Providing an indication that a patient has a contagious respiratory illness based on magnetic resonance imaging data

Abstract

Embodiments of the invention generally relate to a computer-implemented method for providing output data comprising an indication that a patient has a contagious respiratory disease, a computer-implemented method for providing a trained function, a providing system, a training system, a computer program and a computer readable medium. The computer-implemented method for providing output data comprising an indication that a patient has a contagious respiratory disease comprises the following steps: - receiving magnetic resonance imaging data acquired using a magnetic resonance imaging system, wherein the magnetic resonance imaging data comprises a lung region of the patient; - applying a trained function to the magnetic resonance imaging data, wherein output data is generated, wherein the trained function is based on an artificial neural network and the output data comprises an indication that the patient has a contagious respiratory disease; - providing the output data.

Description

[0001] Embodiments of the present invention generally relate to a computer-implemented method for providing output data including an indication that a patient has an infectious respiratory disease, a computer-implemented method for providing a trained function, a providing system, a training system, a computer program, and a computer-readable medium.

[0002] In magnetic resonance imaging (MRI) systems (also known as magnetic resonance computed tomography) systems, when using a fundamental field magnet, the subject of examination (e.g., a patient, a healthy training volunteer, or a phantom) is exposed to a relatively high fundamental magnetic field, such as 1.5 Tesla, 3 Tesla, or 7 Tesla. Furthermore, gradient pulses are activated using a gradient coil arrangement. A radio frequency (RF) antenna arrangement emits RF pulses (e.g., excitation pulses) from a suitable antenna, causing the nuclear spins of specific atoms excited by these RF pulses to tilt relative to the magnetic field lines of the fundamental magnetic field at a defined flip angle. During nuclear spin relaxation, an RF signal, known as a magnetic resonance signal, is emitted, received by a suitable RF antenna, and then further processed. Finally, the desired image data can be reconstructed from the raw data acquired in this way.

[0003] Infectious respiratory diseases are caused by pathogens (viruses, bacteria, or other pathogenic microorganisms) that can infect the respiratory system (i.e., the lungs, throat, and airways) of infected individuals. Such infectious respiratory diseases are usually spread through the mucus and saliva (also known as "respiratory secretions") that a person expels when coughing, sneezing, talking, or laughing.

[0004] A recent example of an infectious respiratory disease is COVID-19 (Coronavirus Disease 2019). COVID-19 is caused by Severe Acute Respiratory Coronavirus 2 (SARS-CoV-2). Common symptoms of COVID-19 include fever, cough, and difficulty breathing. In severe cases, COVID-19 can lead to pneumonia, severe acute respiratory syndrome, and multiple organ failure. In most cases, patients infected with COVID-19 experience mild to moderate symptoms that do not require hospitalization. However, COVID-19 is fatal for a significant proportion of those infected. Due to COVID-19's high reproduction number (R0) and high transmissibility, tools for rapid testing and assessment are crucial for tracing and slowing its spread.

[0005] In current clinical practice, COVID-19 is diagnosed using RT-PCR (reverse transcription polymerase chain reaction). The sensitivity of such diagnostic tools is currently under discussion. Furthermore, the limited supply of RT-PCR test kits has led to the spread of undetected COVID-19.

[0006] Recent research has shown that accurate diagnoses can also be performed based on computed tomography (CT) scans and chest X-rays—including the use of deep learning image analysis algorithms, see: Gozes et al., “Rapid AI Development Cycle for the Coronavirus (COVID-19) Pandemic: Initial Results for Automated Detection & Patient Monitoring using Deep Learning CT Image Analysis”, 2020, arXiv: 2003.05037. One drawback of this approach is radiation exposure, which places an additional burden on patients and may hinder the acquisition of the large amounts of training data required to train the deep learning image analysis algorithms.

[0007] The article by Salehi et al., “Long-term pulmonary consequences of Coronavirus Disease 2019 (COVID-19)” (Thorac Imaging, Vol. 35, No. 4, July 2020), demonstrates that long-term functional impairment may persist even after recovery from infectious lung diseases such as COVID-19 or SARS. Measuring the successive recovery of such patients (particularly their lungs) in longitudinal follow-up imaging studies has high clinical value. However, radiation exposure from CT and chest X-ray imaging can be problematic when performing such longitudinal follow-up imaging studies.

[0008] Assessment of lung function parameters can also be performed using non-invasive pulmonary function tests (PFTs) such as spirometry or plethysmography, without subjecting the patient to ionizing radiation. A drawback of these non-invasive methods is that the information obtained through PFTs is typically not spatially resolved. Therefore, only a holistic measurement of lung function can be derived from the PFT. Spatially resolved functional parameters of the lungs, such as ventilation and / or perfusion parameters, can also be obtained based on scintillation imaging using appropriate isotopes. However, this method involves the necessity of using radioactive materials and results in long scan times.

[0009] The object of the present invention is to enable the provision of output data, including indications of a patient having an infectious respiratory disease, in an effective and radiation-dosage-efficient manner.

[0010] A computer-implemented method of the present invention for providing output data including at least one embodiment of an indication that a patient has an infectious respiratory disease includes:

[0011] - Receive magnetic resonance imaging (MRI) data acquired using an MRI system, including data on the patient's lung regions.

[0012] - A trained function is applied to magnetic resonance imaging data, wherein output data is generated, wherein the trained function is based on an artificial neural network, and the output data includes indications about whether the patient has an infectious respiratory disease.

[0013] - Provides output data.

[0014] Receiving magnetic resonance imaging (MRI) data may include acquiring MRI data via an MRI system or loading previously acquired MRI data from an image database. MRI data including a patient's lung region specifically means that the MRI data shows or images a region of the patient's lungs. Specifically, at least one portion of the MRI shows at least a portion of the patient's lung region. The lung region (also referred to as the lung area) typically includes the patient's right lung and / or left lung. Furthermore, the lung region typically includes the patient's bronchioles and alveoli. The lung region may also include at least a portion of the patient's diaphragm and / or trachea and / or pleura. MRI data typically includes reconstructed data, which includes the spatial distribution of MRI signal values ​​and / or the spatial distribution of quantitative MRI values ​​(e.g., T1 maps and / or T2 maps) and / or the spatial distribution of MRI functional data. MRI data may also include k-space data representing the spatial MRI frequencies before reconstruction.

[0015] The received magnetic resonance imaging (MRI) data can be fed into a trained function to generate output data. Therefore, the MRI data forms the input to the trained function, and the output data forms the output of the trained function.

[0016] Typically, trained functions mimic human cognitive functions associated with other aspects of human thought. In particular, by training based on training data, trained functions can adapt to new situations and detect and infer patterns.

[0017] Typically, the parameters of a trained function can be tuned through training. Specifically, supervised training, semi-supervised training, unsupervised training, reinforcement learning, and / or active learning can be used. Furthermore, representation learning (an alternative term is "feature learning") can be used. In particular, the parameters of a trained function can be iteratively tuned through a series of training steps.

[0018] In this way, the output data is provided based on machine learning methods (also known as deep learning methods), which are based on artificial neural networks. Artificial neural networks (ANNs) are, in particular, networks of artificial neurons that are mimicked in computer programs.

[0019] Specifically, the artificial neural network can be a deep neural network, a convolutional neural network, or a convolutional deep neural network. Furthermore, the neural network can be an adversarial network, a deep adversarial network, and / or a generative adversarial network.

[0020] In this context, artificial neural networks are typically based on a network of many artificial neurons. These neurons are usually arranged in different layers. Typically, an artificial neural network includes an input layer and an output layer, where the neuron outputs are the only visible output. The layers located between the input and output layers are usually called hidden layers.

[0021] Typically, the architecture and / or topology of an artificial neural network are first established, and then trained during the training phase for a specific task or multiple tasks. In such cases, training an artificial neural network usually involves changing the weights of the connections between two artificial neurons. Training an artificial neural network can also include: developing new connections between artificial neurons, deleting existing connections between artificial neurons, adjusting the thresholds of artificial neurons, and / or inserting or deleting artificial neurons.

[0022] Specifically, the artificial neural network has been pre-trained appropriately to provide output data. In this case, medical training data records have been used to train the artificial neural network. Typically, medical training data records are obtained from examination subjects that are different from the subjects being examined.

[0023] Specifically, magnetic resonance imaging data can be passed to the input layer of an artificial neural network, particularly the neurons / nodes of the input layer. The output layer of the artificial neural network—particularly at least one node of the output layer—can include output data, i.e., an indication that the patient has an infectious respiratory disease.

[0024] Specifically, the trained function can be trained in such a way that the indication of a patient having an infectious respiratory disease is based on abnormalities detected in magnetic resonance imaging data. Therefore, the trained function can be specifically trained to detect such abnormalities. In the case of COVID-19, such abnormalities to be detected by the trained function can include at least one of the following abnormalities in MRI data associated with COVID-19 (according to the following studies: Fields et al., “Imaging of COVID-19: CT, MRI, and PET”, Semin Nucl Med, November 30, 2020, https: / / doi.org / 10.1053 / j.semnuclmed.2020.11.003; and Langenbach et al., “MRI appearance of COVID-19 infection”, Diagn Interv Radiol 2020, 26:377-378):

[0025] –Ground glass-like shading (GGO)

[0026] - Signs of edema

[0027] - Focal pulmonary infiltration,

[0028] - Hyperintense consolidation on T2-weighted images

[0029] - False emptiness

[0030] -Crazy paving patterns

[0031] - Pleural effusion.

[0032] Therefore, a trained function can be trained using training magnetic resonance imaging data, which includes abnormalities indicating infectious respiratory diseases.

[0033] Specifically, providing output data includes displaying an indication on an output unit (e.g., a display unit) that the patient has an infectious respiratory disease. Alternatively or supplementarily, providing output data may include storing the indication that the patient has an infectious respiratory disease in a database. Alternatively or supplementarily, the output data may be transmitted to another processing unit that may perform further processing on the output data based on the indication that the patient has an infectious respiratory disease.

[0034] Infectious respiratory diseases can be any infectious respiratory disease that can be identified on a magnetic resonance imaging (MRI) image. For example, an infectious respiratory disease can be at least one disease selected from the group consisting of: Middle East Respiratory Syndrome (MERS), Severe Acute Respiratory Syndrome (SARS), pneumonia, coronavirus disease 2019 (COVID-19), and tuberculosis. Of course, one may consider other infectious respiratory diseases not mentioned in this list that relate to the method of the present invention.

[0035] Indications regarding a patient having an infectious respiratory illness may include a binary classifier indicating whether the patient has a respiratory illness or not. Alternatively, indications regarding a patient having an infectious respiratory illness may include a value indicating the probability that the patient has a respiratory illness. Alternatively or supplementally, indications regarding a patient having an infectious respiratory illness may include an assessment value indicating the severity of the infectious respiratory illness the patient has. Output data may include an indication of whether the patient has a specific infectious respiratory illness. Alternatively, output data may distinguish between different infectious respiratory illnesses and healthy patients. Indications regarding a patient having an infectious respiratory illness may also simply include a prompt and / or warning message to the physician that the physician should review the acquired MRI data, which may indicate that the patient has an infectious respiratory illness, more carefully.

[0036] Alternatively, the output data may form intermediate data, which is further processed to determine an indication that the patient has an infectious respiratory disease. For example, the intermediate data may include segmentations of regions in magnetic resonance imaging data that contain typical abnormalities indicative of an infectious respiratory disease. The intermediate data can then be further processed, for example, by measuring the size of the segmented regions, to determine an indication that the patient has an infectious respiratory disease. In this case, the output data is only provided to a processing application that performs further processing on the output data. The determination that the patient has an infectious respiratory disease based on the further processing of the output data can then be provided to, for example, an output unit (e.g., a display unit).

[0037] According to one embodiment of the method of the present invention, the magnetic resonance imaging system has a main magnetic field strength of less than 1.0 Tesla. According to one embodiment of the method of the present invention, the magnetic resonance imaging system has a main magnetic field strength of less than 0.7 Tesla. According to one embodiment of the method of the present invention, the magnetic resonance imaging system has a main magnetic field strength between 0.5 Tesla and 0.6 Tesla.

[0038] Generally, a magnetic resonance imaging (MRI) system with such a main magnetic field strength can be considered a low-field MRI system. Using a low-field MRI system to acquire MRI data of the lungs may be particularly advantageous. Due to the use of a low main magnetic field strength, distortion in the MRI images of the lungs is reduced. Furthermore, the altered relaxation time and reduced signal phase loss resulting from using a low-field MRI system can produce higher quality MRI images of the lungs. These effects have been described in the following literature: Campbell-Washburn et al., “Opportunities in Interventional and Diagnostic Imaging by Using High-Performance Low-Field-Strength MRI”, Radiology, November 2019, 293(2):384-393, doi:10.1148 / radiol.2019190452, Epub October 1, 2019. In summary, higher-quality MRI data of the lungs obtained using low-field MRI systems will lead to more robust classification using trained functions and more accurate indications of whether a patient has an infectious respiratory disease.

[0039] According to one embodiment of the method of the present invention, the magnetic resonance imaging system has a system architecture specifically designed for imaging the chest region of a patient. This magnetic resonance imaging system can be considered a dedicated magnetic resonance imaging system for imaging the lungs of a patient. Furthermore, this magnetic resonance imaging system can be considered a practitioner magnetic resonance imaging system and / or a medical point magnetic resonance imaging system. For example, an open system architecture can be used, which is particularly useful for low main magnetic field strengths of less than 1.0 Tesla. Even ultra-low main magnetic field strengths can be considered. One advantage of using a dedicated magnetic resonance imaging system in this application is that such a magnetic resonance imaging system can be rapidly deployed to locations where infectious respiratory diseases are currently severe.

[0040] According to one embodiment of the method of the present invention, the magnetic resonance imaging data includes morphological magnetic resonance imaging data acquired using at least one magnetic resonance imaging pulse sequence selected from the group consisting of: T2-weighted HASTE pulse sequence, T2-weighted BLADE pulse sequence, T2-weighted turbine spin echo pulse sequence, T1-weighted gradient echo pulse sequence, true FISP pulse sequence, pulse sequence for measuring free induction attenuation (FID), radial VIBE pulse sequence, helical VIBE pulse sequence, GRASE pulse sequence, and radial TSE pulse sequence.

[0041] HASTE pulse sequences represent half-Fourier acquisition single-shot turbine spin echo imaging. HASTE sequences are also known by other names such as SS-FSE, SSH-TSE, UFSE, single-shot fast SE, FASE, and super FASE. HASTE is a single-shot technique. This means that data from all k-spaces is acquired after a single 90° excitation pulse. HASTE images can be T2-weighted, but can also have other signal weights, depending on the echo time used to acquire the MRI data.

[0042] BLADE pulse sequences are also known by other names such as PROPELLER, MultiVane, RADAR, and JET. A BLADE sequence involves sampling the k-space in a rotating manner using a set of radially oriented strips or "blades." Each blade consists of multiple parallel phase-coded lines that can be collected using fast spin echo or gradient echo methods.

[0043] Turbine spin echo pulse sequences are also known by other names such as TSE, Fast Spin Echo (FSE), or RARE (Fast Acquisition with Relaxation Enhancement). Turbine spin echo pulse sequences are similar to conventional spin echo (SE) pulse sequences in that they use a series of 180° refocusing pulses after a single 90° pulse to generate a series of echoes. However, the phase encoding gradient of each of these echoes is altered, allowing multiple lines in k-space to be acquired within a given repetition time.

[0044] The true FISP pulse sequence is also known by other names, such as FIESTA and T2-FFE. The true FISP pulse sequence is a steady-state coherent pulse sequence using balanced gradients along all three axes.

[0045] Pulse sequences for measuring free induction attenuation (FID) can be, for example, ultrashort echo time (UTE) pulse sequences, point-by-point encoded time-reduced (PETRA) pulse sequences with radial acquisition, or zero echo time (zTE) pulse sequences. UTE pulse sequences typically include echo times of 500 μs or less. PETRA pulse sequences are a hybrid approach of zero echo time and single-point imaging techniques. PETRA pulse sequences anticipate that the outer k-space is filled with a radial half-projection, while the center is measured point-by-point on a Cartesian trajectory.

[0046] Radial Volume Interpolation Breath-Break Check (VIBE) pulse sequences are 3D gradient echo sequences that use a radial star stacking sampling scheme to acquire k-space information.

[0047] The acronym GRASE stands for "Gradient and Spin Echo". GRASE is a hybrid pulse sequence technique that includes both gradient echo and spin echo acquisition.

[0048] According to one embodiment of the method of the present invention, the magnetic resonance imaging (MRI) data includes MRI data acquired using a combination of T2-weighted HASTE pulse sequences and T1-weighted gradient echo sequences. According to another embodiment of the method of the present invention, the MRI data includes MRI data acquired using a combination of helical radial ultrashort echo time-lapse pulse sequences or 3D radial ultrashort echo time-lapse pulse sequences and T2-weighted TSE pulse sequences or PD-weighted TSE pulse sequences. According to yet another embodiment of the method of the present invention, the MRI data includes MRI data acquired using a combination of T2-weighted BALDE pulse sequences and radial VIBE pulse sequences. Such combinations of MRI pulse sequences can be particularly useful because they allow for the time-efficient acquisition of suitable MRI data as input to a trained function.

[0049] According to one embodiment of the method of the present invention, the magnetic resonance imaging data includes functional magnetic resonance imaging (fMRI) data, which includes lung ventilation maps and / or lung perfusion maps. Lung ventilation maps and / or lung perfusion maps can be acquired using magnetic resonance lung ventilation imaging and / or magnetic resonance lung perfusion imaging. Such magnetic resonance lung ventilation imaging and / or magnetic resonance lung perfusion imaging may include dynamic contrast enhancement imaging, for example, including the application of a paramagnetic contrast agent such as Gd-DTPA. Magnetic resonance lung ventilation imaging and / or magnetic resonance lung perfusion imaging may include dynamic contrast enhancement (DCE) imaging. Using magnetic resonance lung ventilation imaging and / or magnetic resonance lung perfusion imaging, areas of reduced and increased parenchymal ventilation and / or perfusion can be specifically identified as areas showing decreased and increased contrast enhancement. Non-contrast enhancement methods can also be used to acquire lung ventilation maps and / or lung perfusion maps, such as pulmonary perfusion artery spin labeling (ASL) technology. Such functional magnetic resonance imaging data can allow for accurate classification of whether a patient has an infectious respiratory disease. Applying trained functions to lung ventilation and / or lung perfusion maps allows for the accurate generation of indications about a patient's infectious respiratory illness, as infectious respiratory illnesses such as COVID-19 can impair lung function, such as ventilation and / or perfusion. For example, ventilation-perfusion mismatch has been observed in moderate to severe COVID-19 cases. Areas with potential lung damage (e.g., ventilation and / or perfusion defects) can be readily detected on lung ventilation and / or lung perfusion maps. Furthermore, assessing the progression of infectious respiratory illness over time based on lung ventilation and / or lung perfusion maps acquired from the same patient at different consecutive time points can be beneficial. For example, the extent of ventilation and / or perfusion defects can be compared between MRI data acquired from the same patient in subsequent imaging studies performed at specific time intervals. The temporal development of damaged lung regions can reveal information about the recovery process.

[0050] According to one embodiment of the method of the present invention, functional magnetic resonance imaging (fMRI) data includes lung ventilation maps and / or lung perfusion maps obtained based on Fourier decomposition (FD) applied to time series of dynamic magnetic resonance (MRI) data. Therefore, dynamic MRI data is acquired from a patient, and FD is applied to the dynamic MRI data to estimate lung ventilation maps and / or lung perfusion maps. The time series of dynamic MRI data can be generated based on dynamic MRI scans, or it can be generated retrospectively from free-breathing MRI scans. Typically, registration is performed to eliminate respiratory motion from the time series of dynamic MRI data. Subsequently, voxel-by-voxel frequency analysis is performed on the registered MRI data. This allows estimation of deformation of the lung parenchyma and / or pulmonary blood flow, which can be used to calculate lung ventilation maps and / or lung perfusion maps. A possible implementation of this method is the phase-resolved functional lung (PREFUL) magnetic resonance imaging proposed by Voskrebenzev et al. for assessing non-communicable respiratory diseases such as COPD or cystic fibrosis (see Voskrebenzev et al., “Feasibility of quantitative regional ventilation and perfusion mapping with phase-resolved functional lung (PREFUL) MRI in healthy volunteers and COPD, CTEPH, and CF patients,” *Magnetic Resonance in Medicine*, Vol. 79, No. 4, August 2017). Alternatively or additionally, matrix bundle decomposition and / or non-uniform Fourier decomposition (NUFD) can be used to compute lung ventilation maps and / or lung perfusion maps. The use of Fourier decomposition methods allows for the estimation of lung ventilation maps and / or lung perfusion maps without the application of contrast agents to the patient during MRI.

[0051] According to one embodiment of the method of the present invention, the magnetic resonance imaging data includes a combination of morphological magnetic resonance imaging data and functional magnetic resonance imaging data, the functional magnetic resonance imaging data including lung ventilation maps and / or lung perfusion maps. Specifically, the morphological magnetic resonance imaging data and the functional magnetic resonance imaging data can jointly form the input vector of an artificial neural network. Therefore, providing an indication that a patient has an infectious respiratory disease can be based on a large input dataset. The magnetic resonance imaging data may include a combination of lung ventilation maps and at least one magnetic resonance imaging pulse sequence selected from the group consisting of: T2-weighted HASTE pulse sequences, T2-weighted BLADE pulse sequences, T2-weighted turbine spin echo pulse sequences, T1-weighted gradient echo pulse sequences, true FISP pulse sequences, pulse sequences for measuring free induction attenuation (FID), and radial VIBE pulse sequences. Magnetic resonance imaging (MRI) data may include a combination of lung perfusion maps and at least one MRI pulse sequence selected from the group consisting of: T2-weighted HASTE pulse sequences, T2-weighted BLADE pulse sequences, T2-weighted turbine spin echo pulse sequences, T1-weighted gradient echo pulse sequences, true FISP pulse sequences, pulse sequences for measuring free induction attenuation (FID), and radial VIBE pulse sequences. MRI data may also include a combination of lung ventilation maps and lung perfusion maps with at least one MRI pulse sequence selected from the group consisting of: T2-weighted HASTE pulse sequences, T2-weighted BLADE pulse sequences, T2-weighted turbine spin echo pulse sequences, T1-weighted gradient echo pulse sequences, true FISP pulse sequences, pulse sequences for measuring free induction attenuation (FID), and radial VIBE pulse sequences.

[0052] In all cases where T2-weighted contrast is mentioned, proton density (PD) contrast can also be used alternatively or additionally to acquire MRI data. Furthermore, using diffusion-weighted imaging (DWI) to acquire MRI data can be advantageous, for example, to identify lung lesions in the acquired MRI data.

[0053] According to one embodiment of the method of the present invention, the magnetic resonance imaging data includes magnetic resonance fingerprint data acquired using magnetic resonance fingerprinting technology. According to another embodiment of the method of the present invention, the magnetic resonance fingerprint data includes at least two spatial resolution maps, said at least two spatial resolution maps being selected from the group consisting of: T1 map, T2 map, and PD map.

[0054] Magnetic resonance fingerprinting is a quantitative magnetic resonance imaging (MRI) method that determines quantitative values ​​of tissue parameters of an examined object, and thus allows for the creation of a tissue parameter map. One advantage of MRI fingerprinting is that it allows for the simultaneous acquisition of multiple tissue parameters, such as T1 relaxation time and T2 relaxation time, in a single measurement. In this way, MRI fingerprinting eliminates the need for multiple different capture sequences to acquire multiple tissue parameters, thereby reducing the complexity and / or acquisition time of MRI examinations.

[0055] Magnetic resonance fingerprinting methods typically employ pseudo-random or incoherent acquisition schemes to acquire the magnetic resonance signal waveform of voxels, and then provide a signal comparison between the magnetic resonance signal waveform and multiple database signal waveforms.

[0056] The acquired magnetic resonance signal waveform indicates how the signal value of the magnetic resonance signal recorded in voxels changes during the acquisition of the magnetic resonance signal waveform. In this case, voxels indicate regions from which the magnetic resonance signal waveform is acquired. It is also conceivable to acquire the magnetic resonance signal waveform from regions with a coarser resolution than voxels. In this case, the magnetic resonance signal waveform can be acquired, for example, by averaging over multiple voxels. In typical applications, multiple magnetic resonance signal waveforms are acquired simultaneously from multiple voxels. For this purpose, multiple raw images can be acquired in time series using the magnetic resonance fingerprinting method, and then multiple magnetic resonance signal waveforms can be formed from the individual voxels of the multiple raw images. In this case, the temporal resolution of the magnetic resonance signal waveform is determined by the time distance between the acquisition of different magnetic resonance signals. In the magnetic resonance fingerprinting method, different scanning parameters are set to acquire different magnetic resonance signals. The scanning parameters can be varied in a pseudo-random or incoherent manner. Possible scanning parameters that vary during the acquisition of the magnetic resonance signal waveform include, for example, echo time, the implementation and / or number of radio frequency pulses, the implementation and / or number of gradient pulses, the diffusion coding sequence, etc. In this way, magnetic resonance signal waveforms, which are characteristics of voxels (called voxel fingerprints), can be obtained using magnetic resonance fingerprinting methods.

[0057] In this configuration, multiple database signal waveforms compared with the acquired magnetic resonance signal waveform are stored in a signal waveform database. In this configuration, different database values ​​of at least one tissue parameter are advantageously assigned to each of the different database signal waveforms. Then, in each case, the specific database signal waveform represents the signal waveform expected in the magnetic resonance fingerprinting method when examining a sample whose material properties correspond to the database values ​​associated with at least one tissue parameter. The database signal waveforms can be determined, for example, in calibration measurements and / or can be simulated. The magnetic resonance fingerprinting method typically specifies that one of the multiple database signal waveforms is assigned to the acquired magnetic resonance signal waveform based on the result of the signal comparison. In this configuration, the signal comparison can be to determine the similarity between the acquired magnetic resonance signal waveform and the multiple database signal waveforms, wherein the database signal waveform is assigned to the voxel exhibiting the greatest similarity to the magnetic resonance signal waveform. The database value of at least one tissue parameter belonging to the assigned database signal waveform can then be set as the measured value of at least one tissue parameter. The measured value of at least one tissue parameter of the voxel obtained by the magnetic resonance fingerprinting method can then be displayed or stored as the result of the signal comparison. Possible tissue parameters that can be quantified using magnetic resonance fingerprinting include: T1 relaxation time, T2 relaxation time, diffusion value (e.g., apparent diffusion coefficient (ADC)), magnetic moment, proton density, resonance frequency, and substance concentration. Other tissue parameters that are considered beneficial by those skilled in the art are also conceivable. Of course, conventional T1-mapping techniques and / or T2-mapping methods can also be used to obtain quantitative T1-maps and / or T2-maps.

[0058] According to one embodiment of the method of the present invention, the magnetic resonance imaging data includes contrast-based magnetic resonance imaging data acquired using at least one technique selected from the group consisting of: inhalation of an oxygen-based contrast agent by the patient, administration of a hyperpolarizing contrast agent to the patient, and administration of a fluorine-19 contrast agent to the patient. Therefore, an appropriate increase in image contrast in the magnetic resonance imaging data can be achieved, further enabling the accurate classification of whether a patient has an infectious respiratory disease using a trained function. Furthermore, lung ventilation and / or lung perfusion maps can be measured based on the magnetic resonance data acquired after the patient inhales an oxygen-based contrast agent, after administration of a hyperpolarizing contrast agent, or after administration of a fluorine-19 contrast agent.

[0059] According to one embodiment of the method of the invention, the magnetic resonance imaging (MRI) data includes MRI data acquired specifically using a single T2-weighted imaging sequence or a PD-weighted imaging sequence. According to another embodiment of the method of the invention, the MRI data includes MRI data acquired specifically using a single-helical radial ultrashort echo time-pulse sequence or a 3D radial ultrashort echo time-pulse sequence. MRI data acquired specifically using such sequences can include sufficient information to allow for the classification of whether a patient has an infectious respiratory disease. Therefore, MRI data can be acquired in a very short measurement time.

[0060] According to one embodiment of the method of the invention, magnetic resonance imaging (MRI) data is acquired during a patient's free breathing. Suitable techniques, such as rapid scan sequences and / or MRI motion correction techniques, make acquisition during free breathing feasible. For example, radial VIBE (StarVIBE) MRI sequences have been shown to be robust to patient motion and can therefore be used effectively during free breathing acquisition. Alternatively or additionally, triggering and / or gating techniques can help acquire MRI data during a patient's free breathing. Therefore, MRI data can be acquired without requiring the patient to hold their breath during acquisition. This can reduce the burden on the patient, especially in cases where infectious respiratory illnesses make breath-holding difficult.

[0061] According to one embodiment of the method of the present invention, the artificial neural network is a convolutional neural network, particularly a deep convolutional neural network (i.e., an artificial neural network that typically includes at least one convolutional layer and typically also includes at least one local pooling layer or a global pooling layer). According to one embodiment of the method of the present invention, the artificial neural network is a recurrent neural network. Specifically, a recurrent neural network is an artificial neural network in which the connections between nodes form a directed graph along a time series. In particular, a recurrent neural network can be interpreted as a directed acyclic graph. Of course, the trained function can also be based on other types of artificial neural networks that are conceived by those skilled in the art.

[0062] According to one embodiment of the method of the present invention, the output data includes an indication of whether the patient has an infectious respiratory disease. Specifically, the indication regarding whether the patient has an infectious respiratory disease may be an indication of whether the patient has an infectious respiratory disease.

[0063] According to one embodiment of the method of the present invention, the infectious respiratory disease is COVID-19 (coronavirus disease 2019).

[0064] According to one embodiment of the method of the present invention, the output data includes a classification of whether the patient has COVID-19 (Coronavirus Disease 2019). In this case, an indication that the patient has an infectious respiratory disease can be formed by classifying whether the patient has COVID-19 (Coronavirus Disease 2019).

[0065] According to one embodiment of the method of the present invention, the output data includes a classification of whether the patient suffers from a disease selected from the group consisting of: pneumonia, MERS, and SARS. In this case, an indication that the patient has an infectious respiratory disease can be formed by classifying whether the patient suffers from a disease selected from the group consisting of: pneumonia, MERS, and SARS.

[0066] According to one embodiment of the method of the present invention, the output data includes the following classifications: whether the patient has COVID-19, whether the patient has at least one different infectious respiratory disease, or whether the patient is healthy. In this case, the indication that the patient has an infectious respiratory disease can be formed by the following classifications: whether the patient has COVID-19, whether the patient has at least one different infectious respiratory disease, or whether the patient is healthy. According to one embodiment of the method of the present invention, at least one different respiratory disease includes at least one infectious respiratory disease selected from the group consisting of: pneumonia, MERS, and SARS.

[0067] According to one embodiment of the method of the present invention, the output data includes the following classification: whether the patient with COVID-19 is in the acute COVID-19 stage or whether the patient with COVID-19 has recovered from COVID-19. In this case, the indication that the patient has an infectious respiratory disease can be formed by the following classification: whether the patient with COVID-19 is in the acute COVID-19 stage or whether the patient with COVID-19 has recovered from COVID-19.

[0068] According to one embodiment of the method of the present invention, the indication regarding a patient having an infectious respiratory disease includes whether the patient has post-acute symptoms following an acute infection with the infectious respiratory disease. To provide an example of such post-acute symptoms, it is known that COVID syndrome (also known as long-term COVID syndrome or chronic COVID syndrome) can persist for a long time after an acute infection with COVID-19, characterized by long-term sequelae that continue beyond the typical recovery period of COVID-19. Therefore, a trained function can be designed in a manner that detects typical signs of such post-acute symptoms in magnetic resonance imaging data. Thus, the method of the present invention can be used to appropriately assess the long-term follow-up examinations of a patient.

[0069] According to one embodiment of the method of the present invention, the output data includes spatially resolved localization data indicating that the patient has an infectious respiratory disease. Therefore, an indication that the patient has an infectious respiratory disease can be formed from the spatially resolved localization data. The trained function can be designed in such a way that it derives the spatially resolved localization data from the spatially resolved magnetic resonance imaging data as input. The spatially resolved localization data can be formed as a two-dimensional or three-dimensional map, particularly a thermal map, providing spatial resolution values ​​indicating that the patient has an infectious respiratory disease. The spatially resolved localization data can have a lower resolution or the same resolution as the magnetic resonance imaging data.

[0070] According to one embodiment of the method of the present invention, the localization data includes an indication of where, in the magnetic resonance imaging (MRI) data, an abnormality indicating that a patient has an infectious respiratory disease exists. In this case, the indication of where the abnormality exists in the MRI data forms spatially resolved localization data indicating that the patient has an infectious respiratory disease. The indication of where the abnormality exists can mark potentially suspicious areas in the MRI data. Therefore, when diagnosing an infectious respiratory disease, the localization data can guide the radiologist when evaluating the MRI data.

[0071] According to one embodiment of the method of the present invention, the localization data includes segmentation of at least one region in magnetic resonance imaging data, where an abnormality indicating that the patient has an infectious respiratory disease exists. Therefore, a quantitative measurement of the spatial dimension indicating infectious respiratory disease can be derived. In the case of longitudinal follow-up studies, the size of the corresponding segmented region where the abnormality exists can be tracked over time. Therefore, the progression of infectious respiratory disease over time can be readily indicated.

[0072] According to one embodiment of the method of the invention, the location data includes a spatially resolved indicator of the severity of an infectious respiratory disease in a patient at a given spatial location. To create such a spatially resolved indicator, a trained function can derive a spatially resolved map of severity scores based on magnetic resonance imaging data as input. Higher tissue deterioration caused by the infectious respiratory disease at a specific spatial location can indicate a higher severity of the infectious respiratory disease in the patient at that specific spatial location. Therefore, useful information about the spatial progression of the infectious respiratory disease can be provided to radiologists, such as which areas of the lungs are most affected by the infectious respiratory disease.

[0073] According to one embodiment of the method of the invention, providing output data includes displaying the location data as an overlay on the magnetic resonance imaging data at the corresponding spatial location. This overlay is advantageously designed as a semi-transparent overlay, allowing the magnetic resonance imaging data to remain visible beneath the location data. Therefore, radiologists can readily and easily assess the morphological magnetic resonance imaging data and the location data indicating that a patient has an infectious respiratory disease.

[0074] According to one embodiment of the method of the present invention, magnetic resonance imaging data is formed as part of input data applied to a trained function to generate output data, wherein the input data further includes additional magnetic resonance imaging data, said additional magnetic resonance imaging data comprising at least one additional organ region of the patient. According to one embodiment of the method of the present invention, the at least one additional organ region comprises at least one body region selected from the group consisting of: the patient's brain region, the patient's vascular region, and the patient's heart.

[0075] Generally, infectious respiratory diseases may affect not only the lungs but also other organs in a patient. For example, recent studies have shown that COVID-19 affects not only the lungs but also other organs such as the brain (see Poyiadji et al., COVID-19–associated Acute Hemorrhagic Necrotizing Encephalopathy: CT and MRI Features, Radiology, 31 March 2020: 201187, doi: 10.1148 / radiol.2020201187). Therefore, considering MRI images of other organ regions in a patient can lead to a more accurate classification of whether the patient is affected by an infectious respiratory disease. This could be beneficial, for example, in high-field MRI where the relaxation time of the lungs is reduced due to rapid signal decoupling.

[0076] According to one embodiment of the method of the present invention, magnetic resonance imaging data forms part of the input data applied to a trained function to generate output data, wherein the method includes receiving temperature mapping data of a patient acquired using a magnetic resonance imaging system, wherein the input data also includes the temperature mapping data. In this way, an indication of whether the patient has an elevated body temperature or even a fever can be directly derived from the magnetic resonance imaging data. Furthermore, the temperature mapping data can indicate the spatial distribution of body temperature within the patient's body. Such temperature mapping data can form suitable additional inputs for use in the trained function to generate an indication that the patient has an infectious respiratory disease.

[0077] According to one embodiment of the method of the present invention, magnetic resonance imaging data forms part of the input data applied to a trained function to generate output data, wherein the method includes receiving patient demographic data and / or patient epidemiological data, wherein the input data further includes demographic data and / or epidemiological data. Demographic data and / or epidemiological data may include at least one or a combination of the following parameters of the patient: age, sex, race, genetic factors, clinical history, history of contact with individuals suffering from infectious respiratory diseases, relevant pre-existing conditions, and other risk factors (e.g., smoking status, BMI, etc.). Such demographic data and / or epidemiological data may form suitable additional inputs for use in the trained function to generate an indication that the patient has an infectious respiratory disease.

[0078] A computer-implemented method of the present invention for providing at least one embodiment of a trained function adapted to determine an indication that a patient has an infectious respiratory disease, the method comprising:

[0079] - Receive input training data, which includes training magnetic resonance imaging (MRI) data acquired using at least one MRI system, wherein the MRI data includes lung regions of the training subject.

[0080] - Receive output training data, wherein the output training data is related to the input training data, and wherein the output training data includes indications regarding the training subject having an infectious respiratory disease.

[0081] - A function is trained based on input and output training data to generate a trained function, wherein the trained function is based on an artificial neural network.

[0082] - Provides trained functions.

[0083] According to one embodiment of a method for providing output data including an indication that a patient has an infectious respiratory disease, the trained function is provided by a method for providing a trained function suitable for determining an indication that a patient has an infectious respiratory disease.

[0084] A providing system of the present invention, comprising at least one embodiment of providing output data including indications that a patient has an infectious respiratory disease, includes:

[0085] - An input interface configured to receive magnetic resonance imaging (MRI) data acquired using a magnetic resonance imaging system, wherein the MRI data includes data on the patient's lung regions.

[0086] - A computational unit configured to apply a trained function to magnetic resonance imaging data, wherein output data is generated.

[0087] The trained function is based on an artificial neural network, and its output data includes indicators that the patient has an infectious respiratory disease.

[0088] - Output interface, which is configured to provide output data.

[0089] The solution according to the invention is generally described with respect to the claimed providing system and the claimed method. Features, advantages, or alternative embodiments described herein can be assigned to other claimed objects, and vice versa. In other words, features described or claimed in the context of the method can be used to improve the claims for the providing system. In this case, the functional features of the method are implemented by the target unit of the providing system.

[0090] A training system of the present invention for providing at least one embodiment of a trained function suitable for determining an indication that a patient has an infectious respiratory disease includes:

[0091] - An input training data interface, configured to receive input training data, which includes training magnetic resonance imaging (MRI) data acquired using at least one MRI system, wherein the MRI data includes lung regions of the training subject.

[0092] - An output training data interface, configured to receive output training data, wherein the input training data is related to the output training data, and wherein the output training data includes indications regarding the training subject having an infectious respiratory disease.

[0093] - A training computation unit, configured to train a function based on input training data and output training data to generate a trained function, wherein the trained function is based on an artificial neural network.

[0094] - Output interface, which is configured to provide trained functions.

[0095] At least one embodiment of the computer program of the present invention includes instructions that, when executed by a computer, cause the computer system to perform an embodiment of the method of the present invention. For this purpose, the computer system must have prerequisites such as a corresponding main memory, a corresponding graphics card, or a corresponding logic unit to enable efficient execution of the corresponding method steps. In at least one embodiment, the computer program is stored, for example, on a computer-readable medium or held on a network or server, and the computer program can be loaded from the computer-readable medium, network, or server into the processor of a local processing unit, which may be directly connected to the computer-readable medium, network, or server or be part of the computer-readable medium, network, or server. Furthermore, control information of the computer program product can be stored on an electronically readable data medium. The control information of the electronically readable data medium can be implemented such that, when the data medium is used in the processing unit, the control information executes at least one embodiment of the method of the present invention. Therefore, the computer program can also represent the electronically readable data medium.

[0096] Furthermore, at least one embodiment of the computer-readable medium includes instructions that, when executed by a computer, cause the computer to perform an embodiment of the method of the present invention. This enables at least one embodiment of the method of the present invention to be performed quickly and robustly in the same repeatable manner. Examples of electronically readable data media are DVDs, magnetic tapes, hard disks, or USB sticks, on which electronically readable control information, particularly software, is stored. When this control information (software) is read from the data medium and stored in a controller and / or processing unit, embodiments of all inventive forms of the previously described method can be performed.

[0097] Furthermore, a computer-readable medium of the present invention according to one embodiment includes a trained function provided by a method for providing a trained function configured to determine an indication that a patient has an infectious respiratory disease. Additionally, the magnetic resonance imaging system of the present invention includes the providing system of the present invention.

[0098] The following is a summary of the implementation methods presented in the claims portion of a U.S. provisional priority application:

[0099] 1. A computer-implemented method for providing output data, said output data including an indication of whether a patient has an infectious respiratory disease, the method comprising the steps of:

[0100] - Receive magnetic resonance imaging (MRI) data acquired using a magnetic resonance imaging system, wherein the MRI data includes lung regions of the patient.

[0101] - The trained function is applied to the magnetic resonance imaging data, wherein output data is generated.

[0102] The trained function is based on an artificial neural network, and the output data includes an indication of whether the patient has an infectious respiratory disease.

[0103] - Provide the output data.

[0104] 2. The computer-implemented method according to Embodiment 1, wherein the magnetic resonance imaging system has a main magnetic field strength of less than 1.0 Tesla.

[0105] 3. The computer-implemented method according to Embodiment 1, wherein the magnetic resonance imaging system has a main magnetic field strength of less than 0.7 Tesla.

[0106] 4. The computer-implemented method according to Embodiment 1, wherein the magnetic resonance imaging system has a main magnetic field strength between 0.5 Tesla and 0.6 Tesla.

[0107] 5. The computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging system has an open system architecture specifically for imaging the chest region of the patient.

[0108] 6. The computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data includes morphological magnetic resonance imaging data acquired using at least one magnetic resonance imaging pulse sequence, the at least one magnetic resonance imaging pulse sequence being selected from the group consisting of: T2-weighted HASTE pulse sequence, T2-weighted BLADE pulse sequence, T2-weighted turbine spin echo pulse sequence, T1-weighted gradient echo pulse sequence, true FISP pulse sequence, pulse sequence for measuring free induction attenuation (FID), and radial VIBE pulse sequence.

[0109] 7. The computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data includes functional magnetic resonance imaging data, and the functional magnetic resonance imaging data includes lung ventilation maps and / or lung perfusion maps.

[0110] 8. The computer-implemented method according to any one of embodiments 6 to 7, wherein the magnetic resonance imaging data includes a combination of the morphological magnetic resonance imaging data and the functional magnetic resonance imaging data, the functional magnetic resonance imaging data including lung ventilation maps and / or lung perfusion maps.

[0111] 9. The computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data includes magnetic resonance imaging data acquired using a combination of a T2-weighted HASTE pulse sequence and a T1-weighted gradient echo sequence.

[0112] 10. The computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data includes magnetic resonance imaging data acquired using a combination of a spiral ultrashort echo time pulse sequence and a T2-weighted TSE pulse sequence.

[0113] 11. The computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data includes magnetic resonance imaging data acquired using a combination of T2-weighted BLADE pulse sequences and radial VIBE pulse sequences.

[0114] 12. The computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data includes magnetic resonance fingerprint data acquired using magnetic resonance fingerprinting technology.

[0115] 13. The computer-implemented method according to embodiment 12, wherein the magnetic resonance fingerprint data includes at least two spatial resolution maps, the at least two spatial resolution maps being selected from the group consisting of: T1 map, T2 map, and PD map.

[0116] 14. The computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data includes contrast agent-based magnetic resonance imaging data acquired by using at least one technique selected from the group consisting of: inhaling an oxygen-based contrast agent by the patient, administering a hyperpolarizing contrast agent to the patient, and administering a fluorine-19 contrast agent to the patient.

[0117] 15. The computer-implemented method according to any one of embodiments 1 to 5, wherein the magnetic resonance imaging data includes magnetic resonance imaging data specifically acquired using a single T2-weighted imaging sequence.

[0118] 16. The computer-implemented method according to any one of embodiments 1 to 5, wherein the magnetic resonance imaging data includes magnetic resonance imaging data specifically acquired using a single-helix ultrashort echo time pulse sequence.

[0119] 17. The computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data is acquired during the patient's free breathing.

[0120] 18. The computer-implemented method according to any one of the foregoing embodiments, wherein the artificial neural network is a convolutional neural network.

[0121] 19. The computer-implemented method according to any one of embodiments 1 to 17, wherein the artificial neural network is a recurrent neural network.

[0122] 20. The computer-implemented method according to any one of the foregoing embodiments, wherein the output data includes a classification of whether the patient has COVID-19 (coronavirus disease 2019).

[0123] 21. The computer-implemented method according to any one of the foregoing embodiments, wherein the output data includes the following classification: whether the patient is healthy or whether the patient has a disease, wherein the disease is selected from the group consisting of: pneumonia, MERS, SARS.

[0124] 22. The computer-implemented method according to any one of the foregoing embodiments, wherein the output data includes the following classification: whether the patient has COVID-19, whether the patient has at least one different infectious respiratory disease, or whether the patient is healthy.

[0125] 23. The computer-implemented method according to embodiment 22, wherein the at least one different respiratory disease includes at least one infectious respiratory disease selected from the group consisting of: pneumonia, MERS, and SARS.

[0126] 24. The computer-implemented method according to any one of the foregoing embodiments, wherein the output data includes the following classification: whether a patient with COVID-19 is in the acute COVID-19 stage or whether a patient with COVID-19 has recovered from COVID-19.

[0127] 25. The computer-implemented method according to any one of the foregoing embodiments, wherein the output data includes location data, the location data including an indication of where an abnormality indicating that the patient has the infectious respiratory disease exists in the magnetic resonance imaging data.

[0128] 26. The computer-implemented method according to embodiment 25, wherein the positioning data includes a segmentation of at least one region in the magnetic resonance imaging data, wherein an abnormality indicating that the patient has the infectious respiratory disease exists in the at least one region.

[0129] 27. The computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data forms a portion of input data to which the trained function is applied to generate the output data, wherein the input data further includes additional magnetic resonance imaging data, the additional magnetic resonance imaging data including at least one additional organ region of the patient.

[0130] 28. The computer-implemented method according to embodiment 27, wherein the at least one additional organ region comprises at least one body region selected from the group consisting of: the patient's brain region, the patient's vascular region, and the patient's heart.

[0131] 29. A computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data forms a portion of input data to which the trained function is applied to generate the output data, wherein the method includes receiving temperature mapping data of a patient acquired using the magnetic resonance imaging system, wherein the input data further includes the temperature mapping data.

[0132] 30. A computer-implemented method according to any one of the foregoing embodiments, wherein the magnetic resonance imaging data forms a portion of input data applied to the trained function to generate the output data, wherein the method includes receiving demographic data and / or epidemiological data of the patient, wherein the input data further includes the demographic data and / or the epidemiological data.

[0133] 31. A computer-implemented method for providing a trained function configured to determine an indication of whether a patient has an infectious respiratory disease, the method comprising the steps of:

[0134] - Receive input training data, said input training data including training magnetic resonance imaging (MRI) data acquired using at least one MRI system, wherein said MRI data includes lung regions of the training subject.

[0135] - Receive output training data, wherein the output training data is related to the input training data, and wherein the output training data includes an indication of whether the training subject has the infectious respiratory disease.

[0136] - A function is trained based on the input training data and the output training data to generate a trained function, wherein the trained function is based on an artificial neural network.

[0137] - Provide the trained function.

[0138] 32. The computer-implemented method according to any one of embodiments 1 to 30, wherein the trained function is provided by the computer-implemented method according to embodiment 31.

[0139] 33. A system for providing output data, the output data including an indication of whether a patient has an infectious respiratory disease, the system comprising:

[0140] - An input interface configured to receive magnetic resonance imaging (MRI) data acquired using a magnetic resonance imaging system, wherein the MRI data includes lung regions of the patient.

[0141] - A computing unit configured to apply a trained function to the magnetic resonance imaging data, wherein output data is generated.

[0142] The trained function is based on an artificial neural network, and the output data includes an indication of whether the patient has the infectious respiratory disease.

[0143] - An output interface configured to provide the output data.

[0144] 34. A training system for providing a trained function adapted to determine an indication of whether a patient has an infectious respiratory disease, the training system comprising:

[0145] - An input training data interface is configured to receive input training data, which includes training magnetic resonance imaging (MRI) data acquired using at least one MRI system, wherein the MRI data includes lung regions of the training subject.

[0146] - An output training data interface, configured to receive output training data, wherein the input training data is related to the output training data, and wherein the output training data includes an indication of whether the training subject has the infectious respiratory disease.

[0147] - A training computation unit configured to train a function based on the input training data and the output training data to generate a trained function, wherein the trained function is based on an artificial neural network.

[0148] - An output interface configured to provide the trained function.

[0149] 35. A computer program including instructions that, when executed by a computer, cause the computer system to perform the method according to any one of embodiments 1 to 32.

[0150] 36. A computer-readable medium including instructions that, when executed by a computer, cause the computer to perform the method according to any one of embodiments 1 to 32.

[0151] 37. A computer-readable medium comprising a trained function provided by the method according to embodiment 32.

[0152] 38. A magnetic resonance imaging system comprising the providing system according to embodiment 33.

[0153] The present invention will now be described and explained in more detail based on the exemplary embodiments shown in the accompanying drawings.

[0154] Figure 1 An implementation of an artificial neural network that can be used as a trained function is shown.

[0155] Figure 2 An implementation of a convolutional neural network that can be used as a trained function is shown.

[0156] Figure 3 An embodiment of the magnetic resonance imaging system of the present invention, having the system provided by the present invention, is shown.

[0157] Figure 4 A first embodiment of a computer-implemented method for providing output data, including indications that a patient has an infectious respiratory disease, is shown.

[0158] Figure 5 A second embodiment of a computer-implemented method for providing output data, including indications that a patient has an infectious respiratory disease, is shown.

[0159] The accompanying drawings are intended to be schematic and the elements shown are not necessarily to scale. Rather, the various elements are shown such that their function and general purpose will become apparent to those skilled in the art. Any connection or coupling between functional blocks, devices, components, or other physical or functional units shown in the drawings or described herein may also be achieved through indirect connections or couplings. Coupling between components may also be established via wireless connections. Functional blocks may be implemented in hardware, firmware, software, or a combination thereof.

[0160] The specific structural and functional details disclosed herein are merely representative and are intended to describe exemplary embodiments of the invention. However, the invention may be practiced in many alternative forms and should not be construed as being limited to the embodiments set forth herein.

[0161] Figure 1An embodiment of the artificial neural network 100 is shown. Alternative terms for "artificial neural network" are "neural network," "artificial neural mesh," or "neural mesh." The artificial neural network 100 can be used as a trained function and applied to magnetic resonance imaging data to generate output data according to an embodiment of the computer-implemented method of the present invention for providing output data including an indication that a patient has an infectious respiratory disease. The artificial neural network 100 can also be a trained function provided by an embodiment of the computer-implemented method of the present invention for providing a trained function adapted to determine an indication that a patient has an infectious respiratory disease.

[0162] It should be noted that Figure 1 The artificial neural network shown is one possible example illustrating an artificial neural network. In practical examples, the number and formation of layers, nodes, and edges of the artificial neural network used in the method of this invention will be different from those in other examples. Figure 1 The diagrams differ from those in the original text. For example, the number of input nodes will typically be higher than... Figure 1 The number of input nodes shown.

[0163] The artificial neural network 100 includes nodes 120, ..., 132 and edges 140, ..., 142, wherein each edge 140, ..., 142 is a directed connection from a first node 120, ..., 132 to a second node 120, ..., 132. Typically, the first nodes 120, ..., 132 and the second nodes 120, ..., 132 are different nodes 120, ..., 132; alternatively, the first nodes 120, ..., 132 and the second nodes 120, ..., 132 may also be the same. For example, in... Figure 1 In the diagram, edge 140 is a directed connection from node 120 to node 123, and edge 142 is a directed connection from node 130 to node 132. Edges 140, ..., 142 from the first node 120, ..., 132 to the second node 120, ..., 132 are also represented as "entry edges" of the second node 120, ..., 132 and "exit edges" of the first node 120, ..., 132.

[0164] In this embodiment, the nodes 120, ..., 132 of the artificial neural network 100 can be arranged in layers 110, ..., 113, wherein these layers may include an inherent order introduced by edges 140, ..., 142 between nodes 120, ..., 132. Specifically, edges 140, ..., 142 may exist only between neighboring layers of a node. In the shown embodiment, there is an input layer 110 that includes only nodes 120, ..., 122 without any entering edges, an output layer 113 that includes only nodes 131, 132 without any leaving edges, and hidden layers 111, 112 located between the input layer 110 and the output layer 113. Typically, the number of hidden layers 111, 112 can be arbitrarily chosen. The number of nodes 120, ..., 122 in the input layer 110 is typically related to the number of input values ​​of the neural network, and the number of nodes 131, 132 in the output layer 113 is typically related to the number of output values ​​of the neural network.

[0165] Regarding the processing of input data for a computer-implemented method according to the present invention for providing output data including indications that a patient has an infectious respiratory disease, magnetic resonance imaging data may be provided to the input layer 110 of an artificial neural network 100. For example, nodes 120, ..., 122 of the input layer 110 may include at least one of the following:

[0166] - Spatial-resolution image values ​​of magnetic resonance imaging data (especially morphological magnetic resonance imaging data).

[0167] - Spatial-resolution image values ​​from functional magnetic resonance imaging (fMRI) data.

[0168] - Spatial-resolution image values ​​from a combination of morphological magnetic resonance imaging (MRI) data and functional magnetic resonance imaging (fMRI) data.

[0169] - Quantitative values ​​contained in magnetic resonance imaging data, such as T1 and / or T2 values, especially spatially resolved quantitative values.

[0170] Output layer 113 then includes output data, namely, an indication that the patient has an infectious respiratory disease.

[0171] Output layer 113 may include individual output nodes 131 and 132, each including a binary classification indicating whether a patient has a respiratory illness such as COVID-19. Individual output nodes 131 and 132 may also include more than one discrete output classification, such as whether a patient is healthy or whether a patient has a disease selected from the group consisting of pneumonia, MERS, and SARS. Another possible discrete output classification may be whether a patient has COVID-19, whether a patient has at least one different infectious respiratory illness, or whether a patient is healthy, wherein at least one different respiratory illness may include at least one infectious respiratory illness selected from the group consisting of pneumonia, MERS, and SARS. Another possible discrete output classification may be whether a patient with COVID-19 is in the acute COVID-19 stage or whether a patient with COVID-19 has recovered from COVID-19. Of course, discrete output classifications may also be distributed across several output nodes 131 and 132, each including a binary classification. Indications regarding a patient having an infectious respiratory illness may also include indications of whether the patient has post-acute symptoms following an acute infection with the infectious respiratory illness.

[0172] Output layer 113 may also include output nodes 131, 132, particularly corresponding to spatially resolved location data indicating that a patient has an infectious respiratory disease. In this case, the number of such output nodes 131, 132 may be less than or equal to the number of nodes 120, ..., 122 of input layer 110. In this case, the location data may include an indication of which location in the magnetic resonance imaging data contains an abnormality indicating that the patient has an infectious respiratory disease. Alternatively or additionally, the location data may include a segmentation of at least one region in the magnetic resonance imaging data in which an abnormality indicating that the patient has an infectious respiratory disease exists. Alternatively or additionally, the location data may include a spatially resolved indication of the severity of the infectious respiratory disease in the patient at the corresponding spatial location.

[0173] Each of these output nodes 131, 132 may, for example, include a binary classification, i.e., whether an abnormality indicating a patient has an infectious respiratory disease exists at a corresponding spatial location. Output nodes 131, 132 may also correspond to a segmentation mask comprising a segmentation of at least one region in the magnetic resonance imaging data where an abnormality indicating a patient has an infectious respiratory disease exists. Some output nodes 131, 132 may also have more than one discrete or continuous value, which includes a spatially resolved indication of the severity of the infectious respiratory disease in the patient at the corresponding spatial location.

[0174] Specifically, (real) numbers can be assigned as values ​​to each node 120, ..., 132 of the neural network 100. Here, x (n) i This represents the value of the i-th node 120, ..., 132 in the nth layer 110, ..., 113. The values ​​of nodes 120, ..., 122 in the input layer 110 are equivalent to the input values ​​of the neural network 100, and the values ​​of nodes 131, 132 in the output layer 113 are equivalent to the output values ​​of the neural network 100. Furthermore, each edge 140, ..., 142 may include a weight as a real number; specifically, the weight is a real number within the interval [-1, 1] or the interval [0, 1]. Here, w... (m,n) i,j Let w represent the weight of the edge between the i-th node (120, ..., 132) of layer m (110, ..., 113) and the j-th node (120, ..., 132) of layer n (110, ..., 113). Furthermore, for the weight w... (n,n+1) i,j Define the abbreviation w (n) i,j .

[0175] Specifically, in order to calculate the output value of neural network 100, the input value is propagated through the neural network. Specifically, the values ​​of nodes 120, ..., 132 in the (n+1)th layer 110, ..., 113 can be calculated based on the values ​​of nodes 120, ..., 132 in the nth layer 110, ..., 113 using the following formula:

[0176]

[0177] In this paper, the function f is the transfer function (another term is "activation function"). Known transfer functions are step functions, sigmoid functions (e.g., logistic functions, generalized logistic functions, hyperbolic tangent functions, arctangent functions, error functions, smooth step functions), or modifier functions. Transfer functions are primarily used for normalization purposes.

[0178] Specifically, the value is propagated layer by layer through the neural network, wherein the value of the input layer 110 is given by the input of the neural network 100, wherein the value of the first hidden layer 111 can be calculated based on the value of the input layer 110 of the neural network, wherein the value of the second hidden layer 112 can be calculated based on the value of the first hidden layer 111, and so on.

[0179] To set the value w of the edge (m,n) i,j Training data must be used to train the neural network 100. Specifically, the training data includes training input data and training output data (denoted as t). iFor the training step, neural network 100 is applied to the training input data to generate computational output data. Specifically, the training data and the computational output data include a number of values ​​equal to the number of nodes in the output layer.

[0180] Regarding a computer-implemented method for providing a trained function suitable for determining an indication that a patient has an infectious respiratory disease, the input training data may be formed from training magnetic resonance imaging (MRI) data acquired using at least one MRI system, wherein the MRI data includes lung regions of the training subject. The training output data may include an indication that the training subject has an infectious respiratory disease.

[0181] Specifically, the comparison between the calculated output data and the training data is used to recursively adjust the weights within the neural network 100 (backpropagation algorithm). Specifically, the weights change according to the following formula:

[0182]

[0183] Where γ is the learning rate, and when the (n+1)th layer is not the output layer, it can be based on δ. (n+1) j Number δ (n) j The recursive calculation is as follows:

[0184]

[0185] When the (n+1)th layer is the output layer 113, the number δ can be... (n) j The recursive calculation is as follows:

[0186]

[0187] Where f' is the first derivative of the activation function, y (n+1) j It is the comparison training value of the j-th node of the output layer 113.

[0188] Figure 2 An embodiment of a convolutional neural network 200 is shown. In the illustrated embodiment, the convolutional neural network 200 includes an input layer 210, a convolutional layer 211, a pooling layer 212, a fully connected layer 213, and an output layer 214. Alternatively, the convolutional neural network 200 may include some convolutional layers 211, some pooling layers 212, some fully connected layers 213, and other types of layers. The order of the layers can be arbitrarily chosen; typically, the fully connected layer 213 is used as the last layer before the output layer 214.

[0189] It should be noted that Figure 2The convolutional neural network shown is one possible example illustrating a convolutional neural network. In actual examples, the number and formation of layers, nodes, and edges of the convolutional neural network used in the method of this invention will be different from those in other examples. Figure 2 The diagrams in the diagrams differ. For example, the construction of the output layer will often differ depending on the type of output required by the actual application. Figure 2 The structure is shown. If, for example, a convolutional neural network desires segmentation of magnetic resonance imaging (MRI) data, the output layer will be formed by a pixel matrix indicating the segmented MRI data. Typically, the input layer 210 and the output 214 can be configured with... Figure 1 It is formed in the same way as the neural network described in the text.

[0190] Specifically, within the convolutional neural network 200, the nodes 220, ..., 224 of a layer 210, ..., 214 can be considered as arranged as a d-dimensional matrix or a d-dimensional image. In particular, in the two-dimensional case, the values ​​of nodes 220, ..., 224 in the nth layer 210, ..., 214, indexed by i and j, can be represented as x. (n) [i,j]. However, the arrangement of nodes 220, ..., 224 in a layer 210, ..., 214 has no effect on the computation performed within the convolutional neural network 200, because these are given only by the structure and weights of the edges.

[0191] Specifically, convolutional layer 211 is characterized in that the structure and weights of the incoming edges form a convolution operation based on a certain number of kernels. Specifically, the structure and weights of the incoming edges are chosen such that the value x of node 220 based on the previous layer 210 is... (n-1) x is the value of node 221 of convolutional layer 211. (n) k Calculated as convolution x (n) k =K k *x (n-1) In the two-dimensional case, convolution is defined as:

[0192]

[0193] Here, the k-th core K kIt is a d-dimensional matrix (a two-dimensional matrix in this embodiment), which is typically small compared to the number of nodes 220, ..., 224 (e.g., a 3×3 or 5×5 matrix). Specifically, this means that the weights of the incoming edges are not independent, but are chosen such that they produce the convolution equation. Specifically, for a 3×3 kernel, regardless of the number of nodes 220, ..., 224 in the corresponding layers 210, ..., 214, there are only 9 independent weights (each entry in the kernel matrix corresponds to one independent weight). Specifically, for convolutional layer 211, the number of nodes 221 in the convolutional layer is equal to the number of nodes 220 in the previous layer 210 multiplied by the number of kernels.

[0194] If the nodes 220 of the previous layer 210 are arranged as a d-dimensional matrix, then using multiple kernels can be interpreted as adding an additional dimension (represented as the "depth" dimension) so that the nodes 221 of the convolutional layer 221 are arranged as a (d+1)-dimensional matrix. If the nodes 220 of the previous layer 210 have already been arranged as a (d+1)-dimensional matrix including the depth dimension, then using multiple kernels can be interpreted as extending along the depth dimension so that the nodes 221 of the convolutional layer 221 are also arranged as a (d+1)-dimensional matrix, where the size of the (d+1)-dimensional matrix relative to the depth dimension is a larger factor than the number of kernels in the previous layer 210.

[0195] The advantage of using convolutional layers 211 is that the spatial local correlation of the input data can be utilized by implementing local connection patterns between nodes of neighboring layers—in particular by connecting each node to a small region of nodes of the previous layer only.

[0196] In the illustrated embodiment, the input layer 210 includes 36 nodes 220 arranged as a two-dimensional 6×6 matrix. The convolutional layer 211 includes 72 nodes 221 arranged as two two-dimensional 6×6 matrices, each of which is the result of convolution of the input layer values ​​with a kernel. Equivalently, the nodes 221 of the convolutional layer 211 can be interpreted as arranged as a three-dimensional 6×6×2 matrix, where the last dimension is the depth dimension.

[0197] Pooling layer 212 can be characterized by the structure and weights of the incoming edges and the activation functions of nodes 222 of pooling layer 212, which form pooling operations based on a nonlinear pooling function f. For example, in the two-dimensional case, it can be based on the values ​​x of nodes 221 of the previous layer 211. (n - 1) x the value of node 222 in pooling layer 212 (n) The calculation is as follows:

[0198] x (n) [i, j] = f(x) (n-1) [id1, jd2], -, x (n-1)[id1+d1-1, jd2+d2-1])

[0199] In other words, by using pooling layer 212, the number of nodes 221, 222 can be reduced by using a single node 222 calculated based on the values ​​of the number of neighboring nodes 221 in the previous layer 211 instead of the number d1·d2 of neighboring nodes 221. Specifically, the pooling function f can be a maximum function, an average function, or an L2 norm function. Specifically, for pooling layer 212, the weights of the incoming edges are fixed and are not changed during training.

[0200] The advantage of using pooling layer 212 is that it reduces the number of nodes 221 and 222 and the number of parameters. This leads to a reduction in the amount of computation in the network and helps control overfitting.

[0201] In the illustrated embodiment, pooling layer 212 uses max pooling with only one node instead of the maximum of the four neighboring nodes, where the value is the largest of the four neighboring nodes. Max pooling is applied to each d-dimensional matrix of the previous layer; in this embodiment, max pooling is applied to each of the two two-dimensional matrices, thereby reducing the number of nodes from 72 to 18.

[0202] The fully connected layer 213 can be characterized by the fact that there are most (in particular all) edges between nodes 222 of the previous layer 212 and nodes 223 of the fully connected layer 213, and that the weight of each edge can be adjusted individually.

[0203] In this embodiment, the nodes 222 of the preceding layer 212 of the fully connected layer 213 are displayed both as a two-dimensional matrix and as unrelated nodes (represented as lines of nodes, where the number of nodes is reduced for better presentation). In this embodiment, the number of nodes 223 in the fully connected layer 213 is equal to the number of nodes 222 in the preceding layer 212. Alternatively, the number of nodes 222 and 223 may be different.

[0204] Furthermore, in this embodiment, the value of node 224 in the output layer 214 is determined by applying the Softmax function to the values ​​of node 223 in the previous layer 213. By applying the Softmax function, the sum of the values ​​of all nodes 224 in the output layer is 1, and all values ​​of all nodes 224 in the output layer are real numbers between 0 and 1. Specifically, if the input data is classified using the convolutional neural network 200, the values ​​of the output layer can be interpreted as the probability that the input data falls into one of the different categories.

[0205] Convolutional neural networks 200 may also include ReLU (an acronym for "Modified Linear Unit") layers. Specifically, the number and structure of nodes contained in the ReLU layer are equal to the number and structure of nodes contained in the previous layer. In particular, the value of each node in the ReLU layer is computed by applying a correction function to the values ​​of the corresponding nodes in the previous layer. Examples of correction functions are f(x) = max(0,x), the tangent hyperbolic function, or the sigmoid function.

[0206] Specifically, a convolutional neural network 200 can be trained based on the backpropagation algorithm. To prevent overfitting, regularization methods can be used, such as the loss of nodes 220, ..., 224, random pooling, the use of artificial data, weight decay based on L1 or L2 norm, or maximum norm constraints.

[0207] Figure 3 An embodiment of the magnetic resonance imaging system 11 of the present invention, having the providing system 27 of the present invention, is shown.

[0208] The magnetic resonance imaging system 11 includes a detector unit formed by a magnet unit 13 having a main magnet 17 for generating a strong and, in particular, constant main magnetic field 18. Furthermore, the magnetic resonance imaging system 11 has a cylindrical patient receiving area 14 for receiving a patient 15, wherein the patient receiving area 14 is cylindrically surrounded in the circumferential direction by the magnet unit 13. The patient 15 can be pushed into the patient receiving area 14 via a patient support facility 16 of the magnetic resonance imaging system 11. For this purpose, the patient support facility 16 has a platform for the patient to lie on, which is movably arranged inside the magnetic resonance imaging system 11. The magnet unit 13 is shielded externally via a housing cladding 31 of the magnetic resonance imaging system.

[0209] The magnet unit 13 also includes a gradient coil unit 19 for generating a magnetic field gradient, which is used for spatial encoding during the imaging process. The gradient coil unit 19 is controlled via a gradient control unit 28. Furthermore, the magnet unit 13 includes: a radio frequency antenna unit 20, which is implemented as a body coil permanently integrated into the magnetic resonance imaging system 10 in the illustrated case; and a radio frequency antenna control unit 29 for exciting polarization, which is generated in the main magnetic field 18 created by the main magnet 17. The radio frequency antenna unit 20 is controlled by the radio frequency antenna control unit 29 and radiates the radio frequency magnetic resonance sequence into the examination space substantially formed by the patient receiving area 14. Furthermore, the radio frequency antenna unit 20 is implemented for receiving magnetic resonance signals, particularly from the patient 15.

[0210] To control the main magnet 17, gradient control unit 28, and radio frequency antenna control unit 29, the magnetic resonance imaging system 11 has a control unit 24. The control unit 24 centrally controls the magnetic resonance imaging system 11, such as executing predetermined gradient echo sequences. Control information such as imaging parameters and reconstructed magnetic resonance images can be provided to the user on the providing unit 25 (in this case, the display unit 25) of the magnetic resonance imaging system 11. Furthermore, the magnetic resonance imaging system 11 has an input unit 26 through which the user can input information and / or parameters during the measurement process. The control unit 24 may include the gradient control unit 28 and / or the radio frequency antenna control unit 29 and / or the display unit 25 and / or the input unit 26.

[0211] Furthermore, the magnetic resonance imaging system 11 includes a recording unit 32. In the present case, the recording unit 32 is formed together by the magnet unit 13, the radio frequency antenna control unit 29, and the gradient control unit 28.

[0212] The magnetic resonance imaging system 11 shown may, of course, include other components typically found in magnetic resonance imaging systems. Furthermore, the general manner in which the magnetic resonance imaging system 11 operates is known to those skilled in the art, and therefore a more detailed description of the other components will be omitted herein.

[0213] also, Figure 3 A providing system 27 for providing output data, including indications that a patient has an infectious respiratory disease, is shown. The providing system 27 includes an input interface 33, a computing unit 34, and an output interface 36. In this way, the providing system 27 is implemented to perform, for example, according to... Figures 4 to 5 The method of the present invention.

[0214] To implement the method of the invention independently, the providing system 27 advantageously loads magnetic resonance imaging (MRI) data from a database via the input interface 33. When the method of the invention is implemented in combination with the MRI system 11 and the providing system 27, the input interface 33 of the providing system 27 will specifically receive MRI data already recorded via the recording unit 32 of the MRI system 11. For this purpose, with regard to data exchange, the providing system 27 (particularly the input interface 33) is advantageously connected to the control unit 24 of the MRI system 11. When the method of the invention is implemented in combination with the MRI system 11 and the providing system 27, output data can be provided on the providing unit 25 of the MRI system 11.

[0215] Figure 4 A first embodiment of a computer-implemented method for providing output data including indications that a patient has an infectious respiratory disease is shown.

[0216] In the first method step 401, magnetic resonance imaging data acquired using a magnetic resonance imaging system is received, wherein the magnetic resonance imaging data includes the patient's lung region.

[0217] For example, a magnetic resonance imaging (MRI) system may have a main magnetic field strength of less than 1.0 Tesla, less than 0.7 Tesla, or between 0.5 Tesla and 0.6 Tesla. Furthermore, the MRI system may have a system architecture specifically designed for imaging the patient's chest region.

[0218] In another method step 402, a trained function is applied to the magnetic resonance imaging data, wherein output data is generated, wherein the trained function is based on an artificial neural network, and the output data includes indications that the patient has an infectious respiratory disease.

[0219] For example, artificial neural networks are convolutional neural networks or recurrent neural networks.

[0220] In another method step 403, output data is provided.

[0221] Figure 5 A second embodiment of a computer-implemented method for providing output data including indications that a patient has an infectious respiratory disease is shown.

[0222] The descriptions given below are basically limited to those related to Figure 4 The differences in the example implementations are explained below. For details regarding maintaining the same method steps, please refer to... Figure 4 The example implementations are described below. The same reference numerals are used to denote substantially the same method steps.

[0223] Figure 5 The embodiments of the method of the present invention shown herein substantially include those based on... Figure 4 Method steps 401, 402, and 403 of a first form of the method of the present invention. Furthermore... Figure 5 The embodiments of the method of the present invention shown may include additional method steps and sub-steps. Further embodiments are also conceivable. Figure 5 An alternative execution sequence for the method, which has only Figure 5 Some additional method steps and / or sub-steps are shown. Of course, for... Figure 5 Alternative execution sequences may also have additional method steps and / or sub-steps.

[0224] The reception of magnetic resonance imaging data in step 401 includes sub-step 501, in which morphological magnetic resonance imaging data acquired using at least one magnetic resonance imaging pulse sequence is received. The at least one magnetic resonance imaging pulse sequence is selected from the group consisting of: T2-weighted HASTE pulse sequence, T2-weighted BLADE pulse sequence, T2-weighted turbine spin echo pulse sequence, T1-weighted gradient echo pulse sequence, true FISP pulse sequence, pulse sequence for measuring free induction attenuation (FID), radial VIBE pulse sequence, helical VIBE pulse sequence, GRASE pulse sequence, and radial TSE pulse sequence.

[0225] The receipt of magnetic resonance imaging data in step 401 includes sub-step 502, in which functional magnetic resonance imaging (fMRI) data, including lung ventilation maps and / or lung perfusion maps, are received. The fMRI data may include lung ventilation maps and / or lung perfusion maps obtained based on Fourier decomposition methods applied to the time series of dynamic magnetic resonance data.

[0226] In particular, magnetic resonance imaging data may include a combination of morphological magnetic resonance imaging data and functional magnetic resonance imaging data, including lung ventilation maps and / or lung perfusion maps.

[0227] Specifically, magnetic resonance imaging (MRI) data may include MRI data acquired using a combination of T2-weighted HASTE pulse sequences and T1-weighted gradient echo sequences. Specifically, MRI data may include MRI data acquired using a combination of helical or 3D radial ultrashort echo time-lapse pulse sequences and T2-weighted or PD-weighted TSE pulse sequences. Specifically, MRI data may include MRI data acquired using a combination of T2-weighted BLADE pulse sequences and radial VIBE pulse sequences. Specifically, MRI data may include MRI fingerprint data acquired using magnetic resonance fingerprinting techniques. In this case, the MRI fingerprint data includes at least two spatially resolved maps selected from the group consisting of: T1 maps, T2 maps, and PD maps. Specifically, MRI data may include contrast-based MRI data acquired using at least one technique selected from the group consisting of: inhalation of an oxygen-based contrast agent by a patient, administration of a hyperpolarizing contrast agent to a patient, and administration of a fluorine-19 contrast agent to a patient. Specifically, the magnetic resonance imaging (MRI) data may include MRI data acquired specifically using a single T2-weighted or PD-weighted imaging sequence. Specifically, the MRI data may include MRI data acquired specifically using a single helical or 3D radial ultrashort echo time-pulse sequence. Specifically, the MRI data may be acquired during the patient's free breathing.

[0228] like Figure 5As shown, the input data 506 of the trained function—magnetic resonance imaging data forming a part of the input data 506—also includes additional magnetic resonance imaging data containing at least one additional organ region of the patient. Such additional magnetic resonance imaging data is received in a further step 503. The at least one additional organ region may specifically include at least one body region selected from the group consisting of: the patient's brain region, the patient's vascular region, and the patient's heart.

[0229] like Figure 5 As shown, the input data 506 of the trained function—magnetic resonance imaging data forming a part of the input data 506—also includes patient temperature mapping data acquired using the magnetic resonance imaging system. Such temperature mapping data is received in a further step 504.

[0230] like Figure 5 As shown, the input data 506 of the trained function—magnetic resonance imaging data forming a part of the input data 506—also includes patient demographics and / or patient epidemiological data. In a further step 505, such demographic and / or epidemiological data is received.

[0231] according to Figure 5 In another method step 403, providing output data includes providing classification and / or location data in sub-step 507. This classification can be for patients with COVID-19 (Coronavirus Disease 2019) or patients without COVID-19. The classification can also be for healthy patients or patients with an illness selected from the group consisting of: pneumonia, MERS, and SARS. The classification can also be for patients with COVID-19, patients with at least one different infectious respiratory illness, or healthy patients. At least one different respiratory illness can include at least one infectious respiratory illness selected from the group consisting of: pneumonia, MERS, and SARS. The classification can also be for patients with COVID-19 in the acute phase or patients who have recovered from COVID-19.

[0232] Spatial-resolved localization data can indicate that a patient has an infectious respiratory disease. The localization data may include indications of the location in the magnetic resonance imaging (MRI) data where an abnormality indicative of an infectious respiratory disease is present. Furthermore, the localization data may include segmentation of at least one region in the MRI data where an abnormality indicative of an infectious respiratory disease is present. The localization data may include a spatially resolved indication of the severity of the infectious respiratory disease at the corresponding spatial location. Providing output data may include displaying the localization data at the corresponding spatial location as an overlay on the MRI data.

[0233] The exemplary embodiments have been described thus; it will be apparent that these exemplary embodiments can be varied in many ways. Such variations should not be considered as departing from the spirit and scope of the invention, and it will be apparent to those skilled in the art that all such modifications are intended to be included within the scope of the appended claims.

Claims

1. A computer-implemented method for providing output data, said output data including indications that a patient has an infectious respiratory disease, the method comprising the steps of: - Receive magnetic resonance imaging data acquired using a magnetic resonance imaging system, wherein the magnetic resonance imaging data includes lung regions of the patient. - The trained function is applied to the magnetic resonance imaging data, wherein output data is generated. The trained function is based on an artificial neural network, and the output data includes indications that the patient has the infectious respiratory disease. - Provide the output data, The indications regarding whether the patient has the infectious respiratory disease include whether the patient has post-acute symptoms following an acute infection with the infectious respiratory disease.

2. The computer-implemented method according to claim 1, wherein, The magnetic resonance imaging system has a main magnetic field strength of less than 1.0 Tesla.

3. The computer-implemented method according to claim 1, wherein, The magnetic resonance imaging system has a main magnetic field strength of less than 0.7 Tesla.

4. The computer-implemented method according to claim 1, wherein, The magnetic resonance imaging system has a main magnetic field strength between 0.5 Tesla and 0.6 Tesla.

5. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging system has a system architecture specifically designed for imaging the patient's chest region.

6. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data includes morphological magnetic resonance imaging data acquired using at least one magnetic resonance imaging pulse sequence selected from the group consisting of: T2-weighted HASTE pulse sequence, T2-weighted BLADE pulse sequence, T2-weighted turbine spin echo pulse sequence, T1-weighted gradient echo pulse sequence, true FISP pulse sequence, pulse sequence for measuring free induction attenuation (FID), radial VIBE pulse sequence, helical VIBE pulse sequence, GRASE pulse sequence, and radial TSE pulse sequence.

7. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data includes functional magnetic resonance imaging data, which includes lung ventilation maps and / or lung perfusion maps.

8. The computer-implemented method according to claim 6, wherein, The magnetic resonance imaging data includes a combination of morphological magnetic resonance imaging data and functional magnetic resonance imaging data, wherein the functional magnetic resonance imaging data includes lung ventilation maps and / or lung perfusion maps.

9. The computer-implemented method according to claim 7, wherein, The functional magnetic resonance imaging data includes lung ventilation maps and / or lung perfusion maps obtained based on Fourier decomposition methods applied to time series of dynamic magnetic resonance data.

10. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data includes magnetic resonance imaging data acquired using a combination of T2-weighted HASTE pulse sequences and T1-weighted gradient echo sequences.

11. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data includes magnetic resonance imaging data acquired using a combination of helical ultrashort echo time pulse sequences or 3D radial ultrashort echo time pulse sequences and T2-weighted TSE pulse sequences or PD-weighted TSE pulse sequences.

12. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data includes magnetic resonance imaging data acquired using a combination of T2-weighted BLADE pulse sequences and radial VIBE pulse sequences.

13. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data includes magnetic resonance fingerprint data acquired using magnetic resonance fingerprinting technology.

14. The computer-implemented method according to claim 13, wherein, The magnetic resonance fingerprint data includes at least two spatial resolution maps, which are selected from the following group: T1 map, T2 map, and PD map.

15. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data includes contrast agent-based magnetic resonance imaging data acquired using at least one technique selected from the group consisting of: inhalation of an oxygen-based contrast agent by the patient, administration of a hyperpolarizing contrast agent to the patient, and administration of a fluorine-19 contrast agent to the patient.

16. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data includes magnetic resonance imaging data acquired specifically using a single T2-weighted imaging sequence or a PD-weighted imaging sequence.

17. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data includes magnetic resonance imaging data acquired specifically using a single helical ultrashort echo time pulse sequence or a 3D radial ultrashort echo time pulse sequence.

18. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data includes magnetic resonance imaging data acquired during the patient's free breathing.

19. The computer-implemented method according to any one of claims 1 to 4, wherein, The artificial neural network is a convolutional neural network.

20. The computer-implemented method according to any one of claims 1 to 4, wherein, The artificial neural network is a recurrent neural network.

21. The computer-implemented method according to any one of claims 1 to 4, wherein, The infectious respiratory disease mentioned is COVID-19 (Coronavirus Disease 2019).

22. The computer-implemented method according to any one of claims 1 to 4, wherein, The output data includes a classification of whether the patient has COVID-19 (coronavirus disease 2019).

23. The computer-implemented method according to any one of claims 1 to 4, wherein, The output data includes the following categories: whether the patient is healthy or whether the patient suffers from the following diseases, which are selected from the following groups: pneumonia, MERS, SARS.

24. The computer-implemented method according to any one of claims 1 to 4, wherein, The output data includes the following categories: whether the patient has COVID-19, whether the patient has at least one different infectious respiratory disease, or whether the patient is healthy.

25. The computer-implemented method according to claim 24, wherein, The at least one different respiratory disease includes at least one infectious respiratory disease selected from the group consisting of: pneumonia, MERS, and SARS.

26. The computer-implemented method according to any one of claims 1 to 4, wherein, The output data includes the following categories: whether a patient with COVID-19 is in the acute COVID-19 stage or whether a patient with COVID-19 has recovered from COVID-19.

27. The computer-implemented method according to any one of claims 1 to 4, wherein, The output data includes spatially resolved location data indicating that the patient has the infectious respiratory disease.

28. The computer-implemented method according to claim 27, wherein, The location data includes indications of where in the magnetic resonance imaging data an abnormality is present that indicates the patient has the infectious respiratory disease.

29. The computer-implemented method according to claim 27, wherein, The location data includes segmentation of at least one region in the magnetic resonance imaging data, where an abnormality indicating that the patient has the infectious respiratory disease exists.

30. The computer-implemented method according to claim 27, wherein, The location data includes a spatially resolved indication of the severity of the infectious respiratory disease suffered by the patient at the corresponding spatial location.

31. The computer-implemented method according to claim 30, wherein, Providing the output data includes displaying the positioning data as an overlay on the magnetic resonance imaging data at the corresponding spatial location.

32. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data forms part of the input data to which the trained function is applied to generate the output data, wherein the input data also includes additional magnetic resonance imaging data, which includes at least one additional organ region of the patient.

33. The computer-implemented method according to claim 32, wherein, The at least one additional organ region includes at least one body region selected from the group consisting of: the patient's brain region, the patient's vascular region, and the patient's heart.

34. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data forms part of the input data to which the trained function is applied to generate the output data, wherein the method includes receiving temperature mapping data of a patient acquired using the magnetic resonance imaging system, wherein the input data further includes the temperature mapping data.

35. The computer-implemented method according to any one of claims 1 to 4, wherein, The magnetic resonance imaging data forms part of the input data to which the trained function is applied to generate the output data, wherein the method includes receiving the patient's demographic data and / or the patient's epidemiological data, wherein the input data further includes the demographic data and / or the epidemiological data.

36. A computer-implemented method for providing a trained function configured to determine an indication that a patient has an infectious respiratory disease, the method comprising the steps of: - Receive input training data, said input training data including training magnetic resonance imaging (MRI) data acquired using at least one MRI system, wherein said MRI data includes lung regions of the training subject. - Receive output training data, wherein the output training data is related to the input training data, and wherein the output training data includes an indication that the training subject has the infectious respiratory disease. - A function is trained based on the input training data and the output training data to generate a trained function, wherein the trained function is based on an artificial neural network. - Provide the trained function, The indications regarding whether the patient has the infectious respiratory disease include whether the patient has post-acute symptoms following an acute infection with the infectious respiratory disease.

37. The computer-implemented method according to claim 36, wherein, The trained function provided by the method implemented by the computer is the trained function according to any one of claims 1 to 35.

38. A system for providing output data, the output data including indications that a patient has an infectious respiratory disease, the system comprising: - An input interface configured to receive magnetic resonance imaging (MRI) data acquired using a magnetic resonance imaging system, wherein the MRI data includes lung regions of the patient. - A computing unit configured to apply a trained function to the magnetic resonance imaging data, wherein output data is generated. The trained function is based on an artificial neural network, and the output data includes indications that the patient has the infectious respiratory disease. - An output interface configured to provide the output data. The indications regarding whether the patient has the infectious respiratory disease include whether the patient has post-acute symptoms following an acute infection with the infectious respiratory disease.

39. A training system for providing a trained function adapted to determine an indication that a patient has an infectious respiratory disease, the training system comprising: - An input training data interface configured to receive input training data, which includes training magnetic resonance imaging (MRI) data acquired using at least one MRI system, wherein the MRI data includes lung regions of the training subject. - An output training data interface configured to receive output training data, wherein the input training data is related to the output training data, and wherein the output training data includes an indication that the training subject has the infectious respiratory disease. - A training computation unit configured to train a function based on the input training data and the output training data to generate a trained function, wherein the trained function is based on an artificial neural network. - An output interface configured to provide the trained function. The indications regarding whether the patient has the infectious respiratory disease include whether the patient has post-acute symptoms following an acute infection with the infectious respiratory disease.

40. A computer program comprising instructions that, when executed by a computer, cause the computer system to perform the method according to any one of claims 1 to 35.

41. A computer-readable medium comprising instructions that, when executed by a computer, cause the computer to perform the method according to any one of claims 1 to 35.

42. A computer-readable medium comprising a trained function provided by the method of claim 36.

43. A magnetic resonance imaging system, the magnetic resonance imaging system comprising the providing system according to claim 38.

Images