation represents the examination region without a contrast agent or after administration of a first amount of a contrast agent; - receiving a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of a contrast agent; - identifying an area in the first representation and / or the second representation, where the identified area comprises a center of frequency space; - reducing the first representation and / or the second representation to the identified area; - feeding the first and second representations obtained after reduction to a machine learning model, the machine learning model being trained on a training data set for learning the effect of the amount of contrast agent on the representation of the examination region in frequency space, - receiving a third representation of the examination region in frequency space from the machine learning model, where the third representation represents the examination region after administration of a third amount of the contrast agent; - supplementing the third representation with those parts of the first received and / or the second received representation that are not present in the identified region; - converting the captured third representation into a representation of the examination area in a real space representation; - outputting a representation of the examination area in a real space representation.
[0137] 3. A computer-implemented method comprising: - receiving a first representation of an examination region of an examination object in frequency space, where the first representation represents the examination region without a contrast agent or after administration of a first amount of a contrast agent; - receiving a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of a contrast agent; - identifying an area in the first representation and / or the second representation, where the identified area comprises a center of frequency space; - reducing the first representation and / or the second representation to the identified area; - feeding the first and second representations obtained after reduction to a machine learning model, the machine learning model being trained on a training data set for learning the effect of the amount of contrast agent on the representation of the examination region in frequency space, - receiving a third representation of the examination region in frequency space from the machine learning model, where the third representation represents the examination region after administration of a third amount of the contrast agent; - supplementing the third representation with those parts of the first received and / or the second received representation that are not present in the identified region; - converting the captured third representation into a representation of the examination area in a real space representation; - outputting a representation of the examination area in a real space representation; The method comprising:
[0138] 4. The computer-implemented method of any one of embodiments 1-3, wherein the first amount of contrast agent is greater than or equal to zero, the second amount of contrast agent is greater than the first amount of contrast agent, and the third amount of contrast agent is greater than the second amount of contrast agent.
[0139] 5. A computer-implemented method according to any one of embodiments 1 to 4, wherein the first and second representations are results of a radiological examination.
[0140] 6. The computer-implemented method of any one of embodiments 1 to 5, wherein the radiological examination is a magnetic resonance imaging examination, a computed tomography examination, or an ultrasound examination.
[0141] 7. A computer-implemented method according to any one of embodiments 1 to 6, wherein the first representation and the second representation are k-space data from a magnetic resonance imaging study.
[0142] 8. A computer-implemented method according to any of embodiments 1-7, wherein the first representation and the second representation are Fourier transformed real space representations.
[0143] 9. The computer-implemented method of any one of embodiments 1 to 8, further comprising training a machine learning model, wherein the training comprises: - receiving a training data set, where the training data set comprises, for a plurality of examination objects, in each case i) a first reference representation of an examination area of the examination object in frequency space, ii) a second reference representation of the examination area of the examination object in frequency space and iii) a third reference representation of the examination area of the examination object in frequency space, where the first reference representation represents the examination area after administration of a first amount of contrast agent, where the first amount may be zero, where the second reference representation represents the examination area after administration of a second amount of contrast agent and where the third reference representation represents the examination area after administration of the third amount of contrast agent, - identifying a region within the referring expression, the identified region including a center of frequency space; - a sub-step of reducing the representations to the identified regions, where a reduced first referring representation, a reduced second referring representation and a reduced third referring representation are obtained for each test object, - for each examination object, providing the reduced first reference representation and the reduced second reference representation to a machine learning model, where the machine learning model is trained to generate a reduced representation of the examination area in frequency space after administration of a third amount of contrast agent based on the reduced first reference representation and the reduced second reference representation, where the training comprises minimizing a loss function, where the loss function quantifies the deviation of the generated reduced representation of the examination area from the reduced third reference representation. The method comprising:
[0144] 10. A computer-implemented method for training a machine learning model, comprising: - receiving a training data set, the training data set comprising, for a plurality of examination objects, in each case i) a first reference representation of an examination area of the examination object in frequency space, ii) a second reference representation of the examination area of the examination object in frequency space and iii) a third reference representation of the examination area of the examination object in frequency space, where the first reference representation represents the examination area after administration of a first amount of a contrast agent, where the first amount may also be zero, where the second reference representation represents the examination area after administration of a second amount of a contrast agent and where the third reference representation represents the examination area after administration of the third amount of a contrast agent, - identifying a region within the referring expression, where the identified region includes a centre of frequency space; - reducing the representations to the identified regions, where a reduced first referring representation, a reduced second referring representation and a reduced third referring representation are obtained for each test object; - feeding, for each examination object, the reduced first reference representation and the reduced second reference representation to a machine learning model, where the machine learning model is trained to generate a reduced representation of the examination area in frequency space after administration of a third amount of contrast agent based on the reduced first reference representation and the reduced second reference representation, where the training comprises minimizing a loss function, where the loss function quantifies the deviation of the generated reduced representation of the examination area from the reduced third reference representation. The method comprising:
[0145] 10. A computer system comprising: Receiving unit, A control and computing unit, and Output unit Equipped with - wherein the control and calculation unit is configured to prompt the receiving unit to receive at least two representations of an examination area of an examination object, where the representations represent the examination area in frequency space, where the representations represent the examination area after administration of different amounts of contrast agent; - wherein the control and calculation unit is configured to reduce the received representation to a specified region, where the specified region includes a center of the frequency space; - where the control and computing unit is configured to feed the reduced representation to a machine learning model, where the machine learning model has been trained based on a training data set for learning an effect of the amount of contrast agent on the representation of the examination region in frequency space; - where the control and computation unit is configured to receive a predicted representation of the examination region from the machine learning model, where the predicted representation represents the examination region in frequency space after administration of a specific amount of contrast agent, - where the control and calculation unit is configured to supplement the predicted representation by those parts of the one or more received representations that are outside the identified region, - where the control and computation unit is configured to convert the captured predicted representation into a representation of the examination area in real space, wherein the control and computation unit is configured to prompt the output unit to output a representation of the examination area in real space. The system.
[0146] 11. A computer system comprising: Receiving unit, A control and computing unit, and Output unit Equipped with - wherein the control and computing unit is configured to prompt the receiving unit to receive at least two representations of an examination area of an examination object, where the representations represent the examination area in frequency space, where the representations represent the examination area after administration of different amounts of contrast agent; - wherein the control and calculation unit is configured to reduce the received representation to a specified region, where the specified region includes a center of the frequency space; - where the control and computing unit is configured to feed the reduced representation to a machine learning model, where the machine learning model has been trained based on a training data set for learning an effect of the amount of contrast agent on the representation of the examination region in frequency space; - where the control and computation unit is configured to receive a predicted representation of the examination region from the machine learning model, where the predicted representation represents the examination region in frequency space after administration of a particular amount of a contrast agent; - where the control and computation unit is configured to supplement the predicted representation with those parts of the one or more received representations that are not present within the identified region, - where the control and calculation unit is configured to convert the captured predicted representation into a representation of the examination area in real space, wherein the control and computation unit is configured to prompt the output unit to output a representation of the examination area in real space. The system.
[0147] 12. A computer system comprising: Receiving unit, A control and computing unit, and Output unit Equipped with - wherein the control and computing unit is configured to prompt the receiving unit to receive at least two representations of an examination area of an examination object, where the representations represent the examination area in frequency space, where the representations represent the examination area after administration of different amounts of contrast agent; - wherein the control and calculation unit is configured to reduce the received representation to a specified region, where the specified region includes a center of the frequency space; - where the control and computing unit is configured to feed the reduced representation to a machine learning model, where the machine learning model has been trained based on a training data set for learning an effect of the amount of contrast agent on the representation of the examination region in frequency space; - where the control and computation unit is configured to receive a predicted representation of the examination region from the machine learning model, where the predicted representation represents the examination region in frequency space after administration of a particular amount of a contrast agent; - where the control and calculation unit is configured to supplement the predicted representation by those parts of the one or more received representations that were cropped during the reduction, and - where the control and calculation unit is configured to convert the captured predicted representation into a representation of the examination area in real space, wherein the control and computation unit is configured to prompt the output unit to output a representation of the examination area in real space. The system.
[0148] 13. A computer program capable of being loaded into the memory of a computer system, which comprises the following: - receiving a first representation of an examination region of an examination object in frequency space, where the first representation represents the examination region without a contrast agent or after administration of a first amount of a contrast agent; - receiving a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of a contrast agent; - identifying a region of frequency space in the first representation and / or the second representation, where the identified region comprises a centre of the frequency space; - reducing the first representation and / or the second representation to the identified area; - feeding the first and second representations obtained after the reduction to a machine learning model, the machine learning model being trained on a training data set for learning the effect of the amount of contrast agent on the representation of the examination region in frequency space, - receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of the contrast agent; - if the frequency space occupied by the third representation is smaller than the frequency space occupied by the received first representation and / or the received second representation: filling the frequency space of the third representation with parts of the received first representation and / or the received second representation that are larger than the third representation; - transforming the third representation into a representation of the examination area in a real space representation; - outputting a representation of the examination area in a real space representation; A computer program that prompts a computer system to execute a program.
[0149] 14. A computer program capable of being loaded into the memory of a computer system, comprising: - receiving a first representation of an examination region of an examination object in frequency space, where the first representation represents the examination region without a contrast agent or after administration of a first amount of a contrast agent; - receiving a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of a contrast agent; - identifying an area in the first representation and / or the second representation, where the identified area comprises a center of frequency space; - reducing the first representation and / or the second representation to the identified area; - feeding the first and second representations obtained after reduction to a machine learning model, the machine learning model being trained on a training data set for learning the effect of the amount of contrast agent on the representation of the examination region in frequency space, - receiving a third representation of the examination region in frequency space from the machine learning model, where the third representation represents the examination region after administration of a third amount of the contrast agent; - supplementing the third representation with those parts of the first received and / or the second received representation that are not present in the identified region; - converting the captured third representation into a representation of the examination area in a real space representation; - outputting a representation of the examination area in a real space representation; A computer program that prompts a computer system to execute a program.
[0150] 15. Use of a contrast agent in a radiological method comprising: - generating a first representation of an examination region of the examined object in frequency space, where the first representation represents the examination region without contrast agent or after administration of a first amount of contrast agent; - generating a second representation of the examination region of the examined object in frequency space, where the second representation represents the examination region after administration of a second amount of contrast agent, where the second amount is different from the first amount; - reducing the first and second representations to a specified region, where the specified region includes a center of frequency space; - feeding the reduced first representation and the reduced second representation to a machine learning model, where the machine learning model has been trained on a training data set for learning an effect of the amount of contrast agent on the representation of the examination region in frequency space; - receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of contrast agent, where the third amount of contrast agent is different from the first amount and the second amount; - filling the frequency space of the third representation by the parts of the first received and / or the second received representation that were cut off during the downscaling, - transforming the filled third representation into a representation of the examination area in a real space representation; - outputting a representation of the examination area in a real space representation; The use comprising:
[0151] 16. Use of a contrast agent in a radiological method comprising: - generating a first representation of an examination region of the examined object in frequency space, where the first representation represents the examination region without contrast agent or after administration of a first amount of contrast agent; - generating a second representation in frequency space of the examination region of the examined object, where the second representation represents the examination region after administration of a second amount of contrast agent, where the second amount is different from the first amount; - reducing the first and second representations to a specified region, where the specified region includes a center of frequency space; - feeding the reduced first representation and the reduced second representation to a machine learning model, where the machine learning model has been trained on a training data set for learning an effect of the amount of contrast agent on the representation of the examination region in frequency space; - receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of the contrast agent, where the third amount is different from the first amount and the second amount; - supplementing the third representation with those parts of the first received and / or the second received representation that are not present in the identified region; - converting the captured third representation into a representation of the examination area in a real space representation; - outputting a representation of the examination area in a real space representation; The use comprising:
[0152] 17. A contrast agent for use in a radiological method, the method comprising: - optionally administering to the subject a first amount of a contrast agent; - generating a first representation of an examination region of the examined object in frequency space, where the first representation represents the examination region without contrast agent or after administration of a first amount of contrast agent; - administering to the subject a second amount of a contrast agent, wherein the second amount is different from the first amount; - generating a second representation in frequency space of the examination region of the examined object, where the second representation represents the examination region after administration of a second amount of contrast agent; - reducing the first and second representations to a specified region, where the specified region includes a center of frequency space; - feeding the reduced first representation and the reduced second representation to a machine learning model, where the machine learning model has been trained on a training data set for learning an effect of the amount of contrast agent on the representation of the examination region in frequency space; - receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of the contrast agent, where the third amount is different from the first amount and the second amount; - supplementing the third representation with those parts of the first received and / or the second received representation that are not present in the identified region; - converting the captured third representation into a representation of the examination area in a real space representation; - outputting a representation of the examination area in a real space representation; Includes.
[0153] 18. A contrast agent for use in a radiological method, the method comprising: - optionally administering to the subject a first amount of a contrast agent; - generating a first representation of an examination region of the examined object in frequency space, where the first representation represents the examination region without contrast agent or after administration of a first amount of contrast agent; - administering to the subject a second amount of a contrast agent, wherein the second amount is different from the first amount; - generating a second representation in frequency space of the examination region of the examined object, where the second representation represents the examination region after administration of a second amount of contrast agent; - reducing the first and second representations to a specified region, where the specified region includes a center of frequency space; - feeding the reduced first representation and the reduced second representation to a machine learning model, where the machine learning model has been trained on a training data set for learning an effect of the amount of contrast agent on the representation of the examination region in frequency space; - receiving from the machine learning model a third representation of the examination region in frequency space, where the third representation represents the examination region after administration of a third amount of contrast agent, where the third amount of contrast agent is different from the first amount and the second amount; - supplementing the third representation with those parts of the first received and / or the second received representation that are not present in the identified region; - converting the captured third representation into a representation of the examination area in a real space representation; - outputting a representation of the examination area in a real space representation; Includes.
[0154] 19. A kit comprising an imaging agent according to the invention according to any of embodiments 13 and 14 and a computer program.
[0155] Figure 11 shows, by way of example, in schematic form, one embodiment of a computer system according to the invention. The computer system (10) comprises a receiving unit (11), a control and calculation unit (12) and an output unit (13).
[0156] A "computer system" is an electronic data processing system that processes data by means of programmable computational rules. Such a system usually comprises a control and computation unit, often called a "computer", which comprises a processor for performing logical operations, a memory for loading computer programs, and peripheral devices.
[0157] In computer technology, a "peripheral" refers to any device that is connected to a computer and used to control the computer and / or as an input and output device. Examples are monitors (screens), printers, scanners, mice, keyboards, joysticks, drives, cameras, microphones, speakers, etc. Built-in ports and expansion cards are also considered peripherals in computer technology.
[0158] Modern computer systems are often divided into desktop PCs, portable PCs, laptops, notebooks, netbooks and tablet PCs, as well as so-called handhelds (e.g. smartphones), all of which may be used to implement the present invention.
[0159] Input into the computer system (e.g. for control by a user) is achieved via input means such as, for example, a keyboard, a mouse, a microphone, a touch-sensitive display, etc. Output is achieved via an output unit (13), which may in particular be a monitor (screen), a printer and / or a data storage medium.
[0160] The computer system (10) according to the invention is configured to predict, from at least two representations of the examination area in frequency space representing the examination area after administration of different amounts of contrast agent, a representation of the examination area after administration of a specific amount of contrast agent, which represents the examination area without the need to actually administer said specific amount.
[0161] The control and calculation unit (12) serves for controlling the receiving unit (11) and the output unit (13), for coordinating data and signals between the various units, for processing the representation of the examination area and for generating the artificial radiological image. It is conceivable that there are several control and calculation units.
[0162] The receiving unit (11) serves to receive a representation of the examination region. The representation may for example be transmitted from a magnetic resonance imaging system, from a computed tomography system or may be read from a data storage medium. The magnetic resonance imaging system or the computed tomography system may be a component of the computer system according to the invention. However, it is also conceivable that the computer system according to the invention is a component of the magnetic resonance imaging system or the computed tomography system. The representation may be transmitted via a network connection or a direct connection. The representation may be transmitted via wireless communication (WLAN, Bluetooth, mobile communication, etc.) and / or wired communication. It is conceivable that there are several receiving units. The data storage medium may also be a component of the computer system according to the invention or may be connected thereto, for example via a network. It is conceivable that there are several data storage media.
[0163] The representation and possibly further data (eg information about the examination subject, image acquisition parameters, etc.) are received by a receiving unit and transmitted to a control and calculation unit.
[0164] The control and computing unit is configured to generate an artificial radiation image based on the received data.
[0165] Via an output unit (13), the artificial radiation image can be displayed (e.g. on a monitor), output (e.g. via a printer) and / or stored on a data storage medium. It is conceivable that there are multiple output units.
[0166] As already explained, the present invention can be used to reduce the amount of contrast agent in a radiological examination. A predictive model can be trained to predict a third representation of an examination area of an examination object based on a first representation of the examination area of an examination object representing the examination area without contrast agent or after administration of a first amount of contrast agent, and based on a second representation of the examination area of an examination object representing the examination area after administration of a second amount of contrast agent, said third representation representing the examination area after administration of a third amount of contrast agent, said third amount being greater than the first amount and the second amount. Thus, to generate a radiological image that looks as if a third amount of contrast agent, which is greater, only the first amount (which may be zero) and the second amount of contrast agent need to be administered. The invention can be used to generate artificial radiation images which show an examination area of an examined subject after administration of one contrast agent, albeit with a different contrast agent administered.
[0167] The invention can be used to generate artificial radiological images showing the results of a different radiological examination (e.g. the results of a computed tomography scan) based on radiological images of one radiological examination (e.g. magnetic resonance imaging).
[0168] Further applications are envisioned. [Brief description of the drawings]
[0169] [Figure 1] FIG. 1 shows a schematic and exemplary relationship between a representation of an examination area in real space and a representation of the examination area in frequency space. [Diagram 2] FIG. 2 (Fig. 2) shows, diagrammatically and exemplarily, how the representations (F1), (F2), and (F3) of the test regions in frequency space as generated in FIG. 1 can be used to train a predictive model (PM). [Diagram 3] FIG. 3 exemplarily and diagrammatically illustrates a preferred embodiment of the method according to the invention for training a machine learning model in the form of a flow chart. [Figure 4] The use of a trained predictive model for prediction is exemplarily and diagrammatically shown in FIG. 4 (FIG. 4). [Diagram 5] FIG. 5 exemplarily and diagrammatically illustrates in the form of a flow chart a preferred embodiment of the method according to the invention for predicting a representation of a test area of a test object. [Figure 6] FIG. 6 exemplarily illustrates diagrammatically the steps in training a predictive model according to a preferred embodiment of the present invention. [Figure 7] FIG. 7 exemplarily and diagrammatically illustrates how the predictive model trained in FIG. 6 can be used for prediction. [Figure 8] FIG. 8 illustrates, in the form of a flow chart, a preferred embodiment of the method according to the present invention for training a machine learning model. [Figure 9] FIG. 9 illustrates, in the form of a flow chart, a schematic representation of a preferred embodiment of a method according to the invention for predicting a representation of a care area of a care object. [Figure 10] FIG. 10 shows, in the form of a flow chart, a further preferred embodiment of a method according to the invention for predicting a representation of a care area of a care object. [Figure 11] FIG. 11 illustrates, by way of example, in schematic form, one embodiment of a computer system according to the present invention.
Claims
1. 1. A computer-implemented method comprising: - receiving a first representation of an examination region of an examination object in frequency space, said first representation representing the examination region without contrast agent or after administration of a first amount of contrast agent; - receiving a second representation of the examination region of the object in frequency space, said second representation being representative of said examination region after administration of a second amount of a contrast agent; - feeding at least a portion of the first representation and at least a portion of the second representation to a machine learning model; receiving a third representation of the examination area in frequency space from the machine learning model, the third representation representing the examination area after administration of a third amount of a contrast agent; - generating a representation of said examination area in a real space representation based on said third representation; - outputting and / or storing said representation of said examination area in said real space representation; The method comprising:
2. 2. The computer-implemented method of claim 1, comprising: Providing at least a portion of the first representation and at least a portion of the second representation to a machine learning model includes: - identifying a region in the received first representation and in the received second representation, the identified region including a center of the frequency space; - reducing said first representation and said second representation to said identified area, resulting in a reduced first representation and a reduced second representation; - feeding said reduced first representation and said reduced second representation to said machine learning model; The method comprising:
3. 3. A computer-implemented method as recited in claim 2, comprising: generating a representation of the examination area in a real space representation based on the third representation, comprising: - supplementing said third representation with those parts of said received first representation and / or said received second representation that are not present in said identified region; - transforming said captured third representation, obtained after said third representation has been captured, into a representation of said examination area in a real space representation; The method comprising:
4. 4. A computer-implemented method according to claim 1, wherein the first representation represents the examination area without contrast agent or after administration of a first amount of the contrast agent, the second representation represents the examination area after administration of a second amount of the contrast agent, the second amount not equal to the first amount, and the third representation represents the examination area after administration of a third amount of the contrast agent, the third amount not equal to the first amount and / or not equal to the second amount.
5. 5. The computer-implemented method of claim 1, wherein the first amount of contrast agent is greater than or equal to zero, the second amount of contrast agent is greater than the first amount of contrast agent, and the third amount of contrast agent is greater than the second amount of contrast agent.
6. 6. A computer-implemented method according to claim 1 , wherein the first representation represents the examination region without a contrast agent or after administration of a first amount of a first contrast agent, the second representation represents the examination region after administration of a second amount of a second contrast agent, and the third representation represents the examination region after administration of a third amount of a third contrast agent, the third contrast agent being different from the first contrast agent and / or different from the second contrast agent, or the second contrast agent being different from the first contrast agent.
7. A computer-implemented method according to any preceding claim, wherein the first and second representations are results of a radiological examination.
8. 8. The computer-implemented method of claim 7, wherein the radiological examination is a magnetic resonance imaging examination, a computed tomography examination, or an ultrasound examination.
9. A computer implemented method according to any preceding claim, wherein the first representation and the second representation are k-space data of a magnetic resonance imaging examination.
10. 10. A computer-implemented method according to any preceding claim, wherein the first representation and the second representation are Fourier transformed real space representations.
11. A computer-implemented method according to any one of claims 1 to 7, comprising: - receiving a first real space representation of the examination area of the examination object, said first real space representation representing said examination area without a contrast agent or after administration of a first amount of a contrast agent; - receiving a second real space representation of the examination area of the examination object, said second real space representation representing said examination area after administration of a second amount of a contrast agent; generating, from said first real space representation, preferably by means of a Fourier transformation, said first representation of said examination area of said object in frequency space; - generating said second representation of said examination area of said object in frequency space from said second real space representation, preferably by means of a Fourier transformation; The method further comprises:
12. 12. The computer-implemented method of claim 1, further comprising training the machine learning model, the training comprising: a sub-step of receiving a training data set, said training data set comprising, for a plurality of examination objects, in each case i) a first reference representation of an examination area of said examination object in frequency space, ii) a second reference representation of the examination area of said examination object in frequency space and iii) a third reference representation of the examination area of said examination object in frequency space, said first reference representation representing said examination area after administration of a first amount of contrast agent, said first amount being zero, said second reference representation representing said examination area after administration of a second amount of contrast agent and said third reference representation representing said examination area after administration of a third amount of contrast agent; - identifying a region within said referring expression, said identified region including the centre of said frequency space; - reducing said representations to said identified regions, in which a reduced first referring representation, a reduced second referring representation and a reduced third referring representation are obtained for each test object; for each examination object: feeding said reduced first and said reduced second reference representation to a machine learning model, said machine learning model being trained to generate a reduced representation of said examination area in frequency space after administration of said third amount of contrast agent based on said reduced first and said reduced second reference representation, said training comprising the minimization of a loss function, said loss function quantifying the deviation of said generated reduced representation of said examination area from said reduced third reference representation; The method comprising:
13. 1. A computer system comprising: a receiving unit; a control and calculation unit; and an output unit; Equipped with the control and calculation unit is configured to prompt the receiving unit to receive at least two representations, a first representation and a second representation, of an examination area of an examination object, the first representation representing the examination area without contrast agent or after administration of a first amount of contrast agent and the second representation representing the examination area after administration of a second amount of contrast agent; the control and computation unit is configured to provide at least a portion of the first representation and at least a portion of the second representation to a machine learning model; the control and computation unit is configured to receive a third representation of the examination region from the machine learning model, the third representation representing the examination region in frequency space after administration of a third amount of a contrast agent; the control and calculation unit is adapted to generate a representation of the examination area in real space based on the third representation; the control and calculation unit is configured to prompt the output unit to output and / or store the representation of the examination area in real space; The system.
14. A computer program loadable into the memory of a computer system, said computer program comprising: - receiving a first representation of an examination region of an examination object in frequency space, said first representation representing said examination region without a contrast agent or after administration of a first amount of a contrast agent; - receiving a second representation of the examination region of the object in frequency space, said second representation being representative of said examination region after administration of a second amount of a contrast agent; - feeding at least a portion of said first representation and at least a portion of said second representation to a machine learning model; receiving from the machine learning model a third representation of the examination area in frequency space, the third representation representing the examination area after administration of a third amount of a contrast agent; - generating a representation of said examination area in a real space representation based on said third representation; - outputting and / or storing said representation of said examination area in said real space representation; and instructing a computer system to execute the computer program.
15. Use of a contrast agent in a radiological method, said radiological method comprising: - generating a first representation of an examination area of an examination object in frequency space, said first representation representing said examination area without said contrast agent or after administration of a first amount of said contrast agent; - generating a second representation of the examination area of the examination object in frequency space, said second representation representing said examination area after administration of a second amount of said contrast agent, said second amount being different from said first amount; - feeding at least a portion of said first representation and at least a portion of said second representation to a machine learning model; receiving a third representation of the examination area in frequency space from the machine learning model, the third representation representing the examination area after administration of a third amount of the contrast agent, the third amount being different from the first amount and the second amount; - generating a representation of said examination area in a real space representation based on said third representation; - outputting and / or storing said representation of said examination area in said real space representation; The use comprising:
16. 1. A contrast agent for use in a radiological method, the method comprising: - generating a first representation of an examination area of an examination object in frequency space, said first representation representing said examination area without said contrast agent or after administration of a first amount of said contrast agent; - generating a second representation of the examination area of the examination object in frequency space, said second representation being representative of said examination area after administration of said second amount of said contrast agent; - feeding at least a portion of said first representation and at least a portion of said second representation to a machine learning model; receiving a third representation of the examination area in frequency space from the machine learning model, the third representation representing the examination area after administration of a third amount of the contrast agent, the third amount of the contrast agent being different from the first amount and the second amount; - generating a representation of said examination area in a real space representation based on said third representation; - outputting and / or storing said representation of said examination area in said real space representation; The contrast agent comprising:
17. A kit comprising an imaging agent and a computer program according to claim 14.
