Systems, methods, and computer program products for contrast radiology using machine learning

Abstract

A method for providing a prediction of a representation of an examination area generated using a medical imaging technique involving a contrast agent includes receiving a first representation in frequency space of an examination area of ​​an examination subject, receiving a second representation in the frequency space of the examination area of ​​the examination subject, providing the first and second representations as inputs to a predictive machine learning model configured to provide as an output a prediction of the representation in the frequency space of the examination area using an amount of the contrast agent administered during the medical imaging technique, receiving an output of the predictive machine learning model based on the input, and converting the output of the predictive machine learning model to a representation in real space of the examination area of ​​the examination subject.

Description

[Technical field]

[0001] REFERENCE TO RELATED APPLICATIONS This application claims the benefit of European Application No. EP21160325.3, filed March 2, 2021, and European Application No. EP21167116.9, filed April 7, 2021, the entire disclosures of which are hereby incorporated by reference in their entireties. background

[0002] 1. Field The present disclosure relates generally to generating radiological images, and in some non-limiting embodiments, to systems, methods, and computer program products for generating predictions of representations of radiological images generated using medical imaging techniques that require contrast agents. [Background technology]

[0003] 2. Technical Considerations Radiology may refer to the branch of medicine that uses medical imaging techniques to diagnose and treat disease. Radiology may involve the application of electromagnetic radiation and mechanical waves, such as during an ultrasound procedure, to produce images for diagnostic, therapeutic, and / or scientific purposes. In some instances, ionizing radiation, such as x-ray radiation, gamma radiation, and / or electrons, may be used to produce images during medical imaging procedures. Radiology also includes other imaging methods, such as computed tomography (CT), positron emission tomography (PET), ultrasonography, and magnetic resonance imaging (MRI), also known as nuclear magnetic resonance imaging (NMRI), although these medical imaging techniques do not use ionizing radiation. In some instances, a contrast agent (e.g., a radiological contrast material, a radiocontrast agent, a contrast media, etc.), and / or a wash, such as saline, may be used in medical imaging techniques, such as angiography, CT, ultrasound, and MRI. Contrast agents may be used (e.g., injected into a patient's bloodstream) to provide contrast enhancement to images generated based on medical imaging techniques. For example, a contrast agent may include a substance or mixture of substances that improves the depiction of structures and / or functions of a patient's (e.g., a human or animal patient's) body during a medical imaging technique.

[0004] CT scans (e.g., computed axial tomography (CAT) scans) may include medical imaging techniques used in radiology to non-invasively obtain detailed images of the body for diagnostic purposes. A device for performing a CT scan may be a CT scanner. A CT scanner may use a rotating X-ray tube and an array of detectors arranged in a gantry to measure X-ray attenuation by different tissues within a patient's body. Multiple X-ray measurements taken from different angles may be processed on a computer using reconstruction algorithms to generate a tomographic (e.g., cross-sectional) image of the body.

[0005] MRI may include medical imaging techniques used in medical diagnosis, particularly for depicting the structure and function of tissues and / or organs within a patient's body. In MRI, the magnetic moments of protons in an examination subject are aligned in a basic magnetic field, resulting in the presence of macroscopic magnetization along the longitudinal direction. The magnetic moments of the examination subject are then deflected from a rest position (e.g., a relaxed position) by irradiation with a radio frequency (HF) pulse (e.g., by excitation with an HF pulse of electromagnetic radiation). The return of the protons from the excited state to the rest position, or magnetization dynamics, can then be detected as a relaxation signal by one or more HF receiving coils of the MRI machine. For spatial encoding, rapidly switched gradient magnetic fields can be superimposed on the basic magnetic field. The relaxation signal initially exists as raw data (e.g., detected MRI data) in frequency space (e.g., frequency domain, spatial frequency space, Fourier space, Fourier depiction, etc.) and can be transformed to real space (e.g., image space) by inverse Fourier transformation. In one embodiment, during the use of native MRI, tissue contrast is determined by the various relaxation times (e.g., T 1 and T 2 The spin-lattice relaxation time (SLT) may be generated by the time t of the magnetization (referred to as the spin-lattice relaxation time and the spin-spin relaxation time) and / or the proton density. The spin-lattice relaxation time can describe the transition of the longitudinal magnetization to its equilibrium magnetization state, and the spin-lattice relaxation time is measured as the time it takes to reach 63.21% of the equilibrium magnetization before resonant excitation. The spin-spin relaxation time can, in a similar manner, describe the transition of the transverse magnetization to its equilibrium magnetization state.

[0006] During CT imaging procedures, iodine-containing solutions can be used as contrast agents. During MRI imaging procedures, superparamagnetic substances (e.g., iron oxide nanoparticles, superparamagnetic iron-platinum particles (SIPP), etc.) and / or paramagnetic substances (e.g., gadolinium chelates, manganese chelates, etc.) can be used as contrast agents. In ultrasound imaging procedures, liquids containing gas-filled microbubbles may be used, which are administered intravenously. MRI contrast agents exert their effect by altering the relaxation time of structures that incorporate the agent. Two groups of substances can be distinguished: paramagnetic substances and superparamagnetic substances. Both groups of substances have unpaired electrons that induce a magnetic field around individual atoms or molecules. Superparamagnetic contrast agents primarily cause T2 shortening, while paramagnetic contrast agents primarily cause T1 shortening. The effect of contrast agents may be indirect, since the agent itself does not emit a signal, but instead merely affects the intensity of the signal in the vicinity of the agent. An example of a superparamagnetic contrast agent is an iron oxide nanoparticle (e.g., superparamagnetic iron oxide (SPIO)). Examples of paramagnetic contrast agents are gadolinium chelates such as gadopentetate dimeglumine (e.g., Magnevist®), gadoteric acid (e.g., Dotarem®, Dotagita®, Cyclolux®), gadodiamide (e.g., Omniscan®), gadoteridol (e.g., ProHance®), and gadobutrol (e.g., Gadovist®). Summary of the Invention [Problem to be solved by the invention]

[0007] summary

[0008] Thus, a system, method, and computer program product are disclosed for providing a prediction of a representation of a region of examination generated using a medical imaging technique that requires a contrast agent. [Means for solving the problem]

[0009] According to certain non-limiting embodiments, provided is a system for providing a prediction of a representation of an examination region generated using a medical imaging technique requiring a contrast agent, comprising: receiving a first representation in frequency space of an examination region of an object, where the first representation includes a representation of the examination region without an amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; receiving a second representation in frequency space of an examination region of an object, where the second representation includes a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent is different from the first amount of contrast agent, and providing an input to a predictive machine learning model, where the predictive machine learning model a predictive machine learning model configured to provide, as an output, a prediction of a representation in frequency space of the examination region of the subject using a third amount of contrast agent administered during a medical imaging technique, where the third amount of contrast agent is greater than the first amount of contrast agent and the second amount of contrast agent; and at least one processor programmed or configured to receive an output of the predictive machine learning model based on the input, transform the output of the predictive machine learning model into a predicted representation in real space of the examination region of the subject, and provide the predicted representation of the examination region of the subject in real space.

[0010] According to certain non-limiting embodiments, provided is a computer program product for providing a prediction of a representation of an examination region generated using a medical imaging technique requiring a contrast agent, the computer program product, when executed by at least one processor, causing the at least one processor to receive a first representation in frequency space of an examination region of an object, where the first representation includes a representation of the examination region without an amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique, receive a second representation in frequency space of an examination region of an object, where the second representation includes a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent is different from the first amount of contrast agent, and generate a prediction of the representation of the examination region using a predictive machine learning model. a predictive machine learning model configured to provide an input to a predictive machine learning model based on the input, where the input to the predictive machine learning model includes at least a portion of the first representation and at least a portion of the second representation, where the predictive machine learning model is configured to provide, as an output, a prediction of a representation in frequency space of the examination region of the examination using a third amount of a contrast agent administered during a medical imaging technique, where the third amount of the contrast agent is greater than the first amount of the contrast agent and the second amount of the contrast agent; receive an output of the predictive machine learning model based on the input; transform the output of the predictive machine learning model into a predicted representation in real space of the examination region of the examination; and provide the predicted representation in real space of the examination region of the examination subject.

[0011] According to certain non-limiting embodiments, provided is a method for providing a prediction of a representation of an examination region generated using a medical imaging technique requiring a contrast agent, comprising: receiving, by at least one processor, a first representation in frequency space of an examination region of an object, where the first representation includes a representation of the examination region without an amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; receiving, by the at least one processor, a second representation in frequency space of the examination region of the object, where the second representation includes a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent is different from the first amount of contrast agent; and providing, by the at least one processor, an input to a predictive machine learning model. receiving, by at least one processor, an output of the predictive machine learning model based on the input; converting, by the at least one processor, the output of the predictive machine learning model into a predicted representation in real space of the examination region of the subject; and providing, by the at least one processor, the predicted representation in real space of the examination region of the subject.

[0012] Further non-limiting embodiments are described in the following numbered paragraphs:

[0013] Item 1: A system for providing a prediction of a representation of an examination region generated using a medical imaging technique requiring a contrast agent, comprising: receiving a first representation in frequency space of an examination region of an object, where the first representation includes a representation of the examination region without an amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; receiving a second representation in frequency space of an examination region of an object, where the second representation includes a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent is different from the first amount of contrast agent; and providing an input to a predictive machine learning model, where the input to the predictive machine learning model includes the first representation in frequency space of an examination region of an object, where the first representation includes a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent is different from the first amount of contrast agent. and at least a portion of the first representation and at least a portion of the second representation, where the predictive machine learning model comprises a trained machine learning model configured to provide, as an output, a prediction of a representation in frequency space of the examination region of the subject using a third amount of a contrast agent administered during a medical imaging technique, where the third amount of the contrast agent is greater than the first amount of the contrast agent and the second amount of the contrast agent; receiving an output of the predictive machine learning model based on the input; converting the output of the predictive machine learning model into a predicted representation in real space of the examination region of the subject; and providing the predicted representation in real space of the examination region of the subject.

[0014] Clause 2: The system of clause 1, wherein the at least one processor is further programmed or configured to: provide a reduced representation in frequency space of the inspection area of ​​the object to identify a portion of the first representation comprising a center in frequency space, a portion of the second representation comprising a center in frequency space, or a portion of the first representation and a portion of the second representation, each comprising a center in frequency space; wherein the input to the predictive machine learning model includes the reduced representation in frequency space of the inspection area of ​​the object; and wherein, when providing the input to the predictive machine learning model, the at least one processor is programmed or configured to: provide the reduced representation in frequency space of the inspection area of ​​the object as an input to the predictive machine learning model.

[0015] Clause 3: The system of clause 2, wherein the at least one processor is further programmed or configured to: provide a supplemented output of the predictive machine learning model to supplement an output of the predictive machine learning model with: a portion of the first representation that does not include a center of frequency space and that was not specified to provide a reduced representation in frequency space of the inspection area of ​​the subject to be inspected, a portion of the second representation that does not include a center of frequency space and that was not specified to provide a reduced representation in frequency space of the inspection area of ​​the subject to be inspected, or a portion of the first representation and a portion of the second representation that do not include a center of frequency space and that were not specified to provide a reduced representation in frequency space of the inspection area of ​​the subject to be inspected; wherein when converting the output of the predictive machine learning model to a predicted representation in real space of the inspection area of ​​the subject to be inspected, the at least one processor is programmed or configured to: convert the supplemented output of the predictive machine learning model to a predicted representation in real space of the inspection area of ​​the subject to be inspected.

[0016] Clause 4: A system as described in any of clauses 1 to 3, wherein a first amount of contrast agent administered during the medical imaging technique is greater than zero, and wherein a second amount of contrast agent administered during the medical imaging technique is greater than the first amount of contrast agent.

[0017] Clause 5: A system as described in any of clauses 1 to 3, wherein a first amount of contrast agent administered during the medical imaging technique is zero, and wherein a second amount of contrast agent administered during the medical imaging technique is greater than the first amount of contrast agent.

[0018] Clause 6: A system as described in any of clauses 1 to 5, wherein at least one processor is further programmed or configured to: generate a first representation in frequency space of the examination area of ​​the object for a first result of the radiological examination; and generate a second representation in frequency space of the examination area of ​​the object for a second result of the radiological examination.

[0019] Clause 7: The system of clause 6, wherein the radiological examination is a magnetic resonance imaging examination, a computed tomography examination, or an ultrasound examination.

[0020] Clause 8: A system as described in clause 6, wherein the radiological examination is a magnetic resonance imaging examination, and wherein, when receiving a first representation in frequency space of an examination region of the subject in frequency space, the at least one processor is programmed or configured to: receive first k-space data related to a magnetic resonance imaging examination of the examination region of the subject; and, when receiving a second representation in frequency space of the examination region of the subject in frequency space, the at least one processor is programmed or configured to: receive second k-space data related to a magnetic resonance imaging examination of the examination region of the subject.

[0021] Clause 9: A system according to any one of clauses 1 to 7, wherein, when receiving a first representation in frequency space of an examination region of the subject in frequency space, at least one processor is programmed or configured to receive a first representation in real space of the examination region of the subject, where the first representation in real space includes a representation of the examination region without an amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; convert the first representation in real space of the examination region of the subject into a first representation in frequency space; and, when receiving a second representation in frequency space of the examination region of the subject in frequency space, at least one processor is programmed or configured to receive a second representation in real space of the examination region of the subject, where the second representation in real space includes a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent is different from the first amount of contrast agent; convert the first representation in real space of the examination region of the subject into a first representation in frequency space.

[0022] Clause 10: The system according to any one of clauses 1 to 9, wherein the at least one processor is further programmed or configured to train a predictive machine learning model based on a training dataset, wherein the training dataset comprises: a set of reference representations in a frequency space of the examination area for each examination object of a plurality of examination objects, each set of reference representations of the examination area of ​​the examination object comprising: a first reference representation in the frequency space of the examination area of ​​the examination object; a second reference representation in the frequency space of the examination area of ​​the examination object; and a third reference representation in the frequency space of the examination area of ​​the examination object; and wherein the first reference representation comprises a reference representation in the frequency space of the examination area with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; wherein the second reference representation comprises a reference representation in the frequency space of the examination area with a second amount of contrast agent administered during the medical imaging technique; and wherein the third reference representation comprises a reference representation in the frequency space of the examination area with a third amount of contrast agent administered during the medical imaging technique.

[0023] Clause 11: The system described in clause 10, wherein when training the predictive machine learning model, at least one processor is programmed or configured to minimize an amount of error provided by an error function, wherein the error function quantifies the deviation between a prediction of a representation in frequency space of the examination region of the subject using a third amount of contrast agent administered during the medical imaging technique and a third reference representation in frequency space of the examination region of the subject.

[0024] Clause 12: The system according to any one of clauses 1 to 11, wherein the at least one processor is further programmed or configured to train a predictive machine learning model based on a training dataset, wherein the training dataset includes a set of reduced referring representations in a frequency space of the test area for each test object of the plurality of test objects, each set of reduced referring representations of the test area of ​​the test object includes: a first reduced referring representation in the frequency space of the test area of ​​the test object; a second reduced referring representation in the frequency space of the test area of ​​the test object; and a third reduced referring representation in the frequency space of the test area of ​​the test object; and wherein the first reduced referring representation is a reduced referring representation in the frequency space of the test area that includes a center of the frequency space. a first reference representation in a frequency space of the examination region including a center of the frequency space, the first reference representation comprising a reference representation of the examination region with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; a second reduced reference representation comprising a reference representation of a portion of the second reference representation in a frequency space of the examination region including a center of the frequency space, the second reference representation comprising a reference representation of the examination region with a second amount of contrast agent administered during the medical imaging technique; and a third reduced reference representation comprising a reference representation of a portion of the third reference representation in a frequency space of the examination region including a center of the frequency space, the third reference representation comprising a reference representation of the examination region with a third amount of contrast agent administered during the medical imaging technique.

[0025] Clause 13: The system described in clause 12, wherein when training the predictive machine learning model, at least one processor is programmed or configured to minimize an amount of error provided by an error function, wherein the error function quantifies the deviation between a prediction of a reduced representation in frequency space of the examination region of the subject using a third amount of contrast agent administered during the medical imaging technique and a third reduced reference representation in frequency space of the examination region of the subject.

[0026] Clause 14: A computer program product for providing a prediction of a representation of an examination region generated using a medical imaging technique requiring a contrast agent, the computer program product, when executed by at least one processor, causing the at least one processor to: receive a first representation in frequency space of an examination region of an object, where the first representation includes a representation of the examination region with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; receive a second representation in frequency space of an examination region of an object, where the second representation includes a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent is different from the first amount of contrast agent; and provide input to a predictive machine learning model, the computer program product comprising: 1. A computer program product comprising: at least one non-transitory computer-readable medium comprising one or more instructions: an input to the predictive machine learning model includes at least a portion of the first representation and at least a portion of the second representation, where the predictive machine learning model includes a trained machine learning model configured to provide as an output a prediction of a representation in frequency space of the examination region using a third amount of a contrast agent administered during a medical imaging technique, where the third amount of the contrast agent is greater than the first amount of the contrast agent and the second amount of the contrast agent; receiving an output of the predictive machine learning model based on the input; converting the output of the predictive machine learning model into a predicted representation in real space of the examination region of the examination; and providing the predicted representation in real space of the examination region of the examination subject.

[0027] Clause 15: The computer program product of clause 14, wherein the one or more instructions further cause at least one processor to identify a portion of the first representation comprising a center of frequency space, a portion of the second representation comprising a center of frequency space, or a portion of the first representation and a portion of the second representation, each comprising a center of frequency space, and provide a reduced representation in frequency space of the inspection area of ​​the object to be inspected; wherein the input to the predictive machine learning model comprises the reduced representation in frequency space of the inspection area of ​​the object to be inspected; and wherein the one or more instructions causing the at least one processor to provide an input to the predictive machine learning model cause the at least one processor to provide the reduced representation in frequency space of the inspection area of ​​the object to be inspected as an input to the predictive machine learning model.

[0028] Clause 16: The computer program product of clause 15, wherein the one or more instructions further cause the at least one processor to: supplement an output of the predictive machine learning model with: a portion of the first representation that does not include a center of frequency space and that was not specified to provide a reduced representation in frequency space of the inspection area of ​​the subject to be inspected, a portion of the second representation that does not include a center of frequency space and that was not specified to provide a reduced representation in frequency space of the inspection area of ​​the subject to be inspected, or a portion of the first representation and a portion of the second representation, each of which does not include a center of frequency space and that was not specified to provide a reduced representation in frequency space of the inspection area of ​​the subject to be inspected; provide the supplemented output of the predictive machine learning model; wherein the one or more instructions causing the at least one processor to transform the output of the predictive machine learning model into a predicted representation in real space of the inspection area of ​​the subject to be inspected cause the at least one processor to: transform the supplemented output of the predictive machine learning model into a predicted representation in real space of the inspection area of ​​the subject to be inspected.

[0029] Clause 17: A computer program product according to any of clauses 14 to 16, wherein a first amount of contrast agent administered during the medical imaging technique is greater than zero, and wherein a second amount of contrast agent administered during the medical imaging technique is greater than the first amount of contrast agent.

[0030] Clause 18: A computer program product according to any of clauses 14 to 16, wherein a first amount of contrast agent administered during the medical imaging technique is zero, and wherein a second amount of contrast agent administered during the medical imaging technique is greater than the first amount of contrast agent.

[0031] Clause 19: A computer program product according to any of clauses 14 to 18, wherein the one or more instructions further cause at least one processor to: generate a first representation in frequency space of the inspection area of ​​the inspected object for a first result of the radiological examination; and generate a second representation in frequency space of the inspection area of ​​the inspected object for a second result of the radiological examination.

[0032] Clause 20: The computer program product of clause 19, wherein the radiological examination is a magnetic resonance imaging examination, a computed tomography examination, or an ultrasound examination.

[0033] Clause 21: A computer program product as described in clause 19, wherein the radiological examination is a magnetic resonance imaging examination, and wherein the one or more instructions causing at least one processor to receive a first representation in frequency space of an examination area of ​​the examination subject cause the at least one processor to: receive first k-space data related to a magnetic resonance imaging examination of the examination area of ​​the examination subject; and wherein the one or more instructions causing the at least one processor to receive a second representation in frequency space of the examination area of ​​the examination subject cause the at least one processor to: receive second k-space data related to a magnetic resonance imaging examination of the examination area of ​​the examination subject.

[0034] Clause 22: A computer program product according to any of clauses 14 to 20, wherein the one or more instructions for causing at least one processor to receive a first representation in frequency space of an examination region of an examined object include causing the at least one processor to: receive a first representation in real space of an examination region of an examined object, where the first representation in real space includes a representation of the examination region without an amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; and converting the first representation in real space of the examination region of the examined object into a first representation in frequency space. and converting the first representation in real space of the examination region of the subject into a first representation in frequency space, wherein the one or more instructions cause the at least one processor to: receive a second representation in real space of the examination region of the subject, where the second representation in real space includes a second representation of the examination region with a second amount of a contrast agent administered during a medical imaging technique, where the second amount of the contrast agent differs from the first amount of the contrast agent; and converting the first representation in real space of the examination region of the subject into a first representation in frequency space.

[0035] Clause 23: A computer program product according to any one of clauses 14 to 22, wherein one or more instructions further cause at least one processor to train a predictive machine learning model based on a training dataset, wherein the training dataset includes: a set of reference representations in a frequency space of an examination area for each examination object of a plurality of examination objects, each set of reference representations of the examination area of ​​the examination object including: a first reference representation in a frequency space of the examination area of ​​the examination object; a second reference representation in a frequency space of the examination area of ​​the examination object; and a third reference representation in a frequency space of the examination area of ​​the examination object; and wherein the first reference representation in the frequency space includes a reference representation of the examination area with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; wherein the second reference representation in the frequency space includes a reference representation of the examination area with a second amount of contrast agent administered during the medical imaging technique; and wherein the third reference representation in the frequency space includes a reference representation of the examination area with a third amount of contrast agent administered during the medical imaging technique.

[0036] Clause 24: A computer program product as described in clause 23, wherein the one or more instructions for causing at least one processor to train a predictive machine learning model cause the at least one processor to: minimize an amount of error provided by an error function, where the error function quantifies the deviation between a prediction of a representation in frequency space of the examination region of the subject using a third amount of a contrast agent administered during the medical imaging technique and a third reference representation in frequency space of the examination region of the subject.

[0037] Clause 25: The computer program product according to any one of clauses 14 to 24, wherein the one or more instructions further cause at least one processor to: train a predictive machine learning model based on a training dataset, wherein the training dataset includes: a set of reduced referring representations in a frequency space of the test area for each test object of the plurality of test objects, each set of reduced referring representations of the test area of ​​the test object includes: a first reduced referring representation in the frequency space of the test area of ​​the test object; a second reduced referring representation in the frequency space of the test area of ​​the test object; and a third reduced referring representation in the frequency space of the test area of ​​the test object; and wherein the first reduced referring representation is a reduced referring representation in the frequency space of the test area that includes a center of the frequency space. a first reference representation including a reference representation of a portion of a first reference representation, where the first reference representation includes a representation of an examination area with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; a second reduced reference representation including a reference representation of a portion of the second reference representation in a frequency space of the examination area including a center of the frequency space, where the second reference representation includes a reference representation of an examination area with a second amount of contrast agent administered during the medical imaging technique; and a third reduced reference representation including a reference representation of a portion of the third reference representation in a frequency space of the examination area including a center of the frequency space, where the third reference representation includes a reference representation of an examination area with a third amount of contrast agent administered during the medical imaging technique.

[0038] Clause 26: A computer program product as described in clause 25, wherein the one or more instructions for causing at least one processor to train a predictive machine learning model cause the at least one processor to: minimize an amount of error provided by an error function, where the error function quantifies the deviation between a prediction of a reduced representation in frequency space of the examination area of ​​the subject using a third amount of a contrast agent administered during a medical imaging technique and a third reduced reference representation in frequency space of the examination area of ​​the subject.

[0039] Clause 27: A method for providing a prediction of a representation of an examination region generated using a medical imaging technique requiring a contrast agent, comprising: receiving, by at least one processor, a first representation in frequency space of an examination region of an object, where the first representation includes a representation of the examination region without an amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; receiving, by the at least one processor, a second representation in frequency space of an examination region of an object, where the second representation includes a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent is different from the first amount of contrast agent; providing, by the at least one processor, input to a predictive machine learning model, where the predictive machine learning model determines the prediction of the representation of the examination region. The method includes: input to the learning model includes at least a portion of the first representation and at least a portion of the second representation, where the predictive machine learning model includes a trained machine learning model configured to provide as an output a prediction of a representation in frequency space of the examination region using a third amount of contrast agent administered during a medical imaging technique, where the third amount of contrast agent is greater than the first amount of contrast agent and the second amount of contrast agent; receiving, by at least one processor, an output of the predictive machine learning model based on the input; converting, by the at least one processor, the output of the predictive machine learning model into a predicted representation in real space of the examination region of the subject; and providing, by the at least one processor, the predicted representation in real space of the examination region of the subject.

[0040] Clause 28: The method of clause 27, further comprising identifying a portion of the first representation comprising a center in frequency space, a portion of the second representation comprising a center in frequency space, or a portion of the first representation and a portion of the second representation, each of which comprises a center in frequency space, and providing a reduced representation in frequency space of the inspection area of ​​the object; wherein the input to the predictive machine learning model comprises the reduced representation in frequency space of the inspection area of ​​the object; and wherein providing the input to the predictive machine learning model comprises: providing a predictive machine learning model using the reduced representation in frequency space of the inspection area of ​​the object as input.

[0041] Clause 29: The method of clause 28, comprising: supplementing an output of the predictive machine learning model with a portion of the first representation that does not include a center of frequency space and that was not specified to provide a reduced representation in frequency space of the inspection area of ​​the subject, a portion of the second representation that does not include a center of frequency space and that was not specified to provide a reduced representation in frequency space of the inspection area of ​​the subject, or a portion of the first representation and a portion of the second representation, each of which does not include a center of frequency space and that were not specified to provide a reduced representation in frequency space of the inspection area of ​​the subject; providing the supplemented output of the predictive machine learning model; wherein transforming the output of the predictive machine learning model to a representation in real space of the inspection area of ​​the subject includes transforming the supplemented output of the predictive machine learning model to a representation in real space of the inspection area of ​​the subject.

[0042] Clause 30: A method according to any of clauses 27 to 29, wherein a first amount of contrast agent administered during the medical imaging technique is greater than zero, and wherein a second amount of contrast agent administered during the medical imaging technique is greater than the first amount of contrast agent.

[0043] Item 31: A method according to any of items 27 to 29, wherein a first amount of contrast agent administered during the medical imaging technique is zero, and a second amount of contrast agent administered during the medical imaging technique is greater than the first amount of contrast agent.

[0044] Clause 32: A method according to any of clauses 27 to 31, further comprising: generating a first representation in frequency space of the examination area of ​​the object for a first result of the radiological examination; and generating a second representation in frequency space of the examination area of ​​the object for a second result of the radiological examination.

[0045] Clause 33: The method of clause 32, wherein the radiological examination is a magnetic resonance imaging examination, a computed tomography examination, or an ultrasound examination.

[0046] Clause 34: A method as described in clause 32, wherein the radiological examination is a magnetic resonance imaging examination, and wherein receiving a first representation in frequency space of an examination area of ​​the examination subject includes: receiving first k-space data related to a magnetic resonance imaging examination of the examination area of ​​the examination subject; and wherein receiving a second representation in frequency space of the examination area of ​​the examination subject includes: receiving second k-space data related to a magnetic resonance imaging examination of the examination area of ​​the examination subject.

[0047] Clause 35: A method according to any one of clauses 27 to 33, wherein receiving a first representation in frequency space of an examination area of ​​the subject includes: receiving a first representation in real space of the examination area of ​​the subject, where the first representation in real space includes a representation of the examination area without an amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; converting the first representation in real space of the examination area of ​​the subject to a first representation in frequency space; and wherein receiving a second representation in frequency space of the examination area of ​​the subject includes: receiving a second representation in real space of the examination area of ​​the subject, where the second representation in real space includes a representation of the examination area with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent is different from the first amount of contrast agent; converting the first representation in real space of the examination area of ​​the subject to a first representation in frequency space.

[0048] Clause 36: A method according to any one of clauses 27 to 35, further comprising: training a predictive machine learning model based on a training dataset, wherein the training dataset comprises: a set of reference representations in the frequency space of the examination area for each examination object of a plurality of examination objects, each set of reference representations in the frequency space of the examination area of ​​the examination object comprising: a first reference representation in the frequency space of the examination area of ​​the examination object; a second reference representation in the frequency space of the examination area of ​​the examination object; and a third reference representation in the frequency space of the examination area of ​​the examination object; and wherein the first reference representation comprises a reference representation of the examination area with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; wherein the second reference representation comprises a reference representation of the examination area with a second amount of contrast agent administered during the medical imaging technique; and wherein the third reference representation comprises a reference representation of the examination area with a third amount of contrast agent administered during the medical imaging technique.

[0049] Clause 37: A method as described in clause 36, wherein training the predictive machine learning model comprises: minimizing an amount of error provided by an error function, wherein the error function quantifies the deviation between a prediction of a representation in frequency space of the examination region of the subject using a third amount of a contrast agent administered during a medical imaging technique and a third reference representation in frequency space of the examination region of the subject.

[0050] Clause 38: The method of any of clauses 27 to 37, further comprising: training a predictive machine learning model based on a training dataset, wherein the training dataset comprises: a set of reduced referring representations in a frequency space of the test area for each test object of a plurality of test objects, each set of reduced referring representations in the frequency space of the test area of ​​the test object comprises: a first reduced referring representation in the frequency space of the test area of ​​the test object; a second reduced referring representation in the frequency space of the test object; and a third reduced referring representation in the frequency space of the test area of ​​the test object; and wherein the first reduced referring representation is a portion of the first referring representation in the frequency space of the test area that includes a center of the frequency space. wherein the first reference representation comprises a reference representation of the examination area with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; wherein the second reduced reference representation comprises a reference representation of a portion of the second reference representation in a frequency space of the examination area including a center of the frequency space, wherein the second reference representation comprises a reference representation of the examination area with a second amount of contrast agent administered during the medical imaging technique; and wherein the third reduced reference representation comprises a reference representation of a portion of the third reference representation in a frequency space of the examination area including a center of the frequency space, wherein the third reference representation comprises a reference representation of the examination area with a third amount of contrast agent administered during the medical imaging technique.

[0051] Clause 39: A method as described in clause 38, wherein training the predictive machine learning model comprises: minimizing an amount of error provided by an error function, wherein the error function quantifies the deviation between a prediction of a reduced representation in frequency space of the examination region of the subject using a third amount of a contrast agent administered during a medical imaging technique and a third reduced reference representation in frequency space of the examination region of the subject.

[0052] These and other features and characteristics of the present disclosure, as well as the function of elements relating to the method of operation and structure, and the combination of parts and economies of manufacture, will become more apparent from a consideration of the following description and the appended claims, taken in conjunction with the accompanying drawings, all of which form a part hereof, and in which like reference numerals indicate corresponding parts in the various views. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only and are not intended as a definition of the limits of the present disclosure. As used in this specification and claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Further advantages and details of the non-limiting embodiments are explained in more detail below with reference to exemplary embodiments shown in the accompanying schematic drawings: [Brief description of the drawings]

[0054] [Figure 1] FIG. 1 (FIG. 1) is a diagram of a non-limiting embodiment of an environment in which the systems, methods, and / or computer program products described herein may be implemented in accordance with the present disclosure; [Diagram 2] FIG. 2 is a diagram of a non-limiting embodiment of components of one or more devices and / or one or more systems of FIG. 1; [Diagram 3] FIG. 3 is a flow chart of a non-limiting embodiment of a process for providing a prediction of a representation of an examination region generated using a medical imaging technique that requires a contrast agent; [Figure 4] FIG. 4 is a flow chart of a non-limiting embodiment of a process for training a predictive machine learning model to provide a prediction of a representation of an examination region; [Figure 5A] 5A-5C are diagrams of non-limiting embodiments of implementations of a process for providing a prediction of a representation of an examination region generated using a medical imaging technique that requires a contrast agent; and [Figure 5B] 5A-5C are diagrams of non-limiting embodiments of implementations of a process for providing a prediction of a representation of an examination region generated using a medical imaging technique that requires a contrast agent; and [Figure 5C] 5A-5C are diagrams of non-limiting embodiments of implementations of a process for providing a prediction of a representation of an examination region generated using a medical imaging technique that requires a contrast agent; and [Figure 6A] 6A-6B are diagrams of a non-limiting embodiment of an implementation of a process for providing a prediction of a representation of an examination region generated using a medical imaging technique that requires a contrast agent. [Figure 6B] 6A-6B are diagrams of a non-limiting embodiment of an implementation of a process for providing a prediction of a representation of an examination region generated using a medical imaging technique that requires a contrast agent.

[0055] explanation

[0056] For purposes of the following description, the terms "end," "upper," "lower," "right," "left," "vertical," "horizontal," "top," "bottom," "lateral," "longitudinal," and derivatives thereof, shall refer to the present disclosure as oriented in the drawings. However, it should be understood that the present disclosure may contemplate various alternative modifications and step sequences unless expressly specified to the contrary. It should also be understood that the specific devices and processes illustrated in the accompanying drawings and described in the following specification are merely exemplary embodiments or aspects of the present disclosure. Thus, specific dimensions and other physical characteristics associated with the embodiments or aspects of the embodiments disclosed herein should not be considered as limiting, unless otherwise indicated.

[0057] As used herein, aspects, components, elements, structures, operations, steps, functions, instructions, and / or the like should not be construed as critical or essential unless expressly described as such. Also, as used herein, the articles "a" and "an" are intended to include one or more items and may be used interchangeably with "one or more" and "at least one." As used in this specification and claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Furthermore, as used herein, the terms "set" and "group" are intended to include one or more items (e.g., related items, unrelated items, combinations of related and unrelated items, etc.) and may be used interchangeably with "one or more" or "at least one." When only one item is intended, the term "a" or similar language is used. Also, as used herein, the terms "has," "have," "having," and the like are intended to be open-ended terms. Additionally, the phrase "based on" is intended to mean "based at least in part on," unless expressly stated otherwise. Additionally, the phrase "based on" means "in response to," and may indicate a condition for automatically triggering a specified operation of an electronic device (e.g., a controller, a processor, a computing device, etc.), as appropriately referred to herein.

[0058] As used herein, the term "system" may refer to one or more computing devices or combinations of computing devices, such as, but not limited to, a processor, a server, a client device, a software application, and / or other similar components. Additionally, references to a "server" or a "processor" as used herein may refer to a previously enumerated server and / or processor, a different server and / or processor, and / or a combination of servers and / or processors that are enumerated as performing a previous step or function. For example, as used herein and in the claims, a first server and / or a first processor enumerated as performing a first step or function may refer to the same or a different server and / or processor that is enumerated as performing a second step or function.

[0059] As used herein, the terms "communication" and "communicating" may refer to the reception, receipt, transmission, transfer, provision, etc. of information (e.g., data, signals, messages, instructions, commands, etc.). A unit (e.g., a device, a system, a component of a device or system, combinations thereof, and / or the like) communicating with another unit means that one unit can directly or indirectly receive information from the other unit and / or transmit information to the other unit. This may refer to a direct or indirect connection that is wired and / or wireless in nature. Furthermore, two units may communicate with each other even though the information transmitted may be modified, processed, relayed, and / or routed between the first and second units. For example, a first unit may communicate with a second unit even if the first unit passively receives information and does not actively transmit information to the second unit. As another example, a first unit can communicate with a second unit if at least one intermediate unit (e.g., a third unit located between the first unit and the second unit) processes information received from the first unit and transmits the processed information to the second unit. In some non-limiting embodiments, information may refer to network packets containing data (e.g., data packets, etc.).

[0060] Some non-limiting embodiments are described herein in relation to a threshold value. As used herein, meeting a threshold value may refer to a value that is greater than the threshold value, more than the threshold value, higher than the threshold value, greater than or equal to the threshold value, less than the threshold value, less than the threshold value, lower than the threshold value, less than or equal to the threshold value, equal to the threshold value, etc.

[0061] In some non-limiting embodiments, the machine learning model can be used as a method to reduce the amount of contrast agent used to generate radiological images. For example, during a first step of the process, a training data set can be generated. For a plurality of individuals (e.g., patients, people, etc.), the training data set can include, for each individual, a native radiological image (e.g., a radiological image generated without an amount of contrast agent administered during the medical imaging technique, such as a zero contrast image), a low contrast radiological image generated (e.g., later) during the medical imaging technique based on the administration of a low amount of contrast agent (e.g., a low contrast image), and a full contrast radiological image (e.g., a full contrast image) generated during the medical imaging technique based on the administration of a standard amount of contrast agent. In a second step, a machine learning model, such as an artificial neural network, can be trained to predict, for each individual included in the training data set, a predicted radiological image (e.g., an artificial radiological image) showing the examination area after the administration of a standard amount of contrast agent based on the native radiological image and the low contrast radiological image. The full contrast radiological image is used for each individual as a reference (e.g., ground truth) when training the predictive machine learning model. In a third step, for a new individual, the trained machine learning model can be used to predict a predicted radiographic image showing the examination area of ​​the new individual based on the native radiographic image and the low-contrast radiographic image, such that the examination area appears as if a standard amount of contrast agent had been administered during the medical imaging technique. A similar process can be found in International Patent Application PCT / US2018 / 055034, filed October 9, 2018, which is incorporated herein by reference in its entirety.

[0062] However, during such a process, coregistration may be required to match individual radiographic images of different individuals such that pixels / voxels correspond to one another. For example, coregistration of radiographic images may be required such that image elements (e.g., pixels or voxels) of a radiographic image of an examination region from an individual correspond to (e.g., show the same as) image elements of another radiographic image of an examination region from that individual. If the image elements of the radiographic images do not match, artifacts may appear in the predicted radiographic image, which may obscure, distort, and / or simulate anatomical structures in the examination region.

[0063] In addition, the above-mentioned process may require the use of complete radiographic images for training the predictive machine learning model and generating the predicted radiographic images. For example, training the predictive machine learning model and generating the predicted radiographic images may require the use of complete radiographic images after administration of various amounts of contrast agent. In this way, the computational complexity for generating the predicted radiographic images may increase rapidly. Furthermore, the computation of the predicted radiographic images may require a significant amount of time, along with expensive hardware that may be required to perform the computations within the desired time. Furthermore, the complete radiographic images may need to be reduced in size to sub-regions (e.g., patches), which may need to be processed separately from each other so as not to overload the computer's memory with excessively large radiographic images. However, such an approach may lead to artifacts at the interface when the sub-regions are processed separately from each other and recombined to form the complete radiographic image (e.g., stitching artifacts). Subsequent removal of such stitching artifacts may require additional computational resources. Furthermore, the approach may provide a risk of errors in the predicted radiographic images that may be misinterpreted by radiologists, which may provide a risk of misdiagnosis based on the complete radiographic images.

[0064] Improved systems, methods, and computer program products are provided for providing predictions of a representation of an examination region generated using a medical imaging technique requiring a contrast agent. Embodiments of the present disclosure may include a system comprising at least one processor programmed or configured to: receive a first representation in frequency space of an examination region of an object, where the first representation includes a representation of the examination region without an amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; receive a second representation in frequency space of an examination region of an object, where the second representation includes a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent is different from the first amount of contrast agent; and provide input to a predictive machine learning model; wherein the input to the predictive machine learning model includes at least a portion of the first representation and at least a portion of the second representation, and wherein the predictive machine learning model provides as an output a prediction of a representation in frequency space of the examination region using a third amount of contrast agent administered during the medical imaging technique, where the third amount of contrast agent is greater than the first amount of contrast agent and the second amount of contrast agent; comprising a trained machine learning model configured to receive an output of the predictive machine learning model based on the input, transform the output of the predictive machine learning model into a predicted representation in real space of the examination region of the subject, and provide the predicted representation of the examination region of the subject in real space.

[0065] In some non-limiting embodiments, the at least one processor is further programmed or configured to: identify a portion of the first representation comprising a center in frequency space, a portion of the second representation comprising a center in frequency space, or a portion of the first representation and a portion of the second representation, each comprising a center in frequency space, and provide a reduced representation in frequency space of the inspection area of ​​the subject, where the input to the predictive machine learning model includes the reduced representation in frequency space of the inspection area of ​​the subject, and where when the input is provided to the predictive machine learning model, the at least one processor is programmed or configured to provide the reduced representation in frequency space of the inspection area of ​​the subject as input to the predictive machine learning model. In some non-limiting embodiments, the at least one processor is further programmed or configured to supplement the output of the predictive machine learning model with a portion of the first representation that does not include the center of frequency space and was not selected to provide a reduced representation in frequency space of the inspection area of ​​the subject to be inspected, a portion of the second representation that does not include the center of frequency space and was not selected to provide a reduced representation in frequency space of the inspection area of ​​the subject to be inspected, or a portion of the first representation and a portion of the second representation, each of which does not include the center of frequency space and was not selected to provide a reduced representation in frequency space of the inspection area of ​​the subject to be inspected, to provide a supplemented output of the predictive machine learning model, wherein when converting the output of the predictive machine learning model to a predicted representation of the inspection area of ​​the subject to be inspected in real space, the at least one processor is programmed or configured to convert the supplemented output of the predictive machine learning model to a predicted representation in real space of the inspection area of ​​the subject to be inspected.

[0066] In some non-limiting embodiments, a first amount of contrast agent administered during a medical imaging technique is greater than zero, and wherein a second amount of contrast agent administered during the medical imaging technique is greater than the first amount of contrast agent.

[0067] In some non-limiting embodiments, a first amount of contrast agent administered during a medical imaging technique is zero, and wherein a second amount of contrast agent administered during the medical imaging technique is greater than the first amount of contrast agent (i.e., greater than zero).

[0068] In some non-limiting embodiments, the at least one processor is further programmed or configured to generate a first representation in frequency space of the examination region of the examined object for a first result of the radiological examination and to generate a second representation in frequency space of the examination region of the examined object for a second result of the radiological examination. In some non-limiting embodiments, the radiological examination is a magnetic resonance imaging examination, a computed tomography examination, or an ultrasound examination.

[0069] In some non-limiting embodiments, when receiving a first representation in frequency space of the examination region of the subject in frequency space, the at least one processor is programmed or configured to receive first k-space data associated with a magnetic resonance imaging (MRI) examination of the examination region of the subject, and wherein when receiving a second representation in frequency space of the examination region of the subject in frequency space, the at least one processor is programmed or configured to receive second k-space data associated with an MRI examination of the examination region of the subject.

[0070] In some non-limiting embodiments, when receiving a first representation in frequency space of an examination region of the subject in frequency space, the at least one processor is programmed or configured to: receive a first representation in real space of the examination region of the subject, where the first representation in real space includes a representation of the examination region without a first amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique; convert the first representation in real space of the examination region of the subject into a first representation in frequency space; and, when receiving a second representation in frequency space of the examination region of the subject in frequency space, the at least one processor is programmed or configured to: receive a second representation in real space of the examination region of the subject, where the second representation in real space includes a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, where the second amount of contrast agent differs from the first amount of contrast agent; convert the first representation in real space of the examination region of the subject into a first representation in frequency space.

[0071] In some non-limiting embodiments, the at least one processor is further programmed or configured to train the predictive machine learning model based on a training dataset, where the training dataset includes a set of reference representations in a frequency space of the examination area for each examination object of the multiple examination objects, where each set of reference representations of the examination area of ​​the examination object includes a first reference representation in the frequency space of the examination area of ​​the examination object, a second reference representation in the frequency space of the examination area of ​​the examination object, and a third reference representation in the frequency space of the examination area of ​​the examination object, where the first reference representation includes a reference representation of the examination area with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique, where the second reference representation includes a reference representation of the examination area with a second amount of contrast agent administered during the medical imaging technique, and where the third reference representation includes a reference representation of the examination area with a third amount of contrast agent administered during the medical imaging technique.

[0072] In some non-limiting embodiments, when training the predictive machine learning model, the at least one processor is programmed or configured to minimize an amount of error provided by an error function, where the error function quantifies the deviation between a prediction of a representation in frequency space of the examination region of the subject using a third amount of a contrast agent administered during the medical imaging technique and a third reference representation in frequency space of the examination region of the subject.

[0073] In some non-limiting embodiments, the at least one processor is further programmed or configured to train the predictive machine learning model based on a training dataset, where the training dataset includes a set of reduced referring representations in a frequency space of the test area for each test object of the plurality of test objects, where each set of reduced referring representations of the test area of ​​the test object includes a first reduced referring representation in the frequency space of the test area of ​​the test object, a second reduced referring representation in the frequency space of the test area of ​​the test object, and a third reduced referring representation in the frequency space of the test area of ​​the test object, where the first reduced referring representation is a reduced referring representation in the frequency space of the test area including a center of the frequency space. The reduced reference representation includes a reference representation of a portion of a first reference representation, where the first reference representation includes a reference representation of an examination area with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique, where the second reduced reference representation includes a reference representation of a portion of the second reference representation in a frequency space of the examination area including a center of the frequency space, where the second reference representation includes a reference representation of an examination area with a second amount of contrast agent administered during the medical imaging technique, and where the third reduced reference representation includes a reference representation of a portion of the third reference representation in a frequency space of the examination area including a center of the frequency space, where the third reference representation includes a reference representation of an examination area with a third amount of contrast agent administered during the medical imaging technique.

[0074] In some non-limiting embodiments, when training the predictive machine learning model, the at least one processor is programmed or configured to minimize an amount of error provided by an error function, where the error function quantifies the deviation between a prediction of a reduced representation in frequency space of the examination region of the subject using a third amount of a contrast agent administered during the medical imaging technique and a third reduced reference representation in frequency space of the examination region of the subject.

[0075] In this way, the system can provide a predicted representation in real space of the examination region of the subject, which reduces the use of contrast agent to provide a representation in real space of the examination region of the subject. Furthermore, the system can provide a predicted representation in real space of the examination region of the subject in a manner that is resistant to errors in co-registration, so that the risk of artifacts, in particular stitching artifacts, can be avoided. In addition, the system may require fewer computing resources to provide the predicted representation than a system that does not use a machine learning model configured to provide as output a prediction of a representation in frequency space of the examination region with a specific amount of contrast agent administered during a medical imaging technique, and the system may allow the computational complexity to be adapted to the hardware configuration of the system. Furthermore, by using a reduced representation in frequency space of the inspection area of ​​the inspected object to provide a predicted representation in real space and / or by using a reduced reference representation in frequency space of the inspection area of ​​the inspected object to train a machine learning model, the system may reduce the amount of time required to provide a predicted representation in real space and / or to train a machine learning model compared to systems that do not use a reduced representation in frequency space of the inspection area of ​​the inspected object to provide a predicted representation in real space or systems that do not use a reduced reference representation in frequency space of the inspection area of ​​the inspected object to train a machine learning model.

[0076]

[0023] Referring now to FIG. 1, FIG. 1 is a diagram of a non-limiting embodiment of an environment 100 in which the devices, systems, methods, and / or computer program products described herein may be implemented. As shown in FIG. 1, the environment 100 includes a radiology machine learning system 102, a medical imaging system 104, and a user device 106. In some non-limiting embodiments, the radiology machine learning system 102 may be interconnected (e.g., establish a connection to communicate, etc.) with the medical imaging system 104 and / or the user device 106 via a communication network 108. In some non-limiting embodiments, the radiology machine learning system 102 may be interconnected with the medical imaging system 104 and the user device 106 via a wired connection, a wireless connection, or a combination of wired and wireless connections.

[0077] The radiation machine learning system 102 may include one or more computing devices configured to communicate with the medical imaging system 104 and / or the user device 106 over a communications network 108. For example, the radiation machine learning system 102 may include a server, a group of servers, and / or other similar devices. In some non-limiting embodiments, the radiation machine learning system 102 may be a component of the medical imaging system 104. In some non-limiting embodiments, the radiation machine learning system 102 may include a cloud computing system.

[0078] The medical imaging system 104 may include one or more computing devices capable of communicating with the radiology machine learning system 102 and / or the user device 106 over a communications network 108. For example, the medical imaging system 104 may include one or more scanners, such as a CT scanner and / or an MRI scanner, that can communicate over the communications network 108 and that can perform medical imaging procedures that include the use of contrast agents (e.g., radiological contrast agents).

[0079] The user device 106 may include one or more computing devices configured to communicate with the radiology machine learning system 102 and / or the medical imaging system 104 over the communications network 108. For example, the user device 106 may include a desktop computer (e.g., a client device that communicates with a server) and / or a portable computer, such as a laptop, tablet, smartphone, etc. In some non-limiting embodiments or aspects, the user device 106 may be associated with a user (e.g., an individual that operates the device).

[0080] The communication network 108 may include one or more wired and / or wireless networks. For example, the communication network 108 may include a cellular network (e.g., a Long Term Evolution (LTE) network, a third generation (3G) network, a fourth generation (4G) network, a fifth generation (5G) network, a code division multiple access (CDMA) network, etc.), a local area network (LAN), a wide area network (WAN), a wireless LAN (WLAN), a private network, an ad-hoc network, an intranet, the Internet, an optical fiber-based network, an Ethernet network, a universal serial bus (USB) network, a cloud computing network, etc., and / or a combination of some or all of these or other types of networks.

[0081] The number and arrangement of systems and / or devices shown in Figure 1 are provided as an example. There may be additional systems and / or devices, fewer systems and / or devices, different systems and / or devices, or systems and / or devices arranged differently than those shown in Figure 1. Furthermore, two or more systems and / or devices shown in Figure 1 may be implemented within a single system or device, or a single system or device shown in Figure 1 may be implemented as multiple distributed systems or devices. Additionally or alternatively, a set of systems or devices (e.g., one or more systems, one or more devices) of environment 100 may perform one or more functions described as being performed by another set of systems or another set of devices of environment 100.

[0082] Reference is now made to FIG. 2, which is a diagram of example components of a device 200. The device 200 may correspond to a radiology machine learning system 102, a medical imaging system 104, and / or a user device 106. In some non-limiting embodiments, the radiology machine learning system 102, the medical imaging system 104, and / or the user device 106 may include at least one device 200 and / or at least one component of the device 200. As shown in FIG. 2, the device 200 may include a bus 202, a processor 204, a memory 206, a storage component 208, an input component 210, an output component 212, and a communication interface 214.

[0083] The bus 202 may include components that enable communication between the components of the device 200. In some non-limiting embodiments, the processor 204 may be implemented in hardware, software, or a combination of hardware and software. For example, the processor 204 may include a processor (e.g., a central processing unit (CPU), a graphics processing unit (GPU), an accelerated processing unit (APU), etc.), a microprocessor, a digital signal processor (DSP), and / or any processing component that can be programmed to perform a function (e.g., a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), etc.). The memory 206 may include a random access memory (RAM), a read only memory (ROM), and / or another type of dynamic or static storage device (e.g., flash memory, magnetic memory, optical memory, etc.) that stores information and / or instructions for use by the processor 204.

[0084] The storage component 208 may store information and / or software related to the operation and use of the device 200. For example, the storage component 208 may include a hard disk (e.g., a magnetic disk, optical disk, magneto-optical disk, solid state disk, etc.), a compact disk (CD), a digital versatile disk (DVD), a floppy disk, a cartridge, a magnetic tape, and / or another type of computer-readable medium along with a corresponding drive.

[0085] Input components 210 may include components that enable device 200 to receive information, such as via user input (e.g., a touchscreen display, a keyboard, a keypad, a mouse, buttons, switches, a microphone, a camera, etc.). Additionally or alternatively, input components 210 may include sensors for sensing information (e.g., a global positioning system (GPS) component, an accelerometer, a gyroscope, an actuator, etc.). Output components 212 may include components that provide output information from device 200 (e.g., a display, a speaker, one or more light emitting diodes (LEDs), etc.).

[0086] The communication interface 214 may include transceiver-like components (e.g., a transceiver, a separate receiver and transmitter, etc.) that enable the device 200 to communicate with other devices, such as via a wired connection, a wireless connection, or a combination of wired and wireless connections. The communication interface 214 may enable the device 200 to receive information from another device and / or provide information to another device. For example, the communication interface 214 may include an Ethernet interface, an optical interface, a coaxial interface, an infrared interface, a radio frequency (RF) interface, a universal serial bus (USB) interface, a Wi-Fi interface, a cellular network interface, etc.

[0087] The device 200 can perform one or more processes described herein. The device 200 may perform these processes based on the processor 204 executing software instructions stored in a computer-readable medium, such as the memory 206 and / or the storage component 208. A computer-readable medium (e.g., a non-transitory computer-readable medium) is defined herein as a non-transitory memory device. A non-transitory memory device may include a memory space located within a single physical storage device or a memory space spread across multiple physical storage devices.

[0088] Software instructions may be loaded into memory 206 and / or storage component 208 from another computer-readable medium or from another device via communication interface 214. When executed, the software instructions stored in memory 206 and / or storage component 208 may cause processor 204 to perform one or more processes described herein. Additionally or alternatively, hardwired circuitry may be used in place of or in combination with software instructions to perform one or more processes described herein. Thus, the embodiments or aspects described herein are not limited to any specific combination of hardware circuitry and software.

[0089] The memory 206 and / or the storage component 208 may include a data storage or one or more data structures (e.g., a database and / or the like). The device 200 may be capable of retrieving information from, storing information in, or retrieving information stored in the data storage or one or more data structures in the memory 206 and / or the storage component 208.

[0090] The number and arrangement of components shown in Figure 2 are provided as an example. In some non-limiting embodiments, device 200 may include additional, fewer, different, or differently arranged components than those shown in Figure 2. Additionally or alternatively, a set of components (e.g., one or more components) of device 200 may perform one or more functions described herein as being performed by another set of components of device 200.

[0091] Reference is now made to FIG. 3, which is a flow chart of a non-limiting embodiment of a process 300 for providing a prediction of a representation of an examination region generated using a medical imaging technique that requires a contrast agent. In some non-limiting embodiments or aspects, one or more of the functions described with respect to the process 300 may be performed (e.g., fully, partially, etc.) by the radiology machine learning system 102. In some non-limiting embodiments or aspects, one or more of the steps of the process 300 may be performed (e.g., fully, partially, and / or similarly) by another device or group of devices separate from and / or including the radiology machine learning system 102, such as the medical imaging system 104 and / or the user device 106.

[0092] As shown in FIG. 3, in step 302, the process 300 can include receiving a first representation in frequency space of an examination region. For example, the radiology machine learning system 102 may receive a first representation in frequency space of an examination region (e.g., a field of view (FOV), an image volume, etc.) in frequency space from the medical imaging system 104. In some non-limiting embodiments, the radiology machine learning system 102 can receive a first representation in frequency space of an examination region of an examination object from the medical imaging system 104. In some non-limiting embodiments, the first representation can include a representation in frequency space of the examination region with no amount of contrast agent administered during a medical imaging technique (e.g., a medical imaging technique that will be performed during a radiology examination) or with a first amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the first amount of contrast agent administered during the medical imaging technique is greater than zero.

[0093] In some non-limiting embodiments, the examination object may include a living organism such as a mammal. In one example, the examination object may include a human. In some non-limiting embodiments, the examination area may be a portion of the examination object undergoing a radiological examination (e.g., a radiological examination including medical imaging techniques). In one example, the examination area may include an organ or a portion of an organ of the examination object. In some non-limiting embodiments, the radiological examination may include a magnetic resonance imaging (MRI) examination, a computed tomography (CT) examination, and / or an ultrasound examination. In some non-limiting embodiments, the examination area may include a volume imaged in a radiological image. The examination area may be defined by a radiologist (e.g., manually by a radiologist), for example on an overview image (localizer). Additionally or alternatively, the examination area may be defined automatically (e.g., automatically by the radiological machine learning system 102, automatically by the medical imaging system 104, etc.). For example, the examination area may be automatically defined based on a protocol (e.g., a particular protocol of the medical imaging system 104).

[0094] In some non-limiting embodiments, the first representation in frequency space of the examination region may include data related to a radiological examination performed by the medical imaging system 104. For example, the first representation in frequency space of the examination region may include k-space data related to an MRI examination (e.g., an MRI examination of the examination region of the examined subject) performed by the medical imaging system 104.

[0095] In some non-limiting embodiments, the radiation machine learning system 102 can transform the first representation in frequency space of the examination region to a first representation in real space (e.g., image space) of the examination region. For example, the radiation machine learning system 102 can use an inverse Fourier transform to transform the first representation in frequency space of the examination region to the first representation in real space of the examination region.

[0096] In some non-limiting embodiments, the radiation machine learning system 102 can generate a first representation in frequency space of the examination region. For example, the radiation machine learning system 102 can receive a first representation in real space of the examination region (e.g., a first radiographic image of the examination region), and the radiation machine learning system 102 can generate a first representation in frequency space of the examination region based on the first representation in real space of the examination region using a Fourier transform. In some non-limiting embodiments, the radiation machine learning system 102 can receive the first representation in real space of the examination region as a two-dimensional (2D) radiographic image of the examination region in real space or as a three-dimensional radiographic image, and the radiation machine learning system 102 can generate a first representation in frequency space of the examination region based on the first representation in real space of the examination region using a 2D Fourier transform or a 3D Fourier transform, as appropriate.

[0097] As shown in FIG. 3, in step 304, the process 300 can include receiving a second representation in frequency space of the examination region. For example, the radiation machine learning system 102 can receive a second representation in frequency space of the examination region (e.g., the same examination region of the first representation) from the medical imaging system 104. In some non-limiting embodiments, the radiation machine learning system 102 can receive a second representation in frequency space of the examination region of the examination object (e.g., the same examination object as the examination region of the first representation) from the medical imaging system 104. In some non-limiting embodiments, the second representation can include a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the second amount of contrast agent is different from the first amount of contrast agent administered during the medical imaging technique associated with the first representation. In some non-limiting embodiments, the second amount of contrast agent is greater than the first amount of contrast agent administered during the medical imaging technique.

[0098] In some non-limiting embodiments, the second representation in frequency space of the examination region may include data associated with a radiological examination performed by the medical imaging system 104. For example, the second representation in frequency space of the examination region may include k-space data associated with an MRI examination performed by the medical imaging system 104.

[0099] In some non-limiting embodiments, the radiation machine learning system 102 can receive k-space data associated with a first representation in frequency space of the examination region and k-space data associated with a second representation in frequency space of the examination region. For example, the radiation machine learning system 102 can receive the first representation in frequency space of the examination region as first k-space data associated with an MRI examination of the examination region (e.g., first k-space data generated by the MRI examination), and the radiation machine learning system 102 can receive the second representation in frequency space of the examination region as second k-space data associated with an MRI examination of the examination region (e.g., second k-space data generated by the MRI examination). In some non-limiting embodiments, the radiation machine learning system 102 can generate the first representation in frequency space of the examination region for a first result of the radiation examination, and the radiation machine learning system 102 can generate the second representation in frequency space of the examination region for a second result of the radiation examination. For example, the radiation machine learning system 102 can generate a first representation in frequency space of the examination region based on k-space data associated with a first result of the MRI examination, and the radiation machine learning system 102 can generate a second representation in frequency space of the examination region based on k-space data associated with a second result of the MRI examination. In some non-limiting embodiments, the first representation in frequency space of the examination region can include first k-space data associated with the MRI examination of the examination region and / or the second representation in frequency space of the examination region can include second k-space data associated with the MRI examination of the examination region.

[0100] In some non-limiting embodiments, the radiation machine learning system 102 can transform the second representation in frequency space of the examination region to a second representation in real space of the examination region. For example, the radiation machine learning system 102 can use an inverse Fourier transform to transform the second representation in frequency space of the examination region to a second representation in real space of the examination region.

[0101] In some non-limiting embodiments, the radiation machine learning system 102 can generate a second representation in frequency space of the examination region. For example, the radiation machine learning system 102 can receive a second representation in real space of the examination region (e.g., a first radiological image of the examination region), and the radiation machine learning system 102 can generate a second representation in frequency space of the examination region based on the second representation in real space of the examination region using a Fourier transform. In some non-limiting embodiments, the radiation machine learning system 102 can receive the second representation in real space of the examination region as a two-dimensional (2D) radiological image or a three-dimensional (3D) radiological image of the examination region, and the radiation machine learning system 102 can generate a second representation in frequency space of the examination region based on the second representation in real space of the examination region using a 2D Fourier transform or a 3D Fourier transform, as appropriate.

[0102] As shown in FIG. 3, in step 306, the process 300 may include providing the first and second representations as inputs to a predictive machine learning model. For example, the radiology machine learning system 102 may provide inputs to a predictive machine learning model (e.g., a prediction algorithm, a prediction model, a predictive model, etc.), where the inputs to the predictive machine learning model include at least a portion of the first representation in frequency space of the examination region and at least a portion of the second representation in frequency space of the examination region. In some non-limiting embodiments, the predictive machine learning model may include a trained machine learning model configured to provide as an output a prediction of a representation in frequency space of the examination region of the examined subject using a third amount of a contrast agent administered during a medical imaging technique (e.g., the same medical imaging technique as the medical imaging technique associated with the first representation in frequency space of the examination region and the second representation in frequency space of the examination region). In some non-limiting embodiments, predicting a representation in frequency space of an examination region using a third amount of contrast agent administered during a medical imaging technique includes predicting an examination region of a subject such that the examination region would be represented if a specific amount of contrast agent (e.g., a standard amount of contrast agent that would be administered to a subject based on parameters of the medical imaging technique and / or the subject) were administered to the examination region (e.g., an examination region of a subject) where a specific amount does not actually need to be administered during a medical imaging technique.

[0103] In some non-limiting embodiments, the third amount of contrast agent is different from the first amount of contrast agent and / or the second amount of contrast agent, for example, the third amount of contrast agent may be greater than the first amount of contrast agent administered during a medical imaging technique associated with a first representation in frequency space of the examination region and may be greater than the second amount of contrast agent administered during a medical imaging technique associated with a second representation in frequency space of the examination region.

[0104] In some non-limiting embodiments, the predictive machine learning model may include an artificial neural network. In some non-limiting embodiments, the artificial neural network may include at least three layers of processing elements. For example, the artificial neural network may include a first layer having input neurons (e.g., nodes), an Nth layer having output neurons, and N-2 hidden layers, where N is a natural number and is greater than 2. In some non-limiting embodiments, the input neurons may be configured to receive a first representation in frequency space of the examination region of the examined subject and a second representation in frequency space of the examination region of the examined subject. The output neurons may be configured to provide (e.g., as an output) at least one prediction of the representation in frequency space of the examination region with a particular amount of contrast agent administered during the medical imaging technique based on the input. In some non-limiting embodiments, the processing elements of the layers between the input neurons and the output neurons are connected to each other in a predetermined pattern with predetermined connection weights. In some non-limiting embodiments, the connection weights may be updated during training of the machine learning model (e.g., training of the artificial neural network).

[0105] In some non-limiting embodiments, the predictive machine learning model may include a convolutional neural network (CNN). In some non-limiting embodiments, the CNN may be capable of processing inputs in the form of a matrix. The CNN may include one or more filters (e.g., one or more convolutional layers) and one or more aggregation layers (e.g., one or more pooling layers), which may be alternated, and the end of the CNN may include one or more layers of fully connected neurons (e.g., one or more dense / fully connected layers). Additionally or alternatively, the predictive machine learning model may include a generative adversarial network (GAN).

[0106] In some non-limiting embodiments, the input to the predictive machine learning model may further include information related to the examination object (e.g., the examination object for which the examination area is obtained based on the radiological examination), information related to the examination area (e.g., the examination area related to the first and second representations), and / or information related to the radiological examination performed on the examination object (e.g., the examination area of ​​the examination object). For example, the input to the predictive machine learning model may include information related to the examination object's gender, age, weight, height, medical history, nature and duration and amount of medications already taken, blood pressure, central venous pressure, respiration rate, serum albumin, total bilirubin, blood glucose, iron content, respiratory capacity, etc. Additionally or alternatively, the input to the predictive machine learning model may include existing conditions of the examination area, medical procedures related to the examination area such as partial resection of the examination area, organ conditions such as whether a liver transplant has been performed, whether an iron liver is present, whether a fatty liver is present, etc. Additionally or alternatively, the input to the predictive machine learning model may include the type of radiological examination performed on the examination object.

[0107] In some non-limiting embodiments, the radiation machine learning system 102 can identify (e.g., select, choose, determine, etc.) a portion (e.g., a region) of a first representation in frequency space of the inspection region that includes a center of the frequency space, a portion (e.g., a region) of a second representation in frequency space of the inspection region that includes a center of the frequency space, or a portion (e.g., a region) of the first representation in frequency space of the inspection region that includes a center of the frequency space, or a portion (e.g., a region) of the ... to provide a reduced representation in frequency space of the inspection region of the inspection region. In some non-limiting embodiments, the radiology machine learning system 102 may identify a portion of a representation in frequency space of the inspection area (e.g., a portion of a first representation in frequency space of the inspection area, a portion of a second representation in frequency space of the inspection area, a portion of a reduced representation in frequency space of the inspection area, a portion of a reference representation in frequency space of the inspection area, etc.) based on manual input received from a user (e.g., a user of the user device 106).

[0108] In one example, the reduced representation in frequency space of the inspection area of ​​the subject may include a portion of the first representation in frequency space of the inspection area that includes a center of the frequency space of the inspection area and a second representation in frequency space of the inspection area. In another example, the reduced representation in frequency space of the inspection area of ​​the subject may include a first representation in frequency space of the inspection area and a portion of the second representation in frequency space of the inspection area that includes a center of the frequency space. In yet another example, the reduced representation in frequency space of the inspection area of ​​the subject may include a portion of the first representation in frequency space of the inspection area that includes a center of the frequency space and a portion of the second representation in frequency space of the inspection area that includes a center of the frequency space.

[0109] In some non-limiting embodiments, the input to the predictive machine learning model includes a reduced representation in frequency space of the inspection region of the inspected object. In some non-limiting embodiments, the radiation machine learning system 102 can provide the reduced representation in frequency space of the inspection region of the inspected object as an input to the predictive machine learning model. By using the reduced representation in frequency space of the inspection region of the inspected object, the radiation machine learning system 102 can reduce the computing resource requirements and the computational complexity used in obtaining the output of the predictive machine learning model based on the reduced representation in frequency space of the inspection region of the inspected object.

[0110] According to the present disclosure, the use of a representation in frequency space of the inspection region may be advantageous over the use of a representation in real space of the inspection region. For example, coregistration of the representations is less important in frequency space than in real space. Coregistration (e.g., image registration) may refer to a process in digital image processing to optimally match (e.g., align) two or more images of the same or similar region. One image may be defined as a reference image and the other image may be defined as a target image. The reference image and the target image to be matched may differ from each other because the reference image and the target image are acquired from different positions, at different times, and / or with different sensors. A compensating transformation may be calculated to optimally match the target image with the corresponding reference image.

[0111] A representation in frequency space of the examination region may be more tolerant of errors in co-registration during training, validation, and prediction procedures than a representation in real space of the examination region. For example, if a representation in frequency space is superimposed with a lower precision with another representation in frequency, the lack of precision has less impact than if a representation in real space of the examination region is superimposed with a lower precision. The reduced impact is due to the properties of the Fourier transform. For example, turns or rotations of a representation (e.g., an image) in real space may lead to image information (e.g., information related to visible structures) being localized in different regions of the image after the image is transformed using the Fourier transform. However, in frequency space, such turns or rotations do not change the region in which the contrast information is provided, since the contrast information of a Fourier transformed image (e.g., a representation of an image in frequency space) is mapped around the center of the frequency space.

[0112] A synonym for the term "center of frequency space" is the term "origin of frequency space". The center of frequency space corresponds, for example, to the point with coordinates 0,0 in a two-dimensional Cartesian coordinate system. Points closer to the center of frequency space represent lower frequencies than points further away from the center. The region in frequency space around the center contains more information regarding the contrast of the representation of the examined region than regions further away from the center.

[0113] A further advantage of the use of a representation in frequency space of the inspection area may be that contrast information is separated from detailed information (e.g. fine structure). It is therefore possible to focus during the training procedure on the contrast information to be learned by the predictive machine learning model, and during the prediction procedure also on the contrast information that can be predicted by the predictive machine learning model. Whereas the contrast information in the representation in real space of the inspection area is usually distributed throughout the representation (e.g. each image element inherently has information on contrast), the contrast information in the representation in frequency space of the inspection area is coded in the center and around the frequency space. Thus, the low frequencies in the representation in frequency space of the inspection area are responsible for the contrast information, while the high frequencies contain detailed information on the fine structure. The use of a representation in frequency space of the inspection area may make it possible to separate the contrast information, to restrict training and prediction to the contrast information, and to reintroduce information on the fine structure after the training procedure and / or the prediction procedure.

[0114] As shown in FIG. 3, in step 308, the process 300 can include determining an output of the predictive machine learning model. For example, the radiation machine learning system 102 can provide an input to the predictive machine learning model, and the predictive machine learning model can generate an output based on the input. The radiation machine learning system 102 can determine the output based on the predictive machine learning model that generates the output. In some non-limiting embodiments, the radiation machine learning system 102 can receive an output based on the predictive machine learning model that generates the output. In some non-limiting embodiments, the output of the predictive machine learning model can include a third representation in frequency space of the examination region, which can be a prediction of the representation in frequency space of the examination region (e.g., the artificial representation). In some non-limiting embodiments, the third representation can include a representation in frequency space of the examination region using a third amount of a contrast agent administered during the medical imaging technique.

[0115] In some non-limiting embodiments, the radiation machine learning system 102 can determine an output of a predictive machine learning model based on k-space data associated with an MRI study of the examination region. For example, the radiation machine learning system 102 can provide first k-space data associated with the MRI study of the examination region and second k-space data associated with the MRI study of the examination region as inputs to the predictive machine learning model. The radiation machine learning system 102 can determine the output based on a predictive machine learning model that uses the first k-space data associated with the MRI study of the examination region and the second k-space data associated with the MRI study of the examination region as inputs to generate the output.

[0116] As shown in FIG. 3, in step 310, the process 300 can include converting the output of the predictive machine learning model into a predicted representation in real space of the examination region. For example, the radiology machine learning system 102 can convert the output of the predictive machine learning model into a predicted representation in real space of the examination region (e.g., a prediction of the representation in real space of the examination region). In some non-limiting embodiments, the predicted representation in real space of the examination region can include a predicted radiological image of the examination region (e.g., a prediction of the radiological image of the examination region) using a third amount of contrast agent administered during a medical imaging technique (e.g., a medical imaging technique associated with a first representation in frequency space of the examination region and a second representation in frequency space of the examination region). In some non-limiting embodiments, the radiology machine learning system 102 can convert the output of the predictive machine learning model into a predicted representation in real space of the examination region using an inverse Fourier transform.

[0117] In some non-limiting embodiments, the radiation machine learning system 102 can perform operations related to the predicted representation in real space of the inspection area. For example, the radiation machine learning system 102 can provide the predicted representation in real space of the inspection area as an output of the radiation machine learning system 102. In such an example, the radiation machine learning system 102 can provide (e.g., transmit) the predicted representation in real space of the inspection area to the user device 106 and / or the medical imaging system 104 (e.g., such that the predicted representation in real space of the inspection area can be output by the user device 106 and / or the medical imaging system 104). In some non-limiting embodiments, the radiation machine learning system 102 can store the predicted representation in real space of the inspection area in a data structure associated with the radiation machine learning system 102. In some non-limiting embodiments, the radiation machine learning system 102 can display the predicted representation in real space of the inspection area on a display device.

[0118] In some non-limiting embodiments, the radiation machine learning system 102 can supplement the output of the predictive machine learning model to provide a supplemented output of the predictive machine learning model. For example, the radiation machine learning system 102 may supplement the output of the predictive machine learning model with a portion of a first representation in frequency space of the inspection area of ​​the inspected object that does not include a center of frequency space and is not specified to provide a reduced representation in frequency space of the inspection area of ​​the inspected object, a portion of a second representation in frequency space of the inspection area of ​​the inspected object that does not include a center of frequency space and is not specified to provide a reduced representation in frequency space of the inspection area of ​​the inspected object, or a portion of the first representation and a portion of the second representation, each of which does not include a center of frequency space and is not specified to provide a reduced representation in frequency space of the inspection area of ​​the inspected object, to provide a supplemented output of the predictive machine learning model. In some non-limiting embodiments, the radiation machine learning system 102 can convert the supplemented output of the predictive machine learning model to a representation in real space of the inspection area of ​​the inspected object.

[0119] In some non-limiting embodiments, the radiology machine learning system 102 can convert the resulting signal intensities for various position coordinates of the predicted representation in real space of the examination area into grayscale and / or color values ​​to provide a digital image in a common image format (e.g., Digital Imaging and Communications in Medicine (DICOM) format).

[0120] Referring now to FIG. 4, FIG. 4 is a flowchart of a non-limiting embodiment of a process 400 for training a predictive machine learning model configured to provide a prediction of a representation of an examination region with a particular amount of contrast agent administered during a medical imaging technique. In some non-limiting embodiments, the predictive machine learning model may be the same as or similar to the predictive machine learning model as described with respect to process 300. In some non-limiting embodiments or aspects, one or more of the functions described with respect to process 400 may be performed (e.g., fully, partially, etc.) by the radiology machine learning system 102. In some non-limiting embodiments or aspects, one or more of the steps of process 300 may be performed (e.g., fully, partially, and / or similarly) by another device or group of devices separate from and / or including the radiology machine learning system 102, such as the medical imaging system 104 and / or the user device 106.

[0121] As shown in FIG. 4, in step 402, the process 400 may include receiving a training data set including a set of reference representations in a frequency space of an examination region for each examination target of the multiple examination targets. In some non-limiting embodiments, the radiation machine learning system 102 may receive a training data set including a set of reference representations (e.g., multiple reference representations) in a frequency space of an examination region for each examination target of the multiple examination targets from the medical imaging system 104 and / or the user device 106. In some non-limiting embodiments, the radiation machine learning system 102 may obtain the training data set from a data structure associated with the radiation machine learning system 102. In some non-limiting embodiments, the examination region may be the same for all examination targets of the multiple examination targets. In some non-limiting embodiments, the reference representation (e.g., a reduced reference representation) of the examination region may refer to a representation in a frequency space of the examination region used during the training process of the predictive machine learning model.

[0122] In some non-limiting embodiments, each reference representation in the frequency space of the examination area of ​​each examination object may further include information related to the examination object (e.g., the examination object for which the examination area is obtained based on a radiological examination), information related to the examination area (e.g., the examination area related to the first and second representations), and / or information related to the radiological examination performed on the examination object (e.g., the examination area of ​​the examination object). For example, the information related to the examination object may include information related to the sex, age, weight, height, medical history, nature and / or duration / or amount of medications already taken, blood pressure, central venous pressure, respiration rate, serum albumin, total bilirubin, blood glucose, iron content, respiratory capacity, etc. of the examination object. Additionally or alternatively, the information related to the examination area may include existing conditions of the examination area, medical procedures related to the examination area such as partial resection of the examination area, organ conditions such as whether a liver transplant has been performed, whether an iron liver is present, whether a fatty liver is present, etc. Additionally or alternatively, the information related to the radiological examination performed on the examination object may include the type of radiological examination performed on the examination object.

[0123] In some non-limiting embodiments, the training data set may include a set (e.g., a plurality) of reference representations in the frequency space of the examination area for each examination object of the plurality of examination objects. Each set of reference representations in the frequency space of the examination area of ​​the examination object may include a first reference representation in the frequency space of the examination area of ​​the examination object, a second reference representation in the frequency space of the examination area of ​​the examination object, and a third reference representation in the frequency space of the examination area of ​​the examination object. In some non-limiting embodiments, the first reference representation may include a reference representation in the frequency space of the examination area with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the second reference representation may include a reference representation in the frequency space of the examination area with a second amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the third reference representation may include a reference representation in the frequency space of the examination area with a third amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the third amount of contrast agent may be a standard amount of contrast agent that would be administered to a subject during a medical imaging technique based on parameters of the medical imaging technique and / or the subject. In some non-limiting embodiments, the third reference representation may be a ground truth representation.

[0124] In some non-limiting embodiments, the first amount of contrast agent administered during the medical imaging technique is greater than zero. In some non-limiting embodiments, the first amount of contrast agent administered during the medical imaging technique is zero. In some non-limiting embodiments, the second amount of contrast agent is different from the first amount of contrast agent administered during the medical imaging technique associated with the first representation. In some non-limiting embodiments, the second amount of contrast agent is greater than the first amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the third amount of contrast agent is different from the first amount of contrast agent and / or the second amount of contrast agent. For example, the third amount of contrast agent may be greater than the first amount of contrast agent administered during the medical imaging technique associated with the first representation in frequency space of the examination region and greater than the second amount of contrast agent administered during the medical imaging technique associated with the second representation in frequency space of the examination region.

[0125] In some non-limiting embodiments, the training data set may include a set of reduced reference expressions in the frequency space of the test area for each test object of the multiple test objects. Each set of reduced reference expressions in the frequency space of the test area of ​​the test object may include a first reduced reference expression in the frequency space of the test area of ​​the test object, a second reduced reference expression in the frequency space of the test area of ​​the test object, and / or a third reduced reference expression in the frequency space of the test area of ​​the test object. In some non-limiting embodiments, the first reduced reference expression may include a representation in frequency space of a portion of the first reference expression in the frequency space of the test area including a center of the frequency space. In some non-limiting embodiments, the second reduced reference expression may include a representation in frequency space of a portion of the second reference expression in the frequency space of the test area including a center of the frequency space. In some non-limiting embodiments, the third reduced reference expression may include a representation in frequency space of a portion of the third reference expression in the frequency space of the test area including a center of the frequency space.

[0126] 4, in step 404, the process 400 may include identifying a portion of the referring representation. For example, the radiology machine learning system 102 may identify a portion (e.g., a region) of the referring representation (e.g., the same portion of each referring representation) of the examination region of the subject to be used to provide a reduced referring representation in the frequency space of the examination region of the subject. This allows the radiology machine learning system 102 to provide a training data set including reduced referring representations in the frequency space of the examination region of the subject of a plurality of examination regions.

[0127] In some non-limiting embodiments, the radiation machine learning system 102 may identify (e.g., automatically identify) a portion (e.g., an identified portion) of each referring expression (e.g., each referring expression of a set of referring expressions) in the frequency space of the examination region of the examination target of the plurality of examination targets to provide a reduced referring expression in the frequency space of the examination region of each examination target. For example, the radiation machine learning system 102 may identify a portion of the referring expression in the frequency space of the examination region of the set of referring expressions in the frequency space of the examination region such that the portion includes a center of the frequency space. In some non-limiting embodiments, the center of the portion may correspond to a center of the frequency space. In some non-limiting embodiments, the portion may have a shape. For example, the identified portion may be circular, angular, concave, and / or convex. In some non-limiting embodiments, the portion may be a cubic shape (e.g., a rectangular parallelepiped) in the case of a 3D frequency space in a Cartesian coordinate system. In some non-limiting embodiments, the portion may be a rectangular shape (e.g., a square) in the case of a 2D frequency space in a Cartesian coordinate system. In some non-limiting embodiments, the identified portion may have the same dimensions as the frequency space. For example, in the case of a 2D representation in a 2D frequency space, a portion may comprise a plane. In another example, in the case of a 3D representation in a 3D frequency space, a portion may comprise a volume.

[0128] As shown in FIG. 4, in step 406, the process 400 may include shrinking the reference representation. For example, the radiation machine learning system 102 may shrink the reference representation in the frequency space of the examination region of the examination object based on the identified portion (e.g., the identified portion) to provide a reduced reference representation in the frequency space of the examination region of the examination object. In some non-limiting embodiments, the radiation machine learning system 102 may shrink each reference representation by removing (e.g., discarding, cutting, etc.) any portion of the reference representation that is not present in the portion of the reference representation. For example, the radiation machine learning system 102 may shrink each reference representation by covering the reference representation with a mask (e.g., a mask corresponding to the identified portion) and removing any portion of the reference representation that is not covered by the mask. In some non-limiting embodiments, when covering the reference representation with the mask, the radiation machine learning system 102 may set one or more color values ​​of image elements (e.g., pixels, voxels) that are not covered by the mask to zero (e.g., black).

[0129] As shown in FIG. 4, in step 408, the process 400 can include training a predictive machine learning model using the multiple reduced referring representations. For example, the radiation machine learning system 102 can train a predictive machine learning model based on the multiple reduced referring representations in a frequency space of an examination area (e.g., an examination area of ​​an examination object). In some non-limiting embodiments, the radiation machine learning system 102 can train the predictive machine learning model using a self-learning algorithm in a supervised machine learning process. In some non-limiting embodiments, the radiation machine learning system 102 can use a self-learning algorithm to generate a statistical model based on a training dataset during machine learning.

[0130] In some non-limiting embodiments, the radiology machine learning system 102 can minimize the amount of error provided by the error function during training. In some non-limiting embodiments, the error function can quantify the deviation between a prediction (e.g., predicted representation) of a representation of the examination region of the subject using a particular amount of contrast agent administered during the medical imaging technique and a ground truth reference representation in frequency space of the examination region of the subject using a particular amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the error function quantifies the deviation between a prediction (e.g., predicted reduced representation) of a downscaled representation of the examination region of the subject using a particular amount of contrast agent administered during the medical imaging technique and a ground truth downscaled reference representation in frequency space of the examination region of the subject.

[0131] In some non-limiting embodiments, the predictive machine learning model may be configured to provide as output a prediction of the representation in frequency space of the examination region using a specific amount of contrast agent administered during a medical imaging technique (e.g., the same medical imaging technique as the medical imaging technique associated with the first representation in frequency space of the examination region and the second representation in frequency space of the examination region). In some non-limiting embodiments, the specific amount of contrast agent is different from the first amount of contrast agent administered during the medical imaging technique associated with the first representation in frequency space of the examination region and different from the second amount of contrast agent administered during the medical imaging technique associated with the second representation in frequency space of the examination region. For example, the specific amount of contrast agent is greater than the first amount of contrast agent administered during the medical imaging technique associated with the first representation in frequency space of the examination region and greater than the second amount of contrast agent administered during the medical imaging technique associated with the second representation in frequency space of the examination region.

[0132] In some non-limiting embodiments, the radiology machine learning system 102 can train a predictive machine learning model based on k-space data. For example, each reduced reference representation in frequency space of an examination region of an examination subject can include k-space data associated with an MRI examination (e.g., an MRI examination performed by the medical imaging system 104). The radiology machine learning system 102 may train a predictive machine learning model based on the k-space data included in each reduced reference representation in frequency space of an examination region of an examination subject for a plurality of examination subjects.

[0133] In some non-limiting embodiments, the radiation machine learning system 102 can train the predictive machine learning model using a backpropagation method. In this way, the radiation machine learning system 102 can train the predictive machine learning model to have maximum confidence in the mapping of inputs to corresponding outputs. The mapping quality can be described by an error function (e.g., a loss function). During the backpropagation method, the goal is to minimize the error function. For the backpropagation method, the predictive machine learning model can be updated (e.g., learned, taught, etc.) by changing the connection weights of the nodes. In the trained state, the connection weights between the processing elements of the predictive machine learning model can include information about the relationship between a first reference representation in the frequency space of the inspection region and a second reference representation in the frequency space of the inspection region, and information about a third reference representation. The information can be used to predict a representation in the frequency space of the inspection based on the first and second representations. In some non-limiting embodiments, the radiation machine learning system 102 can use a cross-validation method to divide the data into a training data set and a validation data set. The training data set may be used in backpropagation training of the connection weights of the nodes, and the validation data set may be used to check the accuracy of predictions of the representations provided by the trained predictive machine learning model based on data not used during training.

[0134] In some non-limiting embodiments, the radiology machine learning system 102 can generate a predictive machine learning model based on a training dataset. For example, the radiology machine learning system 102 can generate a predictive machine learning model to perform predictions of a representation (e.g., a representation in frequency space) of an examination region using a particular amount of a contrast agent administered during a medical imaging technique (e.g., a medical imaging technique performed during a radiology examination).

[0135] In some non-limiting embodiments, the predictive machine learning model may include a machine learning model designed to receive as input data related to a representation in frequency space of the examination region with an amount (e.g., a certain amount, no amount, etc.) of contrast agent administered during the medical imaging technique, and provide as output a prediction of the representation in frequency space of the examination region with a particular amount of contrast agent administered during the medical imaging technique. For example, the predictive machine learning model may be designed to receive data related to a first representation in frequency space of the examination region including a representation of the examination region with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique, and data related to a second representation in frequency space of the examination region including a representation of the examination region with a second amount of contrast agent administered during the medical imaging technique, and provide an output including a prediction of the representation in frequency space of the examination region with a third amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the radiology machine learning system 102 may store (e.g., for later use) the predictive machine learning model.

[0136] In some non-limiting embodiments, as described herein, the radiation machine learning system 102 may process data associated with a representation in frequency space of the examination region of the inspected object to obtain training data (e.g., a training data set) for a predictive machine learning model. For example, the radiation machine learning system 102 may process the data to modify the data into a format that can be analyzed (e.g., by the radiation machine learning system 102) to generate a predictive machine learning model. The modified data (e.g., data resulting from the modification) may be referred to as training data. In some non-limiting embodiments, the radiation machine learning system 102 may process data associated with a representation in frequency space of the examination region of the inspected object during a time interval to obtain training data based on receiving the data. Additionally or alternatively, the radiation machine learning system 102 may process the data to obtain training data based on the radiation machine learning system 102 receiving an instruction from a user of the radiation machine learning system 102 (e.g., a user associated with the user device 106), such as when the radiation machine learning system 102 receives an instruction to generate a predictive machine learning model for a time interval corresponding to the data, the radiation machine learning system 102 processes the data.

[0137] In some non-limiting embodiments, the radiology machine learning system 102 can process data related to the representation in frequency space of the examination region of the examination object by determining predictor variables based on the data. The predictor variables can include metrics related to the examination region or the examination object and can be derived based on data related to the representation in frequency space of the examination region of the examination object. The predictor variables can be analyzed to generate a predictive machine learning model. For example, the predictor variables can include variables related to the examination object, such as the examination object's gender, age, weight, height, medical history, nature and / or duration and / or amount of medications taken, blood pressure, central venous pressure, respiration rate, serum albumin, total bilirubin, blood glucose, iron content, respiratory capacity, etc. Additionally or alternatively, the predictor variables can include variables related to the examination region, such as variables related to pre-existing conditions of the examination region, medical procedures related to the examination region, such as partial resection of the examination region, organ conditions, such as whether a liver transplant has been performed, whether iron liver is present, whether fatty liver is present, etc. Additionally or alternatively, the predictor variables can include variables related to radiology examinations performed on the examination object, such as variables related to the type of radiology examination performed on the examination object.

[0138] In some non-limiting embodiments, the radiology machine learning system 102 can analyze the training data to generate a predictive machine learning model. For example, the radiology machine learning system 102 can use machine learning techniques to generate the predictive machine learning model and analyze the training data. In some non-limiting embodiments, generating a predictive machine learning model (e.g., based on training data obtained from historical data related to a representation in frequency space and / or real space of an examination region of an examination subject) can be referred to as training the predictive machine learning model. Machine learning techniques can include, for example, supervised and / or unsupervised techniques, such as decision trees, random forests, logistic regression, linear regression, gradient boosting, support vector machines, extra trees (e.g., an extension of random forests), Bayesian statistics, learning automata, hidden Markov modeling, linear classifiers, quadratic classifiers, association rule learning, and / or the like. In some non-limiting embodiments, the predictive machine learning model can include a model specific to a particular characteristic, such as a model specific to a particular examination region, a particular examination subject, a particular amount of contrast agent administered during a medical imaging technique, etc. Additionally or alternatively, the predictive machine learning models may be specific to a particular type of examination subject (e.g., a type of mammal, such as a human), including a particular type of examination region. In some non-limiting embodiments, the radiology machine learning system 102 can generate one or more predictive machine learning models for one or more entities, particular groups of entities, and / or one or more users of the one or more entities.

[0139] Additionally or alternatively, when analyzing the training data, the radiation machine learning system 102 may identify one or more variables (e.g., one or more independent variables) as predictor variables (e.g., features) that may be used to make predictions when analyzing the training data. In some non-limiting embodiments, the values ​​of the predictor variables may be inputs to a predictive machine learning model. For example, the radiation machine learning system 102 may identify a subset (e.g., a suitable subset) of the variables as predictor variables that may be used to accurately predict the representation in frequency space of the examination region. In some non-limiting embodiments, the predictor variables may include one or more of the examination region and / or examination subject variables, as described above, that have a significant influence (e.g., an influence that meets a threshold) on the prediction of the representation in frequency space of the examination region using a particular amount of contrast agent administered during a medical imaging technique, as determined by the radiation machine learning system 102.

[0140] In some non-limiting embodiments, the radiation machine learning system 102 can validate the predictive machine learning model. For example, the radiation machine learning system 102 can validate the predictive machine learning model after the radiation machine learning system 102 generates the predictive machine learning model. In some non-limiting embodiments, the radiation machine learning system 102 can validate the predictive machine learning model based on a portion of the training data used for validation. For example, the radiation machine learning system 102 can split the training data into a first portion and a second portion, where the first portion can be used to generate the predictive machine learning model, as described above. In this example, the second portion of the training data (e.g., validation data) can be used to validate the predictive machine learning model.

[0141] In some non-limiting embodiments, the radiology machine learning system 102 can validate the predictive machine learning model by providing validation data associated with a representation in frequency space of an examination region of an examination object (e.g., data of the representation in frequency space of an examination region of an examination object) as input to the predictive machine learning model and determining, based on an output of the predictive machine learning model, whether the predictive machine learning model correctly or inaccurately predicted the representation in frequency space of the examination region with a particular amount of a contrast agent administered during a medical imaging technique. In some non-limiting embodiments, the radiology machine learning system 102 can validate the predictive machine learning model based on a validation threshold. For example, the radiology machine learning system 102 may be configured to validate the predictive machine learning model when the representation in frequency space of an examination region using a particular amount of contrast agent administered during a medical imaging technique (as identified by the validation data) is correctly predicted by the predictive machine learning model (e.g., when the predictive machine learning model accurately predicts 50% of the representation in frequency space of an examination region using a particular amount of contrast agent administered during a medical imaging technique, 70% of the representation in frequency space of an examination region using a particular amount of contrast agent administered during a medical imaging technique, or another threshold amount of the representation in frequency space of an examination region using a particular amount of contrast agent administered during a medical imaging technique, etc.).

[0142] In some non-limiting embodiments, if the radiation machine learning system 102 does not validate a predictive machine learning model (e.g., if the percentage of correctly predicted representations in frequency space of an examination region using a particular amount of contrast agent administered during a medical imaging technique does not meet a validation threshold), the radiation machine learning system 102 may generate one or more additional predictive machine learning models.

[0143] In some non-limiting embodiments, once the predictive machine learning model is validated, the radiology machine learning system 102 can further train the predictive machine learning model and / or generate a new predictive machine learning model based on receiving new training data. The new training data can include additional data related to one or more representations in frequency space of the examination region of the examined subject. In some non-limiting embodiments, the new training data can include data related to the representation in frequency space of the examination region of the examined subject. The radiology machine learning system 102 can use the predictive machine learning model to predict the representation in frequency space of the examination region of the examined subject using a particular amount of contrast agent administered during a medical imaging technique, and compare the output of the predictive machine learning model to the new training data including data related to the representation in frequency space of the examination region of the examined subject. In such an example, the radiology machine learning system 102 can update one or more predictive machine learning models based on the new training data.

[0144] In some non-limiting embodiments, the radiation machine learning system 102 can store the predictive machine learning models. For example, the radiation machine learning system 102 can store the predictive machine learning models in a data structure (e.g., a database, a linked list, a tree, etc.). The data structure can be located within the radiation machine learning system 102 or external (e.g., remote) to the radiation machine learning system 102.

[0145] 5A-5C, which are diagrams of an implementation 500 of a process (e.g., process 300) for providing a prediction of a representation of an examination region generated using a medical imaging technique that requires a contrast agent. As described below with respect to FIGS. 5A-5C, the implementation 500 may include a radiology machine learning system 102 that performs steps of the process. In some non-limiting embodiments, one or more of the steps of the process may be performed (e.g., fully, partially, and / or similarly) by another device or group of devices separate from and / or including the radiology machine learning system 102, such as a medical imaging system 104 and / or a user device 106.

[0146] As shown in FIG. 5A, the radiology machine learning system 102 can generate a representation in frequency space of the examination region of the examination object. As further shown in FIG. 5A, at three different time points t 1 , t 2 and t 3 In the present invention, the radiology machine learning system 102 may generate three representations of an examination region of an examined object. For example, the radiology machine learning system 102 may receive results of a radiology examination having medical imaging techniques performed on the examination region of the examined object by the medical imaging system 104. The examination region may be a lung of a human patient. The results may include three representations in real space of the examination region of the examined object based on the medical imaging techniques performed on the examination region by the medical imaging system 104.

[0147] The result may include three radiographic images including a first radiographic image O1, a second radiographic image O2, and a third radiographic image O3. The first radiographic image O1 may include an image of the examination area generated without an amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique. The second radiographic image O2 may include an image of the examination area generated with a second amount of contrast agent administered during the medical imaging technique. The third radiographic image O3 may include an image of the examination area generated with a third amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the second amount of contrast agent may be different from the first amount. In some non-limiting embodiments, the third amount of contrast agent may be different from the first amount and the second amount. In one example, 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.

[0148] The first representation F1 may comprise a representation of the examination area with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique. The second representation F2 may comprise a representation of the examination area with a second amount of contrast agent administered during the medical imaging technique. The third representation F3 may comprise a representation of the examination area with a third amount of contrast agent administered during the medical imaging technique.

[0149] In some non-limiting embodiments, the radiation machine learning system 102 may generate three representations in frequency space, F1, F2, and F3, of the examination of the inspection object based on the results of the radiation examination performed by the medical imaging system 104. For example, the radiation machine learning system 102 may generate three representations in frequency space, F1, F2, and F3, of the examination based on the real space representations, O1, O2, and O3, of the examination area using a Fourier transform (FT). The real space representations, O1, O2, and O3, of the examination area may be transformed by the radiation machine learning system 102 into frequency space representations, F1, F2, and F3, of the examination area using a FT. The frequency space representations, F1, F2, and F3, of the examination area may be transformed by the radiation machine learning system 102 into real space representations, O1, O2, and O3, of the examination area using an inverse Fourier transform (iFT). In some non-limiting embodiments, a transform other than a FT may be used by the radiation machine learning system 102 to transform the real space representations into frequency space representations. Three key properties of such a transform may include the existence of a clear inverse transform (e.g., a clear link between the representation in real space and the representation in frequency space), locality of contrast information, and robustness with respect to imperfect image registration.

[0150] As shown in FIG. 5B, the representations F1, F2, and F3 in frequency space of the examination region may be used to train a predictive machine learning model. The representations F1, F2, and F3 in frequency space of the examination region may be included in a training data set of the examination object. The radiology machine learning system 102 may use multiple training data sets of multiple examination objects to train a predictive machine learning model.

[0151] In some non-limiting embodiments, the predictive machine learning model is trained to provide as output, based on the first representation F1 and the second representation F2, a prediction of a representation in frequency space of the examination region with a particular amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the particular amount of contrast agent corresponds to a third amount of contrast agent administered during the medical imaging technique.

[0152] As further shown in FIG. 5B, the radiation machine learning system 102 can provide input to the predictive machine learning model, and the radiation machine learning system 102 can receive an output of the predictive machine learning model based on the input. In some non-limiting embodiments, the first representation F1 and the second representation F2 can be provided as input to the predictive machine learning model. The output of the predictive machine learning model can include a prediction F3* of the representation in the frequency space of the examination region using a third amount of contrast agent administered during the medical imaging technique. As part of the training procedure, the radiation machine learning system 102 can compare the prediction F3* of the representation in the frequency space of the examination region using a third amount of contrast agent administered during the medical imaging technique with a third representation F3 in the frequency space of the examination region using a third amount of contrast agent administered during the medical imaging technique. The deviation between the prediction F3* of the representation and the third representation F3 can be used in a backpropagation method to train the predictive machine learning model to reduce the deviation to a defined minimum value. If the predictive machine learning model is trained based on multiple training datasets of multiple examination subjects and the predicted F3* of the representation in frequency space of the examination region using a third amount of contrast agent administered during the medical imaging technique reaches a defined accuracy, the predictive machine learning model may be considered trained and may be used for prediction.

[0153] As shown in FIG. 5C, the radiology machine learning system 102 can use a predictive machine learning model to provide a prediction of a representation in frequency space of an examination region of an examined object with a particular amount of contrast agent administered during a medical imaging technique. In some non-limiting embodiments, the radiology machine learning system 102 can receive a first representation in frequency space F'1 of an examination region of an examined object. The first representation in frequency space F'1 can include a representation of the examination region with no amount of contrast agent administered during a medical imaging technique or with a first amount of contrast agent administered during a medical imaging technique. In some non-limiting embodiments, the radiology machine learning system 102 can receive a second representation in frequency space F'2 of an examination region of an examined object. In some non-limiting embodiments, the second representation in frequency space F'2 can include a representation of the examination region with a second amount of contrast agent administered during a medical imaging technique, the second amount of contrast agent being greater than the first amount of contrast agent.

[0154] As further shown in FIG. 5C, the radiology machine learning system 102 can provide input to a predictive machine learning model. The input to the predictive machine learning model can include a first representation in frequency space F′1 and a second representation in frequency space F′2. The predictive machine learning model can include a trained machine learning model configured to provide as output a prediction F′3* of a representation in frequency space of the examination region using a third amount of a contrast agent administered during a medical imaging technique. In some non-limiting embodiments, the third amount of the contrast agent is greater than the first amount of the contrast agent and the second amount of the contrast agent.

[0155] 5C, the radiation machine learning system 102 can receive an output F'3* of the predictive machine learning model based on the input and use iFT to transform the output of the predictive machine learning model into a predicted representation O'3* of the inspection area of ​​the inspected object in real space. In some non-limiting embodiments, the radiation machine learning system 102 can provide a predicted representation O'3* of the inspection area of ​​the inspected object in real space.

[0156] 6A-6B, which are diagrams of an implementation 600 of a process (e.g., process 300) for providing a prediction of a representation of an examination region generated using a medical imaging technique that requires a contrast agent. As described below with respect to FIGS. 6A-6B, the implementation 600 may include a radiology machine learning system 102 that performs steps of the process. In some non-limiting embodiments, one or more of the steps of the process may be performed (e.g., fully, partially, and / or similarly) by another device or group of devices separate from and / or including the radiology machine learning system 102, such as a medical imaging system 104 and / or a user device 106.

[0157] As shown in FIG. 6A, in the same or similar manner as implementation 500, the representations F1, F2, and F3 in frequency space of the examination region may be used to train a predictive machine learning model. The representations F1, F2, and F3 in frequency space of the examination region may be included in a training data set of the examination object. The radiology machine learning system 102 may use multiple training data sets of multiple examination objects to train a predictive machine learning model.

[0158] Similar to implementation 500, the first representation F1 may include a representation of the examination area with no amount of contrast agent administered during the medical imaging technique or with a first amount of contrast agent administered during the medical imaging technique, the second representation F2 may include a representation of the examination area with a second amount of contrast agent administered during the medical imaging technique, and the third representation F3 may include a representation of the examination area with a third amount of contrast agent administered during the medical imaging technique.

[0159] In some non-limiting embodiments, in the same or similar manner as in implementation 500, the radiology machine learning system 102 may generate three representations F1, F2, and F3 in frequency space of the examination region of the examined object based on the results of the radiology examination performed by the medical imaging system 104. In some non-limiting embodiments, the predictive machine learning model is trained to provide as output, based on a portion of the first representation and a portion of the second representation, a prediction of a portion of the representation in frequency space of the examination region with a specific amount of contrast agent administered during the medical imaging technique. In some non-limiting embodiments, the specific amount of contrast agent corresponds to a third amount of contrast agent administered during the medical imaging technique.

[0160] As further shown in FIG. 6A, the radiology machine learning system 102 generates a first reduced representation F1 in frequency space of the examination region of the examination object. red Further, the radiology machine learning system 102 may identify a portion A of the first representation F1 that includes a center in frequency space to provide a second reduced representation F2 in frequency space of the examination region of the examination object. red Further, the radiology machine learning system 102 may identify a portion A of the second representation F2 that includes a center in frequency space to provide a third reduced representation F3 in frequency space of the examination region of the examination object. red A portion A of the third representation F3 that includes a center of the frequency space can be identified to provide:

[0161] As further shown in FIG. 6A, the radiology machine learning system 102 may receive a first reduced representation F1 in frequency space of the examination region of the object as an input to the predictive machine learning model. red and a second reduced representation F2 in frequency space of the test area of ​​the test object red and the radiology machine learning system 102 can receive an output of the predictive machine learning model based on the input. In some non-limiting embodiments, the output can be a predicted F3* of a portion of the representation in frequency space of the examination region using a third amount of a contrast agent administered during a medical imaging technique.red may include.

[0162] As part of the training procedure, the radiology machine learning system 102 may be configured to calculate a predicted F3* of a portion of the representation in frequency space of the examination region using a third amount of contrast agent administered during the medical imaging technique. red into a third reduced representation F3 in the frequency space of the test area to be tested. red The prediction of part of the representation in frequency space, F3* red and the third reduced representation in frequency space F3 red The deviation between may be used in a backpropagation method to train a predictive machine learning model to reduce the deviation to a defined minimum value. The predictive machine learning model is trained based on a plurality of training data sets of a plurality of test subjects, and predicts F3* of a portion of the representation in frequency space using a third amount of a contrast agent administered during a medical imaging technique. red If ̃ reaches a defined accuracy, the predictive machine learning model may be considered trained and may be used for prediction.

[0163] As shown in FIG. 6B, the radiology machine learning system 102 can use a predictive machine learning model to provide a prediction of a representation in frequency space of an examination region of an examined object with a particular amount of contrast agent administered during a medical imaging technique. In some non-limiting embodiments, the radiology machine learning system 102 can receive a first representation in frequency space F'1 of an examination region of an examined object. The first representation in frequency space F'1 can include a representation of the examination region with no amount of contrast agent administered during a medical imaging technique or with a first amount of contrast agent administered during a medical imaging technique. In some non-limiting embodiments, the radiology machine learning system 102 can receive a second representation in frequency space F'2 of an examination region of an examined object. In some non-limiting embodiments, the second representation in frequency space F'2 can include a representation of the examination region with a second amount of contrast agent administered during a medical imaging technique, and the second amount of contrast agent can be greater than the first amount of contrast agent.

[0164] As further shown in FIG. 6B, the radiology machine learning system 102 calculates a first reduced representation F′1 in frequency space of the examination region of the examination object. red Further, the radiology machine learning system 102 may identify a portion A of the first representation F'1 that includes a center in frequency space to provide a second reduced representation F'2 in frequency space of the examination region of the examination object. red A portion A of the second representation F'2 that includes the center of the frequency space can be identified to provide:

[0165] As further shown in FIG. 6B, the radiation machine learning system 102 can provide an input to the predictive machine learning model. The input to the predictive machine learning model is a first reduced representation F′1 in frequency space. red and the second reduced representation in frequency space F'2 red The predictive machine learning model may include a prediction F'3* of a reduced representation in frequency space of the examination region using a third amount of a contrast agent administered during a medical imaging technique. redIn some non-limiting embodiments, the third amount of contrast agent is greater than the first amount of contrast agent and the second amount of contrast agent.

[0166] Further, as shown in FIG. 6B, the radiology machine learning system 102 calculates the output F′3* of the predictive machine learning model based on the input. red and the radiology machine learning system 102 may receive the supplemented output F′1 of the predictive machine learning model. D1 +F'3* red To provide the output of the predictive machine learning model, F'3* red a first reduced representation F′1 in frequency space that does not include the center of the frequency space and is a first reduced representation F′2 in frequency space of the test area to be tested red The part of the first expression F'1 that was not specified to provide D1 In some non-limiting embodiments, the radiation machine learning system 102 can use iFT to convert the supplemented output of the predictive machine learning model into a predicted representation O'3* of the inspection area of ​​the inspected object in real space. In some non-limiting embodiments, the radiation machine learning system 102 can provide a predicted representation O'3* of the inspection area of ​​the inspected object in real space.

[0167] Although the above systems, methods, and computer program products have been described in detail for purposes of illustration based on what are currently considered to be the most practical and preferred embodiments or aspects, it should be understood that such details are for this purpose only, and that the disclosure is not limited to the described embodiments or aspects, but on the contrary, is intended to cover modifications and equivalent arrangements within the spirit and scope of the appended claims. For example, it should be understood that the present disclosure contemplates that, to the extent possible, at least one feature of any embodiment or aspect can be combined with at least one feature of any other embodiment or aspect.

Claims

1. 1. A system for providing a prediction of a representation of a region of examination generated using a medical imaging technique requiring a contrast agent, comprising: receiving a first representation in frequency space of an examination region of an examination object, wherein said first representation comprises a representation of said examination region without an amount of contrast agent administered during a medical imaging technique or with a first amount of said contrast agent administered during said medical imaging technique; receiving a second representation in the frequency space of the examination region of the subject, where the second representation includes a representation of the examination region with a second amount of the contrast agent administered during the medical imaging technique, where the second amount of the contrast agent is different from the first amount of the contrast agent; providing input to a predictive machine learning model, where the input to the predictive machine learning model includes at least a portion of the first representation including a center of the frequency space and at least a portion of the second representation including a center of the frequency space, where the predictive machine learning model includes a trained machine learning model configured to provide as output a prediction of a representation in the frequency space of the examination region using a third amount of the contrast agent to be administered during the medical imaging technique, where the third amount of the contrast agent is greater than the first amount of the contrast agent and the second amount of the contrast agent; receiving an output of the predictive machine learning model based on the input; Transforming the output of the predictive machine learning model into a predicted representation in real space of the inspection region of the inspection object; and The system comprises at least one processor programmed or configured to provide the predicted representation in the real space of the examination area of ​​the examination object.

2. 10. The system of claim 1, wherein the at least one processor further comprises: to identify a portion of the first representation that includes a center of the frequency space, a portion of the second representation that includes a center of the frequency space, or a portion of the first representation and a portion of the second representation, each of which includes a center of the frequency space, to provide a reduced representation in the frequency space of the examination area of ​​the examination object; where the input to the predictive machine learning model includes the reduced representation in the frequency space of the inspection region of the inspection object; and wherein, when providing the input to the predictive machine learning model, the at least one processor: The system is programmed or configured to provide the reduced representation in the frequency space of the inspection region of the inspection object as the input to the predictive machine learning model.

3. 3. The system of claim 2, wherein the at least one processor further comprises: providing a supplemental output of the predictive machine learning model; a portion of the first representation that does not include the center of the frequency space and that has not been specified to provide the reduced representation in frequency space of the examination area of ​​the examination object; a portion of the second representation that does not include the center of the frequency space and that has not been specified to provide the reduced representation in frequency space of the examination area of ​​the object of examination; or with a portion of the first representation and a portion of the second representation that do not each include the center of the frequency space and that were not specified to provide the reduced representation in the frequency space of the examination area of ​​the examination object, programmed or configured to supplement the output of the predictive machine learning model; wherein when converting the output of the predictive machine learning model into the predicted representation in the real space of the inspection area of ​​the inspection object, the at least one processor: The system is programmed or configured to convert the supplemented output of the predictive machine learning model into the predicted representation in real space of the inspection area of ​​the inspection object.

4. 4. The system of claim 1, wherein the first amount of contrast agent administered during the medical imaging technique is greater than zero, and wherein the second amount of contrast agent administered during the medical imaging technique is greater than the first amount of contrast agent.

5. 4. The system of claim 1, wherein the first amount of contrast agent administered during the medical imaging technique is zero, and wherein the second amount of contrast agent administered during the medical imaging technique is greater than the first amount of contrast agent.

6. 4. The system of claim 1, wherein the at least one processor further comprises: generating the first representation in the frequency space of the examination area of ​​the examination object related to a first result of a radiological examination; and The system is programmed or configured to generate the second representation in the frequency space of the examination area of ​​the object for a second result of the radiological examination.

7. The system of claim 6 , wherein the radiological examination is a magnetic resonance imaging examination, a computed tomography examination, or an ultrasound examination.

8. 7. The system of claim 6, wherein the radiological examination is a magnetic resonance imaging examination; wherein, upon receiving the first representation in frequency space of the examination area of ​​the examination object in frequency space, the at least one processor: programmed or configured to receive first k-space data associated with the magnetic resonance imaging examination of the examination region of the examination subject; and wherein, upon receiving the second representation in frequency space of the examination area of ​​the examination object in frequency space, the at least one processor: The system is programmed or configured to receive second k-space data related to the magnetic resonance imaging examination of the examination region of the subject.

9. 4. The system of claim 1, wherein upon receiving the first representation in frequency space of the examination area of ​​the examination object in frequency space, the at least one processor: receiving a first representation in real space of an examination area of ​​an examined subject, where said first representation in real space comprises a representation of said examination area without an amount of contrast agent administered during a medical imaging technique or with a first amount of said contrast agent administered during said medical imaging technique; a first representation in real space of the examination area of ​​the examination object to a first representation in frequency space; wherein, upon receiving the second representation in frequency space of the examination area of ​​the examination object in frequency space, the at least one processor: receiving a second representation in the real space of the examination region of the subject, where the second representation in real space includes a representation of the examination region with a second amount of the contrast agent administered during the medical imaging technique, where the second amount of the contrast agent is different from the first amount of the contrast agent; The system is programmed or configured to transform the second representation in real space of the examination area of ​​the examination object into the second representation in frequency space.

10. 4. The system of claim 1, wherein the at least one processor further comprises: The method includes the steps of: training the predictive machine learning model based on a training dataset, the training dataset comprising: a set of reference representations in the frequency space of a test area for each test object of a plurality of test objects, each set of reference representations of the test area of ​​a test object comprising: a first reference representation in the frequency space of the test area of ​​the test object; a second reference representation in the frequency space of the test area of ​​the test object; and a third reference representation in the frequency space of the test area of ​​the test object; and wherein the first reference representation comprises a reference representation in the frequency space of the examination region with no amount of contrast agent administered during a medical imaging technique or with a first amount of the contrast agent administered during the medical imaging technique; wherein the second reference representation comprises a reference representation in the frequency space of the examination region using a second amount of the contrast agent administered during the medical imaging technique; wherein the third reference representation comprises a reference representation in the frequency space of the examination region using a third amount of the contrast agent administered during the medical imaging technique.

11. 11. The system of claim 10, wherein when training the predictive machine learning model, the at least one processor: The system is programmed or configured to minimize an amount of error provided by an error function, where the error function quantifies the deviation between a prediction of a representation in the frequency space of the examination region of the subject using the third amount of the contrast agent administered during the medical imaging technique and the third reference representation in the frequency space of the examination region of the subject.

12. 4. The system of claim 1, wherein the at least one processor further comprises: The method includes the steps of: training the predictive machine learning model based on a training dataset, the training dataset comprising: a set of reduced referring representations in the frequency space of a test area for each test object of a plurality of test objects, each set of reduced referring representations of the test area of ​​a test object comprising: a first reduced reference representation in the frequency space of the test area of ​​the test object; a second reduced reference representation in the frequency space of the test area of ​​the test object; and a third reduced reference representation in the frequency space of the test area of ​​the test object; and wherein the first reduced reference representation comprises a reference representation of a portion of the first reference representation in the frequency space of the examination area including a center of the frequency space, wherein the first reference representation comprises a reference representation of the examination area with no amount of contrast agent administered during a medical imaging technique or with a first amount of the contrast agent administered during the medical imaging technique; wherein the second reduced reference representation comprises a reference representation of a portion of the second reference representation in the frequency space of the examination area that includes a center of the frequency space, and wherein the second reference representation comprises a reference representation of the examination area using a second amount of the contrast agent administered during the medical imaging technique; wherein the third reduced reference representation includes a reference representation of a portion of the third reference representation in the frequency space of the examination area including a center of the frequency space, and wherein the third reference representation includes a reference representation of the examination area using a third amount of the contrast agent administered during the medical imaging technique.

13. 13. The system of claim 12, wherein when training the predictive machine learning model, the at least one processor: The system is programmed or configured to minimize an amount of error provided by an error function, where the error function quantifies the deviation between a prediction of a reduced representation in the frequency space of the examination region of the subject using the third amount of the contrast agent administered during the medical imaging technique and the third reduced reference representation in the frequency space of the examination region of the subject.

14. 1. A computer program product for providing a prediction of a representation of a region of examination generated using a medical imaging technique requiring a contrast agent, the computer program product, when executed by at least one processor, causing the at least one processor to: receiving a first representation in frequency space of an examination region of an examination object, wherein said first representation comprises a representation of said examination region without an amount of contrast agent administered during a medical imaging technique or with a first amount of said contrast agent administered during said medical imaging technique; receiving a second representation in the frequency space of the examination region of the subject, where the second representation includes a representation of the examination region with a second amount of the contrast agent administered during the medical imaging technique, where the second amount of the contrast agent is different from the first amount of the contrast agent; providing input to a predictive machine learning model, wherein the input to the predictive machine learning model includes at least a portion of the first representation including a center of the frequency space and at least a portion of the second representation including a center of the frequency space, wherein the predictive machine learning model includes a trained machine learning model configured to provide as output a prediction of a representation in the frequency space of the examination region using a third amount of the contrast agent administered during the medical imaging technique, wherein the third amount of the contrast agent is greater than the first amount of the contrast agent and the second amount of the contrast agent; receiving the output of the predictive machine learning model based on the input; Transforming the output of the predictive machine learning model into a predicted representation in real space of the inspection region of the inspection object; and at least one non-transitory computer-readable medium comprising one or more instructions for causing the predicted representation in the real space of the examination area of ​​the examination object to be provided.

15. 1. A method for providing a prediction of a representation of a region of examination generated using a medical imaging technique requiring a contrast agent, comprising: receiving, by at least one processor, a first representation in frequency space of an examination region of an examined object, wherein said first representation includes a representation of said examination region without an amount of contrast agent administered during a medical imaging technique or with a first amount of said contrast agent administered during said medical imaging technique; receiving, by the at least one processor, a second representation in the frequency space of the examination region of the subject, where the second representation includes a representation of the examination region with a second amount of the contrast agent administered during the medical imaging technique, where the second amount of the contrast agent is different from the first amount of the contrast agent; providing, by the at least one processor, an input to a predictive machine learning model, where the input to the predictive machine learning model includes at least a portion of the first representation including a center of the frequency space and at least a portion of the second representation including a center of the frequency space, where the predictive machine learning model includes a trained machine learning model configured to provide as an output a prediction of a representation in the frequency space of the examination region using a third amount of the contrast agent administered during the medical imaging technique, where the third amount of the contrast agent is greater than the first amount of the contrast agent and the second amount of the contrast agent; receiving, by the at least one processor, the output of the predictive machine learning model based on the input; converting, by the at least one processor, the output of the predictive machine learning model into a predicted representation in real space of the inspection region of the inspection object; and providing, by the at least one processor, the predicted representation in the real space of the examination area of ​​the examination object. The method comprising:

Images