Method for providing at least one evaluation characteristic variable of image quality of at least one magnetic resonance image
The calculation of magnetic resonance image quality feature variables through multiple quality modules and machine learning methods is solved, and the problem of time-consuming and subjective impact of image quality evaluation in the prior art is achieved, and rapid and robust image quality evaluation and diagnostic quality improvement are achieved.
Patent Information
- Application Number
- CN202410608198.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-05-17
- Filing Date
- 2024-05-16
- Publication Date
- 2024-11-22
AI Technical Summary
Existing MRI image quality assessment methods are time-consuming and subjectively affected, making it difficult to provide fast and robust assessments under stressful situations, especially in situations where inexperienced medical operators are prone to errors.
Multiple quality modules (such as NRMS, PSNR, DSS, GMSD, etc.) are used to combine machine learning methods to calculate image quality feature variables through evaluation units, providing fast and objective image quality evaluation, and supporting individualized adjustment and automated execution.
It realizes a rapid and robust assessment of image quality during magnetic resonance examination, reduces manual and subjective errors, supports medical operators to obtain direct image quality feedback, and can take timely remedial measures to improve the quality of diagnostic images.
Smart Images

Figure CN119001569A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method for providing at least one evaluation characteristic variable for the image quality of at least one magnetic resonance image. The invention also relates to a system having a magnetic resonance device, a quality unit and an evaluation unit, wherein the system is designed to carry out the method for providing at least one evaluation characteristic variable for the image quality of at least one magnetic resonance image. Background Art
[0002] Regardless of the grammatical gender of a particular term, it includes people with either male or female gender identities.
[0003] Magnetic resonance tomography (MRT) is an efficient imaging method with high soft tissue contrast and without harmful radiation, especially X-ray radiation. However, problems with image quality may occur in MRT images, such as artifacts, which are particularly due to insufficient settings or lack of patient consistency. To date, image quality assurance has been performed manually by medical operators during the measurement time. However, this process of manual image quality assurance is very time-consuming. In addition, this process of manual image quality assurance is influenced by the subjective criteria of the medical operator.
[0004] In the case of magnetic resonance tomography (MRT), image impressions and characteristics vary strongly with different contrasts, anatomical structures, patients, MR equipment and locations. In addition, radiologists have personal preferences for acceptable image quality and their sensitivity relative to the severity of image quality issues. Despite great progress in the field of automated image quality assessment, it remains a major challenge to find sufficiently robust methods to take into account the variability of MR image impressions and at the same time allow for user-specific tolerance adjustments.
[0005] In particular, erroneous quality assessments occur in stressful situations and in the case of inexperienced medical operators. Automated and objective methods for assessing image quality (IQA) that can be used at the scanner can provide great utility in order to ensure the quality of the recorded images. Summary of the invention
[0006] The invention is therefore based on the object, inter alia, of providing a rapid and robust assessment of the image quality during a magnetic resonance examination. This object is achieved by the features of the invention. Advantageous embodiments are described in the following description.
[0007] The invention is based on a method for providing at least one evaluation characteristic variable of the image quality of at least one magnetic resonance image, the method comprising the following steps:
[0008] - providing at least one magnetic resonance image for an assessment of image quality, wherein the provision is made to a quality unit,
[0009] - determining at least two different quality measures of the at least one magnetic resonance image, wherein the quality unit comprises two or more quality modules and each of the two or more quality modules is configured to determine one quality measure of the two or more different quality measures and to provide the two or more quality measures to the evaluation unit,
[0010] - determining at least one evaluation characteristic variable by means of the evaluation unit, wherein the at least one evaluation characteristic variable is determined as a function of two or more different quality measures, and
[0011] - providing at least one evaluation characteristic variable.
[0012] Providing at least one magnetic resonance image preferably comprises detecting at least one magnetic resonance image during a magnetic resonance examination and providing the detected magnetic resonance image to the quality unit. Furthermore, providing may also comprise providing at least one stored magnetic resonance image from a storage unit. The providing is preferably performed by means of a providing unit comprised by a computing unit of the magnetic resonance device.
[0013] The quality unit comprises two or more different quality modules. Preferably, the quality unit comprises a plurality of different quality modules. Each of the different quality modules is designed to determine a quality measure, in particular a unique quality measure. The individual quality modules and thus the different quality measures are hereby independent of each other and hereby respectively assess different image characteristics of at least one magnetic resonance image. For example, the different image characteristics can include noise and / or different noise parameters of at least one magnetic resonance image.
[0014] When determining the corresponding quality measure, the different quality modules can take into account, for example, the following quality characteristics:
[0015] -Normalized root mean square (NRMS), also known as scatter index, including statistical error indicators,
[0016] -Peak signal-to-noise ratio (PSNR) consists of the ratio between the maximum possible power of the signal and the strength of the interfering noise,
[0017] - DCT (discrete cosine transform) subbands similarity (DSS). DSS exploits important features of human visual perception by measuring the variation of structural information in subbands within the DCT range and weighting the quality estimates for said subbands.
[0018] - Gradient magnitude similarity deviation (GMSD). Image gradients are sensitive to image distortion, and different local structures in a distorted image are damaged to different degrees. With GMSD, the image quality index is calculated using pixel-by-pixel gradient magnitude similarity (GMS) combined with the standard deviation.
[0019] -The Haar wavelet-based perceptual similarity index (HaarPSI) is a similarity measure for images that aims to correctly estimate the perceptual similarity of two images to a human observer.
[0020] - The mean deviation similarity index (MDSI) uses the gradient magnitude to measure structural distortion in the image and uses chromaticity features to measure color distortion in the image. These two similarity maps are combined to form a gradient-chromaticity similarity map and the final quality assessment is calculated from this.
[0021] -The mean structural similarity index measure (MSSIM) measures the similarity between two given images.
[0022] - The multi-scale structural similarity index measure (MSSSIM) comprises a more advanced form of SSIM that operates at multiple scales while scanning multiple levels of reduction.
[0023] - Visual Information Fidelity (VIF) includes a complete reference index for evaluating image quality based on natural scene statistics and knowledge of image information extracted from the human visual system.
[0024] - Visual saliency-based index (VSI). Here, visual saliency is used when calculating the local quality map of the distorted image. In addition, visual saliency is also used as a weighting function when summarizing the quality factor in order to reflect the significance of the local area.
[0025] - Deep image structure and texture similarity (DISTS) describes an image quality method that combines correlation in spatial averages ("texture similarity") with correlation in feature maps ("structural similarity").
[0026] - Use learned perceptual image patch similarity (LPIPS) in order to assess the perceptual similarity between two images. LPIPS basically computes the similarity between the activations of two image patches for a predefined network.
[0027] - The perceptual image error metric (PieAPP) measures the perceptual error of the distorted image with respect to the reference and the dataset it belongs to.
[0028] -Total variation (TV) identifies several slightly different concepts involving the structure of the range of a function or measure.
[0029] - Blind referenceless image spatial quality evaluator (BRISQUE) is a model that uses only image pixels to compute features. The model has proven to be extremely efficient, since no transformations are required to compute its features. The model is based on a spatial NSS (Natural Scene Statistics) model of local normalized luminance coefficients in spatial regions, and on a model based on pairwise products for the coefficients.
[0030] - Natural image quality evaluator (NIQE) measures the distance between NSS-based features computed from the image and features obtained from the image database used to train the model. Features are modeled as multi-dimensional Gaussian distributions.
[0031] Preferably, each quality module considers a unique quality characteristic of the quality characteristics for determining the corresponding quality metric. In addition, the quality unit may include other quality modules with other quality characteristics. Each quality metric may include, for example, a number, the value of which represents a metric for quality. For example, the number may include a value between 0 and 1, and the larger the value of the number, the higher the quality sought for the category. If the smaller the number sought, the higher the quality, then the value of 1 minus the number may also be sought to achieve a uniform assessment in the evaluation unit.
[0032] The individual quality modules each determine a quality metric and / or a quality-related image characteristic. The inventors have recognized that three to four quality metrics provided by three to four quality modules are sufficient for a robust assessment of the image quality. However, it is also conceivable that more than four quality metrics provided by more than four quality modules are used for the assessment of the image quality.
[0033] In order to determine each quality metric, each quality module can also include a trained machine learning method. Generally, a trained machine learning method imitates the cognitive function that associates humans with other human thinking. In particular, by training based on training data, the machine learning method can adapt to new situations and recognize patterns and perform extrapolation. Another term for "trained machine learning method" is "trained function" or "trained machine learning model". In general, the parameters of the machine learning method can be adjusted by means of training. In particular, supervised training, partially supervised training, unsupervised training, reinforcement learning and / or active learning can be used. In addition, representation learning (the alternative term is "feature learning") can be used. The parameters of the machine learning method can be adjusted iteratively through multiple training steps. In addition, a back propagation algorithm can be used in the training of a neural network. The machine learning method can include a neural network, a support vector machine, a decision tree and / or a Bayesian network, and / or the machine learning method can be based on k-means clustering, Q-learning, a genetic algorithm and / or an association rule. In particular, the neural network can be a deep neural network, a convolutional neural network or a convolutional deep neural network. Furthermore, the neural network may be an adversarial network, a deep adversarial network, and / or a generative adversarial network.
[0034] Artificial neural network (KNN, English: artificial neural network-ANN) is a network composed of artificial neurons, especially simulated in a computer program. Artificial neural networks are typically based on the networking of multiple artificial neurons. Artificial neurons are typically arranged on different layers. Usually, an artificial neural network includes an input layer and an output layer, and the neuron output of the output layer is visible as the only neuron output of the artificial neural network. The layer between the input layer and the output layer is typically called a hidden layer. Typically, the framework and / or topology of the artificial neural network is first created and then trained in a training phase for a specific purpose or multiple purposes in the training phase. Here, the training of the artificial neural network typically includes changing the weight of the connection between two artificial neurons of the artificial neural network. The training of the artificial neural network can also include developing new connections between artificial neurons, deleting existing connections between artificial neurons, adjusting the threshold of artificial neurons and / or adding or deleting artificial neurons. The artificial neural network has been trained in a manner suitable for determining quality metrics in advance.
[0035] The mass unit can be included in this case by the magnetic resonance device. For example, the mass unit can be included in the computing unit of the magnetic resonance device. In addition, it is also conceivable that the mass unit is included in the cloud. For this purpose, the computing unit and / or the magnetic resonance device can include a data interface, which is designed for exchanging data with the mass unit.
[0036] The determination of at least one evaluation characteristic variable is performed by means of an evaluation unit. At least two quality metrics, preferably at least three or four quality metrics, are used to determine at least one evaluation characteristic variable. In addition to the quality metrics, other information for determining the evaluation characteristic variable may also be considered by the evaluation unit, such as other image information and / or DICOM information and / or device characteristics, in particular magnetic field strength and / or gradient field strength, and / or weights of various quality metrics, etc. In order to determine at least one evaluation characteristic variable, the evaluation unit may also include a trained machine learning method.
[0037] The evaluation unit can be included in this case by the magnetic resonance system. For example, the evaluation unit can be included in this case by a computing unit of the magnetic resonance system. Furthermore, it is also conceivable that the evaluation unit is included by the cloud. For this purpose, the computing unit and / or the magnetic resonance system can include a data interface, which is designed for exchanging data with the evaluation unit.
[0038] The at least one evaluation characteristic variable is preferably provided to a medical operator, such as, for example, a medical technical radiology assistant. In this case, it is preferably provided to a user interface, in particular an output unit of the user interface, such as a monitor and / or a display, by means of an evaluation unit.
[0039] The evaluation characteristic variable can, for example, include a simple classification of the evaluation characteristic values into at least two evaluation categories of the image quality. For example, the evaluation characteristic variable can be shown here as an evaluation traffic light. Here, the color green can symbolize and / or represent a good evaluation characteristic variable, while the color red can symbolize and / or represent a poor evaluation characteristic variable. In addition, the color orange can also be output for an evaluation characteristic variable of a medium degree. Alternatively, only "+" or "good" or "-" or "poor" etc. can also be displayed to the user.
[0040] Preferably, when assessing the image quality, the method for providing at least one assessment characteristic variable for the image quality of at least one magnetic resonance image is performed automatically, ideally when recording the magnetic resonance data during a magnetic resonance examination or, for example, when creating a diagnosis. Preferably, the magnetic resonance system has a computing unit for this purpose for performing the method for providing at least one assessment characteristic variable for the image quality of at least one magnetic resonance image.
[0041] The computing unit comprises at least one computing module and / or a processor, wherein the computing unit is designed to execute the method according to the invention for providing at least one evaluation characteristic variable of the image quality of at least one magnetic resonance image. Thus, the computing unit is especially designed to execute computer-readable instructions in order to execute the method according to the invention. The computing unit especially comprises a storage unit, wherein computer-readable information is stored on the storage unit, wherein the computing unit is designed to download the computer-readable information from the storage unit and execute the computer-readable information in order to execute the method according to the invention for providing at least one evaluation characteristic variable of the image quality of at least one magnetic resonance image.
[0042] The components of the computing unit can be designed for the most part in the form of software components. In principle, however, in particular when it comes to particularly fast calculations, these components can also be partially implemented in the form of software-supported hardware components, such as FPGAs or the like. Likewise, the required interfaces, for example when they only involve receiving data from other software components, can be designed as software interfaces. However, the interfaces can also be designed as hardware-structured interfaces that are controlled by appropriate software. It is of course also conceivable that a plurality of the components mentioned are implemented in combination in the form of a single software component or software-supported hardware component.
[0043] The invention has the advantage that a rapid and robust assessment of the image quality can be provided to a medical operator during a magnetic resonance examination. In this way, the medical operator can receive direct feedback on the image quality, in particular during the magnetic resonance examination. In this way, the medical operator can also be advantageously supported, and manual and / or subjective errors in the assessment of the image quality can be reduced and / or prevented, and an objective assessment of the image quality can be performed. In particular, the individual quality measures can cover different quality issues in the magnetic resonance images, so that the error susceptibility of the method is reduced.
[0044] Furthermore, in the case of insufficient image quality, the medical operator can immediately initiate remedial measures for improving the image quality as a function of the evaluation characteristic variable, for example carrying out a remeasurement and / or initiating reassurance measures for reassuring the patient if the patient moves during the measurement and / or selecting an alternative measurement sequence for remeasurement, etc. In this way, the diagnosis based on the acquired magnetic resonance data, in particular the magnetic resonance images, can also be improved, since magnetic resonance images with high image quality are now available.
[0045] In an advantageous development of the method according to the invention, provision can be made that during a magnetic resonance examination at least one evaluation characteristic variable for providing the image quality of at least one magnetic resonance image is carried out. Preferably, during the acquisition of image data, in particular during the acquisition of magnetic resonance data, a magnetic resonance image is directly reconstructed and at least one evaluation characteristic variable for the image quality is directly ascertained and provided for the magnetic resonance image. This makes it possible to advantageously support, preferably automatically support, a user in the evaluation and / or estimation of the image quality during a magnetic resonance examination.
[0046] In an advantageous development of the method according to the invention, it can be provided that at least one of the two or more quality modules comprises a trained machine learning method, wherein the trained machine learning method is trained to determine a quality measure of the magnetic resonance image to be evaluated by comparison with at least one reference image. Preferably, the reference image comprises an ideal copy of the magnetic resonance image without any quality restrictions.
[0047] The machine learning method has in particular been trained beforehand in a manner suitable for determining a quality metric. In order to train the machine learning method, in particular a training image data set is used, wherein a magnetic resonance image is compared with at least one reference image with respect to the quality metric to be determined, and deviations and / or differences of the magnetic resonance image from the at least one reference image determine the corresponding quality metric. If two or more of the quality modules have a machine learning method, the machine learning method is preferably trained for determining the corresponding quality metric. The medical training data set is typically obtained from different training personnel and / or training patients. Preferably, the medical training data set comprises data sets for different body regions, so that a preferably automatic assessment of the image quality can be performed for magnetic resonance examinations of different body regions.
[0048] This can provide a medical operator with a fast and robust assessment of the image quality during a magnetic resonance examination. In particular, this can advantageously support the medical operator and can reduce and / or prevent manual and / or subjective errors in assessing the image quality.
[0049] In an advantageous development of the method according to the invention, it can be provided that at least one of the two or more quality modules comprises a trained deep learning method, wherein the trained deep learning method is trained to determine a quality measure for the magnetic resonance image to be evaluated without comparison with a reference image.
[0050] Deep learning methods (English: deep learning) include the following machine learning methods: the method uses an artificial neural network with a large number of intermediate layers between the input layer and the output layer, and thus forms a large-scale internal structure. In order to train the deep learning method, a training image data set is used in particular, in which magnetic resonance images with different image qualities are associated with corresponding quality metrics. In addition, the training data of the deep learning method can also include reference images. If two or more quality modules in the quality module have a deep learning method, the deep learning method is preferably trained for determining the corresponding quality metric. In particular, the neural network can be trained (with or without the help of a reference image) to predict the image quality metric at the test time even in the absence of a reference image, which requires a perfect reference image for direct calculation. Therefore, for example, a neural network can be trained to predict MSSIM without the need for a reference image at the test time. This has significant technical and strategic advantages, because reference images are not available in clinical reality, and the method can still achieve the consideration of a large number of quality metrics, which requires a reference image for direct calculation.
[0051] The medical training data set is typically obtained from different training personnel and / or training patients. Preferably, the medical training data set comprises data sets of different body regions, so that a preferably automatic assessment of the image quality can be performed for magnetic resonance examinations of different body regions. The quality measure from the deep learning method can comprise the number of predicted points for training tasks in the field of image recognition and / or image quality.
[0052] This embodiment has the advantage that a robust determination of the quality measure is possible. Deep learning methods are also characterized in that they can be trained with the aid of small training data sets, so that rapid adaptation to changes, such as changes in hardware conditions and / or the selection of another body region, etc., can also be achieved. In addition, this advantageously allows different, in particular individualized, quality issues to be taken into account when determining the corresponding quality measure.
[0053] In an advantageous improvement of the method according to the invention, it can be provided that the evaluation unit includes a machine learning method for determining at least one evaluation characteristic variable from at least two quality measures. The machine learning method can include support vector machines (SVM), support vector regression (SVR), decision trees, linear regression, linear forests, etc. In addition, the machine learning method can also include deep learning methods, such as, for example, convolutional neural networks (CNN) or fully connected networks.
[0054] Preferably, the machine learning method is trained in such a way that an evaluation characteristic variable for evaluating the image quality of at least one magnetic resonance image can be determined from a combination of a plurality of mutually independent quality measures. Furthermore, the machine learning method can also be adapted to the user for training, so that, for example, the user's quality criteria are taken into account when evaluating the image quality of at least one magnetic resonance image. For example, individualized quality criteria, such as noise components in magnetic resonance images, can be defined and / or preset by the user and taken into account in the training phase of the machine learning method. Furthermore, the machine learning method for determining the evaluation characteristic variable can be trained based on the evaluation of an expert, such as a doctor, and thereby provide an optimal consistency between the expert evaluation and the provided quality measure.
[0055] In an advantageous development of the method according to the invention, it can be provided that the evaluation unit uses at least one statistical method to determine at least one evaluation characteristic variable from at least two quality measures. The statistical method can include the arithmetic mean and / or median and / or standard deviation and / or maximum and / or minimum value of the quality measure, etc. In this way, the evaluation characteristic variable can be determined in a simple manner.
[0056] In an advantageous improvement of the method according to the invention, it can be proposed that in order to obtain at least one evaluation characteristic variable of at least one magnetic resonance image, other image parameters and / or DICOM parameters are taken into account. Other image parameters and / or DICOM parameters may include, for example, acquisition time and / or pixel size and / or matrix size and / or image resolution and / or sequence parameters and / or patient sex and / or patient age and / or magnetic field strength and / or gradient field strength, etc. In this way, additional information, especially text-related information, for each quality metric can be present, which is taken into account when obtaining the evaluation characteristic variable from each quality metric. For example, when obtaining the evaluation characteristic variable, the weight of each quality metric can be influenced according to the additional information. Therefore, for example, the noise performance can be weighted according to the pixel size and / or sequence characteristics.
[0057] In an advantageous improvement of the method according to the invention, it can be proposed that each of the obtained quality metrics is weighted in the evaluation unit for obtaining at least one evaluation characteristic variable. In this way, different quality metrics with different weights can be considered. Here, it can also be proposed that the weighting of each quality metric can be at least partially performed by a user, especially a medical operator. Here, for example, when obtaining at least one evaluation characteristic variable, the user can specify the individual weights of each quality metric via a user interface. The following advantages can be achieved: simple and fast, especially individualized adjustments can be achieved when obtaining at least one evaluation characteristic value. In addition, the individual quality metrics provided can also be weighted differently for different sequences, and the quality metrics, for example, have different degrees of consistency with expert evaluation in the case of different sequences. For example, the quality module and / or quality metric with the best consistency with the expert evaluation can also be highly weighted for the corresponding sequence. On the contrary, the quality module and / or quality metric with the worst consistency with the expert evaluation can be lowly weighted for the corresponding sequence.
[0058] In an advantageous improvement of the method according to the invention, it can be proposed that, in order to determine at least one evaluation characteristic variable, two or more quality metrics are compared with at least one input threshold value in the evaluation unit in order to classify the individual quality metrics. The classification of the individual quality metrics can, for example, respectively include two categories, wherein the two categories are divided into "good" and "poor". In addition, the classification can also include more than two categories. Here, for each of the two or more quality metrics, different input threshold values for classifying the individual quality metrics can exist and / or be used. Classifying two or more quality metrics has the following advantages: it is not the absolute value of the quality metric determined, but a normalization and / or evaluation of the quality metric that affects the determination of at least one evaluation parameter. For example, the value (1-MSSIM) given by the inversion (umgedreht) of the quality metric for the module MSSIM cannot be lower than 0.933 for the grade "good", while the quality metric for the module BRISQUE has reached this category at a value of 0.598. This can advantageously prevent the undesired misinterpretation of the individual quality metrics.
[0059] In an advantageous development of the method according to the invention, it can be provided that at least one evaluation characteristic variable comprises a classification of the image quality into at least two quality classes, wherein the quality class is selected according to the evaluation number determined. The evaluation number preferably comprises a number between 0 and 1, the smaller the number, the higher the image quality of the magnetic resonance image data, wherein the evaluation number is determined by the evaluation unit according to at least two quality metrics and other parameters. The at least two quality classes here comprise a quality class with high image quality and a quality class with low and / or insufficient image quality. Preferably, the determined evaluation number is compared with at least one threshold value in order to group and / or classify the evaluation characteristic number. Preferably, the threshold value comprises a boundary between two quality classes.
[0060] The quality class with high image quality can include output elements, in particular visual output elements, which can be output to the user by means of a user interface. The visual output elements of the quality class with high image quality can include "good" and / or "+" and / or a smiling emoticon (Smiley) and / or a green dot, etc.
[0061] The quality class with low and / or insufficient image quality can include an output element, in particular a visual output element, which can be output to a user by means of a user interface. The visual output element of the quality class with low and / or insufficient image quality can include "poor" and / or "-" and / or a crying emoticon and / or a red dot, etc.
[0062] Furthermore, the evaluation characteristic variable may also include a classification of the image quality into three quality classes, wherein the classification provides an additional quality class with medium quality. The quality class with medium image quality may include output elements, in particular visual output elements, which can be output to a user by means of a user interface. The visual output elements of the quality class with medium image quality may include an "average" and / or neutral-looking emoticon and / or an orange or yellow dot, etc.
[0063] The embodiment of the invention has the advantage that a simple and direct check of the quality of the provided magnetic resonance images can be achieved for the user. In particular, evaluation characteristic variables for evaluating the image quality can be output directly during the magnetic resonance measurement.
[0064] In an advantageous improvement of the method according to the invention, it can be provided that the evaluation characteristic variable comprises the evaluation number sought. Here, in addition to displaying the quality class of the evaluation characteristic variable, the evaluation number sought can be output and / or displayed. In addition, by outputting the evaluation number, the user can obtain a detailed assessment of the image quality. For example, the user can recognize from the evaluation number that, although there is a low quality class, the evaluation number is very close to the threshold value of the next higher quality class. This can enable the user to realize individualized grading of the image quality of the magnetic resonance image data.
[0065] In an advantageous development of the method according to the invention, it can be provided that at least one evaluation characteristic variable is output to a user. Preferably, the output is performed visually by means of a display unit of a user interface of the magnetic resonance device, such as a monitor and / or a display. This makes it possible to detect the quality of the provided magnetic resonance images simply and directly. In particular, the evaluation characteristic variable can be output directly to a user, such as a doctor, during the detection of magnetic resonance data, so that the user can react directly to poor image quality.
[0066] Furthermore, the invention is based on a system having a magnetic resonance apparatus, a quality unit and an evaluation unit, wherein the system is designed to carry out a method for providing at least one evaluation characteristic variable of the image quality of at least one magnetic resonance image.
[0067] The magnetic resonance system preferably comprises a medical and / or diagnostic magnetic resonance system which is designed and / or constructed for detecting medical and / or diagnostic image data, in particular medical and / or diagnostic magnetic resonance image data, of a patient. The magnetic resonance system comprises a magnet unit and / or a scanner unit having a basic magnet, a gradient coil unit and a radio frequency antenna unit.
[0068] The basic magnet is designed to generate a uniform basic magnetic field with a defined magnetic field strength, for example with a magnetic field strength of 0.55 T or 1.5 T or 3 T or 7 T, etc. The basic magnet is especially designed to generate a strong, constant and uniform basic magnetic field. The uniform basic magnetic field is preferably arranged and / or present in the patient receiving area of the magnetic resonance device.
[0069] In order to perform a magnetic resonance examination, a patient, in particular a region of the patient to be examined, is positioned in a patient accommodation region of a magnetic resonance device. The patient accommodation region is at least partially surrounded by a magnet unit, in particular cylindrically surrounded by the magnet unit. The field of view (FOV) and isocenter of the magnetic resonance device are preferably arranged in the patient accommodation region. The FOV preferably includes a detection region of the magnetic resonance device, in which conditions exist for detecting medical image data, in particular magnetic resonance image data, in the patient accommodation region, such as a uniform basic magnetic field. The isocenter of the magnetic resonance device preferably includes a region and / or point within the magnetic resonance device, which has optimal and / or ideal conditions for detecting medical image data, in particular magnetic resonance image data. In particular, the isocenter includes the most uniform magnetic field region within the magnetic resonance device.
[0070] Magnetic resonance image data, in particular magnetic resonance images, are provided by means of a magnetic resonance scanner for evaluation of the image quality.
[0071] The quality unit is configured to determine at least two different quality measures of at least one magnetic resonance image. For this purpose, the quality unit comprises two or more quality modules, wherein each of the two or more quality modules is configured to determine one quality measure of the two or more different quality measures. Furthermore, the quality unit is also configured to provide the two or more quality measures to the evaluation unit.
[0072] The evaluation unit is designed to ascertain at least one evaluation characteristic variable. In this case, the at least one evaluation characteristic variable is ascertained as a function of two or more different quality measures.
[0073] The mass unit and / or the evaluation unit can be included in this case by the magnetic resonance device. For example, the mass unit and / or the evaluation unit can be included in this case by the computing unit of the magnetic resonance device. Alternatively, the mass unit and / or the evaluation unit can also be included by the cloud. For this purpose, the computing unit and / or the magnetic resonance device can include a data interface configured for exchanging data with the mass unit and / or the evaluation unit.
[0074] Furthermore, the magnetic resonance system comprises a user interface which is designed to output at least one evaluation characteristic variable.
[0075] The system according to the invention has the advantage that a rapid and robust assessment of the image quality can be provided to a medical operator during a magnetic resonance examination. In this way, the medical operator can in particular receive direct feedback on the image quality during the magnetic resonance examination. In this way, the medical operator can also be advantageously supported and manual and / or subjective errors in the assessment of the image quality can be reduced and / or prevented, and an objective assessment of the image quality can be performed. In particular, the individual quality measures can cover different quality issues in the magnetic resonance images, so that the error susceptibility of the method is reduced.
[0076] The advantages of the system according to the invention correspond substantially to the advantages of the method according to the invention for providing at least one evaluation characteristic variable of the image quality of at least one magnetic resonance image, which have been listed in detail above. The features, advantages or alternative embodiments mentioned here can also be transferred to other claimed subjects, and vice versa.
[0077] The invention is also based on a computer program product, which includes a program and can be directly loaded into a memory of a programmable computing unit, and has program means in order to execute a method for providing at least one evaluation characteristic variable of the image quality of at least one magnetic resonance image when the program is run in the computing unit. In this case, the computer program may require program means, such as program libraries and auxiliary functions, in order to implement the corresponding embodiments of the method. In this case, the computer program may include: software with source code that still needs to be compiled and linked or only needs to be interpreted; or executable software code that only needs to be loaded into the corresponding computing unit for execution.
[0078] The computer program product according to the present invention can be directly loaded into the memory of a programmable computing unit and has a program code mechanism so that when the computer program product is run in the computing unit, the method according to the present invention is executed. The computer program product can be a computer program or include a computer program. Thus, the method according to the present invention can be executed quickly, identically and repeatedly and robustly. The computer program product is configured so that it can execute the method steps according to the present invention by means of a computing unit. The computing unit must have prerequisites, such as corresponding working memory, corresponding graphics card or corresponding logic unit, so that the corresponding method steps can be executed efficiently. The computer program product is stored on a computer-readable medium or saved on a network or server, for example, from which the computer program product can be loaded into the processor of a local computing unit, which is directly connected to the magnetic resonance device or can be formed as a part. In addition, the control information of the computer program product can be stored on an electronically readable data carrier. The control information of the electronically readable data carrier can be designed so that the control information executes the method according to the present invention when the data carrier is used in the computing unit. Therefore, the computer program product can also display the electronically readable data carrier. Examples of electronically readable data carriers are DVDs, magnetic tapes, hard disks or USB sticks on which electronically readable control information, in particular software (see above) is stored. If the control information (software) is read from the data carrier and stored in a control device and / or a computing unit, all embodiments of the method described above according to the invention can be performed. The invention can therefore also be based on the computer-readable medium and / or the electronically readable data carrier. BRIEF DESCRIPTION OF THE DRAWINGS
[0079] Further advantages, features and details of the invention result from the exemplary embodiments described below and with reference to the drawings.
[0080] The accompanying drawings show:
[0081] Figure 1 A schematic diagram showing a magnetic resonance apparatus according to the present invention,
[0082] Figure 2 The method according to the invention is shown, and
[0083] Figure 3 A flow chart of the method is shown. DETAILED DESCRIPTION
[0084] exist Figure 1Schematically shown in FIG. 1 is a magnetic resonance device 10. The magnetic resonance device 10 comprises a magnet unit 11 having a base magnet 12, a gradient coil unit 13 and a radio frequency antenna unit 14. In addition, the magnetic resonance device 10 has a patient receiving area 15 for receiving a patient 16. In the present embodiment, the patient receiving area 15 is cylindrically designed and is cylindrically surrounded by the magnet unit 11 in the circumferential direction. However, in principle, different designs of the patient receiving area 15 are conceivable at any time.
[0085] In order to position the patient 16, in particular the area to be examined of the patient 16, in the patient receiving area 15, the magnetic resonance device 10 has a patient support device 17. The patient support device 17 has a movable examination bed 18. The examination bed 18 is designed to movably position the patient 16, in particular the area to be examined of the patient 16, in the patient receiving area 15.
[0086] The basic magnet 12 of the magnet unit 11 is designed to generate a strong and especially constant basic magnetic field 19. Here, the basic magnet 12 can be designed, for example, as a superconducting basic magnet 12 or as a permanent magnet. The gradient coil unit 13 of the magnet unit 11 is designed to generate a magnetic field gradient for position encoding during imaging. The gradient coil unit 13 is controlled by means of a gradient control unit 20 of the magnetic resonance device 10. The radio frequency antenna unit 14 of the magnet unit 10 is designed to excite the polarization occurring in the basic magnetic field 19 generated by the basic magnet 12. The radio frequency antenna unit 14 is controlled by a radio frequency antenna control unit 21 of the magnetic resonance device 10 and emits a radio frequency magnetic resonance sequence into the patient accommodation area 15 of the magnetic resonance device 10.
[0087] The magnetic resonance system 10 has a system control unit 22 for controlling the base magnet 12, the gradient control unit 20 and for controlling the radio frequency antenna control unit 21. The system control unit 22 centrally controls the magnetic resonance system 10, such as, for example, the execution of a predetermined imaging gradient echo sequence. In addition, the system control unit 22 includes an evaluation unit (not shown in detail) for evaluating the medical image data acquired during the magnetic resonance examination.
[0088] The magnetic resonance system 10 further comprises a user interface 23, which is connected to the system control unit 22. Control information, such as imaging parameters, and reconstructed magnetic resonance images can be displayed to a medical operator on a display unit and / or output unit 24, such as at least one monitor, of the user interface 23. The user interface 23 further comprises an input unit 25, by means of which information and / or parameters can be input by the medical operator during the measurement process.
[0089] In the present exemplary embodiment, the magnetic resonance system 10 has a computing unit 26, which includes a mass unit 27 and an evaluation unit 28. In an alternative embodiment of the magnetic resonance system 10, the computing unit 26, in particular the mass unit 27 and the evaluation unit 28 can also be designed separately from the magnetic resonance system 10. In this case, the computing unit 26, in particular the mass unit 27 and the evaluation unit 28 can also be arranged in the cloud.
[0090] The magnetic resonance system 10 shown may of course include other components that are usually present in a magnetic resonance system 10. Otherwise, the general mode of operation of the magnetic resonance system 10 is known to a person skilled in the art, so that a detailed description of the other components is omitted.
[0091] exist Figure 2 and Figure 3 2 shows a method according to the invention for providing at least one evaluation characteristic variable 29 for the image quality of at least one magnetic resonance image 30. The method is here automatically performed by a computing unit 26 which has the necessary software and / or computer program for this purpose.
[0092] In a first method step 100 , at least one magnetic resonance image 30 is provided. Preferably, the magnetic resonance image 30 acquired during the magnetic resonance examination is provided directly for providing at least one evaluation characteristic variable. In this case, it is provided to the quality unit 27 by means of the calculation unit 26 .
[0093] In a subsequent second method step 101, at least two different quality measures of at least one magnetic resonance image 30 are determined by means of the quality unit 27. For this purpose, the quality unit 27 comprises two or more quality modules 31 for determining two or more different quality measures. The two or more quality modules 31 are designed to be independent of one another. The two or more quality measures are also independent of one another. Preferably, three or four quality measures are determined by means of three or four quality modules 31 in order to provide at least one evaluation characteristic variable.
[0094] For example, different quality modules 31 may include different standard image quality characteristics. When determining the corresponding quality metric, the quality module 31 may take into account the following quality characteristics:
[0095] -Normalized root mean square (Peak signal-to-noise ratio - NRMS)
[0096] -Peak signal-to-noise ratio (PSNR)
[0097] -DCT (discrete cosine transform) subband similarity (DSS)
[0098] - Gradient Magnitude Similarity Deviation (GMSD)
[0099] -Haar wavelet-based perceptual similarity index (HaarPSI)
[0100] -Mean Deviation Similarity Index (MDSI)
[0101] - Mean Structural Similarity Index Measure (MSSIM)
[0102] -Multi-Scale Structural Similarity Index Measure (MSSSIM)
[0103] -Visual Information Fidelity (VIF)
[0104] - Based on Visual Saliency Index (VSI)
[0105] - Deep Image Structure and Texture Similarity (DISTS)
[0106] -Learning Perceptual Image Patch Similarity (LPIPS)
[0107] -Perceptual Image Error Metric (PieAPP)
[0108] -Total variation (TV)
[0109] - Reference-free Image Spatial Quality Estimator (BRISQUE)
[0110] - Natural Image Quality Estimator (NIQE).
[0111] Furthermore, the two or more quality modules 31 can each include a trained machine learning method, wherein the respective trained machine learning method is trained to determine a quality measure of the magnetic resonance image 30 to be evaluated by comparison with at least one reference image. The machine learning method has in particular been trained in advance in a manner suitable for determining the quality measure. In order to train the machine learning method for predicting / ascertaining a quality measure defined based on a reference image, in particular a training image data set is used, wherein the magnetic resonance image is compared with at least one reference image with respect to the quality measure to be determined, and deviations and / or differences of the magnetic resonance image from the at least one reference image determine the corresponding quality measure.
[0112] Alternatively, two or more quality modules 31 may also include trained deep learning methods, wherein each trained deep learning method is trained to determine a quality measure of a magnetic resonance image to be evaluated without comparison with a reference image. In order to train the deep learning method, a training image data set is used in particular, wherein magnetic resonance images with different image qualities are associated with different quality measures. In addition, the training data of the deep learning method may also include reference images. In this case, the medical training data set preferably includes data sets of different body regions, so that a preferably automatic evaluation of the image quality can be performed for different body regions. The quality measure from the deep learning method may include the number of predicted points for training for tasks in image recognition and / or image quality.
[0113] Furthermore, in the second method step 101 , two or more quality measures are provided to the evaluation unit 28 by means of the quality unit 27 .
[0114] In a subsequent third method step 102 , at least one evaluation characteristic variable 29 is ascertained by means of the evaluation unit 28 , wherein the at least one evaluation characteristic variable 29 is ascertained as a function of two or more different quality measures.
[0115] In order to determine at least one evaluation characteristic variable 29 from at least two different quality measures, the evaluation unit 28 preferably includes a machine learning method. The machine learning method may include a support vector machine (SVM), a support vector regression (SVR), a decision tree, a linear regression, a linear forest, etc. In addition, the machine learning method may also include a deep learning method, such as, for example, a convolutional neural network (CNN) or a fully connected network.
[0116] The machine learning method can be trained in this case so that from a combination of a plurality of mutually independent quality measures, an evaluation characteristic variable 29 can be determined for evaluating the image quality of the at least one magnetic resonance image 30. Furthermore, the machine learning method can also be adapted to the user for training, in order to take into account the user's quality criteria, for example, when evaluating the image quality of the at least one magnetic resonance image 30. Furthermore, the machine learning method for determining the evaluation characteristic variable can be trained based on the evaluation of an expert, for example a physician, so that an optimal agreement between the expert evaluation and the provided quality measure is provided.
[0117] Furthermore, the evaluation unit 28 may include and / or use at least one statistical method for determining at least one evaluation characteristic variable from at least two different quality measures. The statistical method may include an arithmetic mean and / or a median and / or a standard deviation and / or a maximum and / or a minimum value of the quality measure.
[0118] Furthermore, in the third method step 102, other image parameters and / or DICOM parameters, such as acquisition time and / or pixel size and / or matrix size and / or image resolution and / or sequence parameters and / or patient sex and / or patient age and / or magnetic field strength and / or gradient field strength, etc., may be taken into account when determining at least one evaluation characteristic variable 29.
[0119] Furthermore, in the third method step 102, the individual quality measures can be weighted when determining the at least one evaluation characteristic variable 29. The weights of the individual quality measures can also be specified by the user, for example by means of the input unit 25 of the user interface 23, in order to take into account individual adjustments and / or individual settings when determining the at least one evaluation characteristic variable 29.
[0120] Furthermore, in the third method step 102, the evaluation unit 28 compares two or more quality measures with at least one input threshold value in order to determine at least one evaluation characteristic variable 29 for classifying the individual quality measures. The input threshold value can be adapted individually to the respective quality measure.
[0121] Furthermore, in the third method step 102, the determined evaluation characteristic variable 29 is classified, wherein the classification of the image quality comprises at least two quality classes 32, 33, 34. The selection of the quality classes 32, 33, 34 is carried out here in dependence on the determined evaluation number, which is compared with a threshold value. The at least two quality classes 32, 33, 34 of the evaluation characteristic variable 29 here comprise a quality class 32 with high image quality and a quality class 34 with low and / or insufficient image quality. Furthermore, the evaluation characteristic variable 29 may also comprise a classification of the image quality into three quality classes 32, 33, 34, wherein the classification provides an additional quality class 33 with medium quality.
[0122] Furthermore, the ascertained evaluation characteristic variable 29 also includes the ascertained evaluation number. In an alternative embodiment, the ascertained evaluation characteristic variable 29 can also include only a classification of the evaluation characteristic variable 29 into two or three quality categories 32 , 33 , 34 .
[0123] Subsequently, at least one evaluation characteristic variable 29 is provided in a fourth method step 103 . In this case, it is provided by means of the evaluation unit 18 to the user interface 23 , in particular to a display unit and / or output unit 24 of the user interface 23 .
[0124] In the fifth method step 104 that follows, at least one evaluation characteristic variable 29 is output to the user. The output is performed by means of a display unit and / or an output unit 24 of the user interface 23. In this case, visual output elements of the respective quality categories 32, 33, 34 for the evaluation characteristic value 29 are generated and output. The visual output elements of the quality category 32 with high image quality can include "good" and / or "+" and / or a smiling emoticon and / or a green dot, etc. The visual output elements of the quality category 34 with low and / or insufficient image quality can include "bad" and / or "-" and / or a crying emoticon and / or a red dot, etc. The visual output elements of the quality category 33 with medium image quality can include "medium" and / or a neutral-looking emoticon and / or an orange or yellow dot, etc.
[0125] Although the details of the present invention have been explained and described in detail through the preferred embodiments, the present invention is not limited to the disclosed examples, and other variations may be derived therefrom by those skilled in the art without departing from the scope of protection of the present invention.
Claims
1. A method for providing at least one evaluation characteristic variable of the image quality of at least one magnetic resonance image, the method comprising the following steps: - providing at least one magnetic resonance image to a quality unit for evaluation of the quality of said image, - determining at least two different quality metrics of the at least one magnetic resonance image, wherein the quality unit comprises two or more quality modules and each of the two or more quality modules is configured to determine one quality metric of the two or more different quality metrics and to provide the two or more quality metrics to the evaluation unit, - ascertaining the at least one evaluation characteristic variable by means of the evaluation unit, wherein the at least one evaluation characteristic variable is ascertained as a function of the two or more different quality measures, and - providing said at least one evaluation characteristic variable.
2. The method according to claim 1, It is characterized in that During the magnetic resonance examination, the at least one evaluation characteristic variable of the image quality of at least one magnetic resonance image is provided.
3. The method according to any one of the preceding claims, It is characterized in that At least one quality module of the two or more quality modules comprises a trained machine learning method, wherein the trained machine learning method is trained to determine a quality measure of the magnetic resonance image to be evaluated by comparison with at least one reference image.
4. The method according to any one of the preceding claims, It is characterized in that At least one quality module of the two or more quality modules comprises a trained deep learning method, wherein the trained deep learning method is trained to determine a quality measure of the magnetic resonance image to be evaluated without comparison with a reference image.
5. The method according to any one of the preceding claims, It is characterized in that The evaluation unit comprises a machine learning method for ascertaining the at least one evaluation characteristic variable from at least two quality measures.
6. The method according to any one of the preceding claims, It is characterized in that The evaluation unit uses at least one statistical method to ascertain the at least one evaluation characteristic variable from at least two quality measures.
7. The method according to any one of the preceding claims, It is characterized in that To ascertain at least one evaluation characteristic variable of the at least one magnetic resonance image, further image parameters and / or DICOM parameters are taken into account by the evaluation unit.
8. The method according to any one of the preceding claims, It is characterized in that The individual ascertained quality measures are weighted in the evaluation unit in order to ascertain the at least one evaluation characteristic variable.
9. The method according to any one of the preceding claims, It is characterized in that In order to ascertain the at least one evaluation characteristic variable, the two or more quality measures are each compared with at least one input threshold value in the evaluation unit in order to classify the individual quality measures.
10. The method according to any one of the preceding claims, It is characterized in that The at least one evaluation characteristic variable comprises a classification of the image quality into at least two quality classes, wherein a quality class is selected as a function of the ascertained evaluation number.
11. The method according to claim 10, It is characterized in that The at least one evaluation characteristic variable comprises the ascertained evaluation number.
12. The method according to any one of claims 10 or 11, It is characterized in that The at least one evaluation characteristic variable is output to a user.
13. A system comprising a magnetic resonance device, a mass unit and an evaluation unit, The system is designed to carry out the method according to claim 1 for providing at least one evaluation characteristic variable for the image quality of at least one magnetic resonance image.
14. A computer program product comprising a program and being directly loadable into a memory of a programmable computing unit, the computer program product having program means for executing the method for providing at least one evaluation characteristic variable for the image quality of at least one magnetic resonance image according to any one of claims 1 to 12 when the program is run in the computing unit.
