Method for magnetic resonance image quality assessment and magnetic resonance imaging system

Abstract

A magnetic resonance image quality evaluation method and a magnetic resonance imaging system. The present application provides a magnetic resonance image quality evaluation method, a magnetic resonance imaging system and a computer readable storage medium. The method comprises: acquiring a magnetic resonance image; processing the magnetic resonance image to extract an image parameter related to a preset index; and determining an image quality evaluation result related to the corresponding index based on the image parameter.

Description

Technical Field

[0001] This invention relates to the field of medical imaging, and in particular to a method for assessing the quality of magnetic resonance images, a magnetic resonance imaging system, and a computer-readable storage medium. Background Technology

[0002] In magnetic resonance imaging (MRI), the quality of the generated MRI images has a significant impact on disease diagnosis and treatment. Doctors or technicians typically judge image quality visually based on experience. With the development of deep learning technology, those skilled in the art have proposed using deep learning to evaluate the quality of MRI images. However, such evaluations are often inaccurate or too general, failing to meet the needs of deeper or broader applications. Summary of the Invention

[0003] One aspect of the present invention provides a method for assessing the quality of magnetic resonance images, comprising:

[0004] Acquire magnetic resonance images; process the magnetic resonance images to extract image parameters related to preset indicators; and determine image quality assessment results related to the corresponding indicators based on the image parameters.

[0005] On the other hand, the method also includes: selecting and outputting the image quality assessment results of the corresponding magnetic resonance images from magnetic resonance image association information pre-stored in the database.

[0006] On the other hand, the associated information includes one or more of the following: time information for generating the magnetic resonance image, scanning parameters when performing the magnetic resonance scan, human anatomical information for performing the magnetic resonance scan, and product information for performing the magnetic resonance scan.

[0007] On the other hand, the method also includes generating status indication information of the magnetic resonance imaging system based on the image quality assessment results.

[0008] On the other hand, the method also includes: selecting the preset indicators and outputting image quality evaluation results related to the selected indicators.

[0009] On the other hand, the step of selecting the preset index includes: selecting one from at least two image quality analysis modes, wherein the at least two image quality analysis modes include different combinations of indexes.

[0010] On the other hand, the preset indicators include one or more of the following: signal-to-noise ratio, contrast-to-noise ratio, motion artifacts, shadow intensity, and subjective rating.

[0011] On the other hand, the steps for processing the magnetic resonance image include: preprocessing the magnetic resonance image; scaling the preprocessed magnetic resonance image according to multiple scaling ratios to obtain multiple scale images; detecting feature points in the multiple scale images respectively; and generating the feature distribution of the feature points.

[0012] On the other hand, the step of detecting feature points in the plurality of scale images respectively includes: performing high-frequency information enhancement on the plurality of scale images, and detecting the feature points in the scale images after high-frequency information enhancement; the step of generating the feature distribution of the feature points includes: determining the main features describing the feature points, and generating a histogram of the main feature distribution.

[0013] On the other hand, the step of determining the image quality assessment result related to the corresponding index based on the image parameters includes: performing machine learning on the feature distribution through a trained first machine learning model to output a subjective score of the corresponding magnetic resonance image.

[0014] On the other hand, the step of processing the magnetic resonance image includes: generating a histogram of the intensity of the magnetic resonance image, and performing a first function fitting on the histogram of the intensity to generate a signal distribution; generating a histogram of the gradient of the magnetic resonance image, and performing a second function fitting on the histogram of the gradient to generate a noise distribution.

[0015] On the other hand, the first function is a Gaussian function, and the second function is a gamma function.

[0016] On the other hand, the step of determining the image quality assessment result related to the corresponding index based on the image parameters includes: obtaining at least one of the signal-to-noise ratio and the contrast-to-noise ratio based on the signal distribution and the noise distribution; and normalizing the signal-to-noise ratio and the contrast-to-noise ratio based on the scanning parameters applied when performing the magnetic resonance scan.

[0017] On the other hand, the steps for processing the magnetic resonance image include: extracting edge information of the magnetic resonance image; obtaining distribution information of the edge information in multiple directions; generating a histogram of the distribution information in multiple directions; and performing function fitting on the histogram of the distribution information in multiple directions and determining the fitting parameters.

[0018] On the other hand, the mean-free contrast normalization coefficient of the magnetic resonance image is extracted as the edge information; the mean-free contrast normalization coefficient is subjected to multi-directional filtering to obtain the multi-directional filtering coefficient as the distribution information of the edge information in multiple directions; and the fitting parameters are determined by fitting a Gaussian function to the histogram of the multi-directional filtering coefficient.

[0019] On the other hand, the step of determining the image quality assessment result related to the corresponding index based on the image parameters includes: performing machine learning on the fitted parameters through a trained second machine learning model to output a judgment result on whether motion artifacts exist.

[0020] On the other hand, the steps of processing the magnetic resonance image include: performing uniformity correction on the magnetic resonance image to obtain a corrected image; and determining the difference between the magnetic resonance image and the corresponding corrected image.

[0021] The step of determining the image quality assessment result related to the index based on the image parameters includes: determining the shading of the magnetic resonance image based on the difference.

[0022] On the other hand, the method also includes generating an analysis report related to the performance of the magnetic resonance imaging system based on the image quality assessment results.

[0023] Another aspect of the present invention provides a computer-readable storage medium comprising a stored computer program, wherein the magnetic resonance image quality assessment method described above is executed when the computer program is run.

[0024] Another aspect of the present invention also provides a magnetic resonance imaging system, comprising:

[0025] Image generating apparatus for performing magnetic resonance scanning to generate magnetic resonance images; and

[0026] A processor for executing any of the above-mentioned magnetic resonance image quality assessment methods.

[0027] On the other hand, the system also includes a memory containing a pre-established database, which includes magnetic resonance image association information, including one or more of the following: time information of the image generation device generating the magnetic resonance image, scanning parameters when performing the magnetic resonance scan, human anatomical information of the person performing the magnetic resonance scan, and product information of the person performing the magnetic resonance scan.

[0028] On the other hand, the processor includes a result output module, which is used to receive magnetic resonance image association information determined by the user through a user interface, and output the image quality assessment result of the corresponding magnetic resonance image through the user interface based on the magnetic resonance image association information.

[0029] On the other hand, the output is also used to receive user-defined metrics for evaluating image quality through the user interface, and to output image quality evaluation results related to the selected metrics.

[0030] It should be understood that the brief description provided above is intended to introduce some concepts further described in the detailed embodiments in a simplified form. This is not intended to identify key or essential features of the claimed subject matter, the scope of which is uniquely defined by the claims following the detailed description. Furthermore, the claimed subject matter is not limited to the implementation of any shortcomings mentioned above or in any paragraph of this disclosure. Attached Figure Description

[0031] The invention will be better understood by referring to the accompanying drawings and by reading the following description of non-limiting embodiments, in which:

[0032] Figure 1 A schematic diagram of an exemplary MRI system 100 according to some embodiments is shown.

[0033] Figure 2 A flowchart 200 of a magnetic resonance image quality assessment method according to an embodiment of the present invention is shown.

[0034] Figure 3 A flowchart 300 of a magnetic resonance quality assessment method according to another embodiment of the present invention is shown.

[0035] Figure 4 A user interface 400 of an example of the present invention is shown.

[0036] Figure 5 A flowchart 500 of a magnetic resonance image quality assessment method according to another embodiment of the present invention is shown.

[0037] Figure 6 A flowchart 600 illustrates a magnetic resonance image quality assessment method according to another embodiment of the present invention.

[0038] Figure 7 A flowchart 700 of a magnetic resonance image quality assessment method according to an embodiment of the present invention is shown.

[0039] Figure 8 The results of quality assessment of multiple magnetic resonance images under subjective scoring indicators using the method of embodiments of the present invention are shown.

[0040] Figure 9 A flowchart 900 of a magnetic resonance image quality assessment method according to another embodiment of the present invention is shown.

[0041] Figure 10 An example of the noise distribution used in this invention to generate magnetic resonance images is shown.

[0042] Figure 11 The results of quality assessment of multiple magnetic resonance images under the signal-to-noise ratio index using the method of embodiments of the present invention are shown.

[0043] Figure 12 A flowchart 1200 of a magnetic resonance image quality assessment method according to another embodiment of the present invention is shown.

[0044] Figure 13 The flowchart 1300 of another embodiment of the magnetic resonance image quality assessment method of the present invention is shown.

[0045] Figure 14 The results of quality assessment of multiple magnetic resonance images under the shadow index using the method of embodiments of the present invention are shown.

[0046] Figure 15 An MRI system 1500 according to an embodiment of the present invention is shown.

[0047] The accompanying drawings illustrate the components, systems, and methods described in the magnetic resonance imaging methods and systems. Together with the following description, the drawings illustrate and explain the structural principles, methods, and concepts described herein. In the drawings, the thickness and dimensions of components may be enlarged or otherwise modified for clarity. Well-known structures, materials, or operations are not shown or described in detail to avoid obscuring the described components, systems, and methods. Detailed Implementation

[0048] The following describes specific embodiments of the present invention. It should be noted that, in order to provide a concise description, this specification cannot exhaustively describe all features of the actual embodiments. It should be understood that, in the actual implementation of any embodiment, just as in any engineering or design project, various specific decisions are often made to achieve the developer's specific goals and to meet system-related or business-related constraints, and this can change from one embodiment to another. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this invention, some design, manufacturing, or production modifications based on the technical content disclosed herein are merely conventional technical means and should not be construed as insufficient content of this disclosure.

[0049] Unless otherwise defined, the technical or scientific terms used in the claims and description shall have the ordinary meaning understood by one of ordinary skill in the art. The terms “first,” “second,” and similar terms used in this specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. The terms “an” or “a” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising” or “including” and similar terms mean that the elements or objects preceding “comprising” or “including” encompass the elements or objects listed following “comprising” or “including” and their equivalents, and do not exclude other elements or objects. The terms “connected” or “linked” and similar terms are not limited to physical or mechanical connections, nor are they limited to direct or indirect connections. Furthermore, it should be understood that references to “an embodiment” or “an embodiment” in this disclosure are not intended to exclude the existence of additional embodiments that also include the referenced features.

[0050] refer to Figure 1 This diagram illustrates an exemplary MRI (Magnetic Resonance Imaging) system 100 according to some embodiments. Operation of the MRI system 100 is controlled by an operator workstation 110, which includes an input device 114, a control panel 116, and a display 118. The input device 114 may be a joystick, keyboard, mouse, trackball, touch-activated screen, voice control, or any similar or equivalent input device. The control panel 116 may include a keyboard, touch-activated screen, voice control, buttons, sliders, or any similar or equivalent control device. The operator workstation 110 is coupled to and communicates with a computer system 120, which enables the operator to (e.g., via the input device) control the generation and viewing of images on the display 118, and to perform human-computer interaction through a user interface displayed on the display 118. This human-computer interaction can be used to: determine scan parameters, perform image processing operations, select images, and view the quality assessment results of selected images, etc. The computer system 120 includes multiple components that communicate with each other via an electrical and / or data connection module 122. The connection module 122 can be a direct wired connection, a fiber optic connection, a wireless communication link, etc. The computer system 120 may include a central processing unit (CPU) 124, a memory 126, and an image processor 128. In some embodiments, the image processor 128 may be replaced by image processing functions implemented in the CPU 124. The computer system 120 may be connected to an archive media device, permanent or backup storage, or a network. The computer system 120 may be coupled to and communicate with a separate MRI system controller 130.

[0051] The MRI system controller 130 includes a set of components that communicate with each other via an electrical and / or data connection module 132. The connection module 132 may be a direct wired connection, a fiber optic connection, a wireless communication link, etc. The MRI system controller 130 may include a CPU 131, a sequence pulse generator 133 that communicates with an operator workstation 110, a transceiver (or RF transceiver) 135, a memory 137, and an array processor 139. In some embodiments, the sequence pulse generator 133 may be integrated into the resonant assembly 140 of the MRI system 100. The MRI system controller 130 may receive commands from the operator workstation 110 to indicate the MRI scan sequence to be performed during an MRI scan. The MRI system controller 130 is also coupled to and communicates with a gradient driver system 150, which is coupled to a gradient coil assembly 142 to generate a magnetic field gradient during an MRI scan.

[0052] The sequence pulse generator 133 may also receive data from a physiological acquisition controller 155, which receives signals from multiple different sensors, such as electrocardiogram (ECG) signals from electrodes attached to the patient, connected to the subject or patient 170 undergoing an MRI scan. The sequence pulse generator 133 is coupled to and communicates with a scan room interface system 145, which receives signals from various sensors associated with the state of the resonant assembly 140. The scan room interface system 145 is also coupled to and communicates with a patient positioning system 147, which sends and receives signals to control the movement of the patient table to the desired position for the MRI scan.

[0053] MRI system controller 130 provides gradient waveforms to gradient driver system 150, the gradient driver system including G x G y and G z Amplifiers, etc. Each G x G y and G zGradient amplifiers excite corresponding gradient coils in gradient coil assembly 142 to generate magnetic field gradients for spatial encoding of MR signals during MRI scans. Gradient coil assembly 142 is disposed within resonant assembly 140, which also includes a superconducting magnet with a superconducting coil 144 that provides a static, uniform longitudinal magnetic field B0 throughout the cylindrical imaging volume 146 during operation. Resonant assembly 140 also includes an RF body coil 148 that provides a transverse magnetic field B1 during operation, which is substantially perpendicular to B0 throughout the cylindrical imaging volume 146. Resonant assembly 140 may also include an RF surface coil 149 for imaging different anatomical structures of a patient undergoing an MRI scan. RF body coil 148 and RF surface coil 149 may be configured to operate in transmit and receive modes, transmit mode, or receive mode.

[0054] The MRI scan subject or patient 170 can be positioned within the cylindrical imaging volume 146 of the resonance assembly 140. The transceiver 135 in the MRI system controller 130 generates RF excitation pulses amplified by the RF amplifier 162 and provides them to the RF body coil 148 via the transmit / receive switch (T / R switch) 164.

[0055] As described above, the RF body coil 148 and RF surface coil 149 can be used to transmit RF excitation pulses and / or receive resulting MR signals from a patient undergoing an MRI scan. MR signals emitted by nuclei excited within the patient during an MRI scan can be sensed and received by the RF body coil 148 or RF surface coil 149 and transmitted back to the preamplifier 166 via a T / R switch 164. The T / R switch 164 can be controlled by a signal from the sequence pulse generator 133 to electrically connect the RF amplifier 162 to the RF body coil 148 during transmit mode and to connect the preamplifier 166 to the RF body coil 148 during receive mode. The T / R switch 164 can also enable the RF surface coil 149 to be used in either transmit or receive mode.

[0056] In some implementations, the MR signal sensed and received by the RF body coil 148 or the RF surface coil 149 and amplified by the preamplifier 166 is stored as a raw k-space data array in memory 137 for post-processing. A reconstructed magnetic resonance image can be obtained by transforming / processing this stored raw k-space data.

[0057] In some implementations, the MR signal sensed and received by the RF body coil 148 or RF surface coil 149 and amplified by the preamplifier 166 is demodulated, filtered, and digitized in the receiving section of the transceiver 135 and transmitted to the memory 137 in the MRI system controller 130. For each image to be reconstructed, the data is rearranged into separate k-space data arrays, and each of these separate k-space data arrays is input to the array processor 139, which is operated to perform a Fourier transform on the data into an array of image data.

[0058] The array processor 139 uses a transformation method, most commonly Fourier transform, to create an image from the received MR signal. These images are transmitted to the computer system 120 and stored in the memory 126. In response to a command received from the operator workstation 110, the image data may be stored in long-term memory, or it may be further processed by the image processor 128 and transmitted to the operator workstation 110 for display on the monitor 118.

[0059] In various implementations, components of the computer system 120 and the MRI system controller 130 may be implemented on the same computer system or multiple computer systems. It should be understood that... Figure 1 The MRI system 100 shown is for illustrative purposes. Suitable MRI systems may include more, fewer, and / or different components.

[0060] The MRI system controller 130 and image processor 128 may each include a computer processor and a storage medium, either individually or jointly. The storage medium records a program for predetermined data processing to be executed by the computer processor. For example, the storage medium may store programs for performing scan processing (e.g., scan procedures, imaging sequences), image reconstruction, image processing, etc. For instance, it may store a program for implementing the magnetic resonance image quality assessment method of the embodiments of the present invention. The storage medium may include, for example, a ROM, floppy disk, hard disk, optical disk, magneto-optical disk, CD-ROM, or a non-volatile memory card.

[0061] refer to Figure 2 The diagram illustrates a flowchart 200 of a magnetic resonance image quality assessment method according to an embodiment of the present invention. In step 201, a magnetic resonance image is acquired. In step 205, the magnetic resonance image is processed to extract image parameters related to a preset index. In step 209, an image quality assessment result related to the index is determined based on the extracted image parameters.

[0062] The aforementioned preset metrics may include one or more of the following: signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), motion artifacts, shading, and subjective (e.g., human visual) ratings.

[0063] In the above embodiments, by performing image processing based on preset indicators to extract image parameters related to the indicators, and then performing quality assessment based on the image parameters, the accuracy of quality assessment can be improved, and the quality assessment results under different indicators may play a role in different application requirements.

[0064] Such quality assessment methods can be used for system performance evaluation, tracing influencing factors in the image generation process, and improving system maintenance.

[0065] For example, current routine maintenance of MRI products is typically scheduled at fixed time intervals, usually based on experience, such as every six months or a quarter. This approach may have the following problems: when product performance is good, the maintenance frequency is too high, resulting in wasted on-site maintenance manpower; when product performance deteriorates, the maintenance frequency is too low, leading to malfunctions or unacceptable image quality before maintenance, and in more serious cases, increased system downtime, causing losses to the hospital. Through embodiments of the present invention, after a new MRI image is generated through MRI scanning, the image quality assessment method of the present invention can be automatically or based on input commands to generate image quality assessment results under multiple indicators. Therefore, the current working status of the system can be determined based on the comprehensive assessment results of these indicators, and timely warnings or automatic communication with remote maintenance engineers can be issued when system performance is abnormal.

[0066] For example, in recent years, third-party suppliers have emerged that can provide product maintenance. Some of their operations during product maintenance (such as replacing key components) may cause serious product performance degradation. However, by performing image quality evaluation on the generated magnetic resonance images under multiple indicators through the embodiments of the present invention, it is possible to support the clarification of the root cause of performance degradation by reviewing the trend of the evaluation results (or other possible methods), which facilitates timely response. Furthermore, the impact of such operations can be quantified by issuing appropriate warnings.

[0067] In one embodiment, the magnetic resonance image acquired in step 203 can be obtained based on the execution of a magnetic resonance scan. For example, once a magnetic resonance scan (or imaging) procedure is performed using a magnetic resonance imaging system similar to system 100 described above and a reconstructed magnetic resonance image is generated, the magnetic resonance image quality assessment method of this embodiment can be automatically or based on input instructions to perform a quality assessment on the generated magnetic resonance image after the procedure is completed. The quality assessment results can be stored in a database.

[0068] refer to Figure 3The diagram illustrates a flowchart 300 of a magnetic resonance imaging (MRI) quality assessment method according to another embodiment of the present invention, which further includes step 313: selecting and outputting the image quality assessment result of the corresponding MRI image based on MRI image association information pre-stored in a database. This database can be stored in the aforementioned computer system 120 (e.g., in memory 126), or it can be stored in an independent storage system capable of communicating with system 100.

[0069] As mentioned earlier, the database includes image quality assessment results, which can be retrieved in various applications.

[0070] For example, the quality assessment result can be retrieved by setting search criteria, which can be achieved through a user interface, such as display 118. Specifically, the database may also include magnetic resonance image association information, and in step 313, the image quality assessment result of the corresponding magnetic resonance image can be selected and output (e.g., displayed) based on this association information.

[0071] By storing image association information in the database instead of the images themselves, redundant image storage (e.g., images already stored in system 100's memory 126) can be avoided. Furthermore, by storing association information and image quality assessment results, embodiments of the invention can be implemented using only a sufficiently lightweight database. However, in other embodiments, this database can also be used to store magnetic resonance images.

[0072] The aforementioned associated information may be generated during the magnetic resonance imaging (MRI) scan procedure and subsequently recorded / collected in a database. For example, when starting a new MRI scan procedure, relevant time information is recorded, and scan settings (such as imaging sequence selection, coil selection, and anatomical structure selection) are performed; this setting information is also collected and stored in the database. This associated information may include one or more of the following: the time information for generating the MRI image, the scan parameters during the MRI scan, the human anatomical information for performing the MRI scan, and the product information for performing the MRI scan.

[0073] Figure 4 An example user interface 400 is shown, including a time determination module 410 and an anatomical structure determination module 420. By operating the time determination module 410, a time range or a specific time can be input, and image quality assessment results of magnetic resonance images generated in the database within that time range or at that specific time are selected and output. By operating the anatomical structure determination module 420, image quality assessment results of magnetic resonance images of a specific anatomical structure can be selected.

[0074] Once the user confirms the above operation, the corresponding image quality assessment result 430 is retrieved. For example, it can be directly displayed on the user interface 400, or it can be further used for system performance analysis, and then the system performance analysis result 440 is displayed on the user interface 400.

[0075] Although the above describes automatically evaluating the image quality of the generated images after the scan, it is also possible to first store the association information in the database, then retrieve the magnetic resonance images related to the association information based on the user's selection of association information, and perform quality evaluation on the retrieved images and output the evaluation results.

[0076] Figure 5 A flowchart 500 of a magnetic resonance image quality assessment method according to another embodiment of the present invention is shown, which further includes step 513: determining and outputting status indication information of the magnetic resonance system based on the image quality assessment results. This status indication information may include the aforementioned system performance analysis results. For example, the status indication information may be used to indicate whether the image quality assessment results based on various indicators show good system performance or performance problems. The status indication information may also include alarm information when performance problems are severe.

[0077] Figure 6 A flowchart 600 of a magnetic resonance image quality assessment method according to another embodiment of the present invention is shown, which further includes step 613: selecting preset indicators and outputting image quality assessment results related to the selected indicators. In this way, image quality assessment results for specific indicators can be output only as needed, rather than all information. For example, for doctors, the indicators they focus on might be subjective ratings, signal-to-noise ratio, etc. For equipment maintenance personnel, they might focus on more indicators, thus providing selectable modules to achieve personalized result output.

[0078] Specifically, in step 613, one can be selected from at least two image quality analysis modes, wherein the at least two image quality analysis modes include different combinations of indicators.

[0079] In one example, the above selection can be achieved through a user interface 400. Specifically, the user interface 400 further includes a clinical mode module 450 and a maintenance mode module 460. When the user operation determines to enable the clinical mode module 450, the output image quality assessment result is only related to the indicators that the doctor is concerned about. When the user operation determines to activate the maintenance mode module 460, the output image quality assessment result is related to more or different indicators.

[0080] The selection of the above modes can also be completed automatically through username detection / authentication. For example, if the current user is identified as a clinician or hospital staff, the first image quality analysis mode (clinical mode) is enabled, while if the current user is identified as a system maintenance personnel, the second image quality analysis mode (maintenance mode) is enabled.

[0081] In embodiments of the present invention, image parameters related to a preset index are extracted by processing magnetic resonance images. Therefore, the parameters extracted by image processing are suitable for outputting evaluation results that are more relevant to the index, thereby improving the accuracy of quality assessment and making the evaluation results more multidimensional.

[0082] refer to Figure 7 The diagram shows a flowchart 700 of a magnetic resonance image quality assessment method according to an embodiment of the present invention, wherein the assessment results are used to obtain subjective scores.

[0083] Step 205 above may specifically include Figure 7 Steps 701, 703, 705 and 707, step 209 includes step 709.

[0084] In step 701, the magnetic resonance image is preprocessed, including normalization, to improve the consistency of the resulting evaluation.

[0085] In step 703, the preprocessed magnetic resonance image is scaled based on multiple scaling ratios to obtain multiple scaled images. For example, two different scaling ratios can be set, which can be less than, equal to, or greater than 1, thus generating two or more images with different scaling ratios. In this way, image features can be fully detected subsequently.

[0086] In step 705, feature points are detected in the multiple scale images respectively. For example, these feature points include pixels in the image that can predict visually saliency features. Specifically, in step 705, a high-frequency enhancement filter can be used to enhance the high-frequency information of the multiple scale images to highlight high-frequency features, and feature points are detected in the enhanced scale images. In one example, a multi-stage filtering algorithm can be used for feature point detection, which can detect more feature points compared to the traditional method of performing only one filtering step.

[0087] In step 707, the feature distribution of the aforementioned feature points is generated. In one example, this may specifically include the following steps: determining the principal features describing the feature point and generating a histogram of the principal feature distribution. Each feature point can be described by multiple features, which may include, for example, the differences between the feature point and its surrounding pixels, as well as other data that can reflect image features. Hundreds of such features could even be generated for description. However, in embodiments of the present invention, only one or more principal features are selected to describe each feature point. These principal features better reflect human visual feedback to the image, thus enabling a more accurate subjective rating with a sufficiently small computational load.

[0088] In one example, the histogram of the main feature distribution can be generated by performing a preset mathematical operation on the main feature (or feature point description data) of each feature point to obtain feature data, and then counting the number of feature points distributed at each feature data point.

[0089] In step 709, the above feature distribution is machine learned by the first machine learning model trained to output the subjective score of the corresponding magnetic resonance image.

[0090] The aforementioned machine learning model may include a deep learning network or a support vector machine model, and its training dataset may include an input dataset and an output dataset, wherein the input dataset may be the main feature distribution of feature points detected in a magnetic resonance image with known subjective ratings, and the output dataset may be the aforementioned known subjective ratings.

[0091] A suitable model can be selected from existing deep learning network models or support vector machine models and embedded into the magnetic resonance imaging system or a remote system that communicates with the magnetic resonance imaging system.

[0092] Figure 8 The method of this invention is shown to evaluate the quality of multiple magnetic resonance images under a subjective scoring index, with corresponding scores shown. The higher the quality under the subjective scoring index, the higher the score, which is consistent with the doctor's experience judgment and can therefore serve as a reliable basis for clinical evaluation.

[0093] refer to Figure 9 The diagram shows a flowchart 900 of a magnetic resonance image quality assessment method according to another embodiment of the present invention, wherein assessment results of signal-to-noise ratio and contrast-to-noise ratio can be obtained.

[0094] Step 205 above may specifically include Figure 7 Steps 901 and 903, and step 209 includes steps 905 and 907.

[0095] In step 901, a histogram of the intensity of the magnetic resonance image is generated, and the histogram is fitted with a first function to generate a signal distribution. In this step, an intensity map of the magnetic resonance image can be generated, and the intensity histogram is obtained based on the signal intensity value distribution in the intensity map. In this embodiment, a Gaussian function is used to fit the intensity histogram to obtain a more accurate signal intensity distribution.

[0096] In step 903, a histogram of the gradient of the magnetic resonance image is generated, and a second function is fitted to the histogram to generate the noise distribution. In this step, the gradient map of the magnetic resonance image can be obtained by performing a first-order derivative operation, further removing non-noise components from the gradient map, and generating a histogram of the gradient based on the gradient map after removing non-noise components. In embodiments of the present invention, high-pass filtering can be applied to the gradient map to remove non-noise components. By fitting the histogram of the gradient with a second function, the influence of noise is enhanced while the influence of the image signal is reduced, thus reflecting the noise distribution more accurately. In embodiments of the present invention, the second function is the gamma function.

[0097] refer to Figure 10 This illustrates an example of the noise distribution used in this invention to generate magnetic resonance images, wherein gradient maps g in the x and y directions are generated, respectively. x1 and g y1 For this gradient map g x1 and g y1 High-pass filtering is performed to remove low-order components that do not reflect noise characteristics, and the gradient map g is obtained. x2 and g y2 To obtain the gradient map g x2 and g y2 A fusion operation is performed to obtain the final gradient map g. m The above fusion operation can be described by the following formula:

[0098]

[0099] Furthermore, the gradient map g is generated. m The histogram is obtained by fitting it with a gamma function to obtain the fitted histogram H. m By comparing the histograms of the gradient maps after high-pass filtering (H... x2 and H y2 ) and the histogram of the original gradient map (H x1 and H y1 As can be seen, the histogram (H) x2 and H y2 It has a more accurate noise distribution; therefore, the histogram H obtained based on the fused data... mIt also has a more accurate noise distribution, which helps to obtain more accurate quality assessment results based on the noise distribution, such as signal-to-noise ratio and contrast-to-noise ratio.

[0100] For example, noise in magnetic resonance images is often related to a variety of scanning parameters, such as:

[0101]

[0102]

[0103] Where N represents noise, T represents the layer thickness of the excited layer, NEX represents the number of excitations of the magnetic resonance signal, rBW represents the RF signal receiving bandwidth, Rec.Matrix represents the size of the reconstructed data, and Acq.Matrix represents the size of the acquired data. Experiments show that the noise distribution obtained using the embodiments of this invention can better satisfy the above specific relationships, therefore its distribution is more accurate.

[0104] In step 905, at least one of signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) is obtained based on the signal distribution and noise distribution. For example, those skilled in the art will understand that the ratio of the peak value in the signal distribution to the peak value in the noise distribution can be used to obtain the SNR, and the difference between two adjacent peak values ​​in the largest signal in the signal distribution can be used to obtain the CNR.

[0105] In step 907, the obtained signal-to-noise ratio is normalized based on the scanning parameters applied during the magnetic resonance scan. This method allows for the comparison of product performance and image quality based on the quality assessment of magnetic resonance images obtained under different scanning parameters.

[0106] Taking slice thickness as an example, if the first magnetic resonance image is obtained under the condition of slice thickness of 4, it has a first noise distribution, while the second magnetic resonance image is obtained under the condition of slice thickness of 2, it has a second noise distribution. By performing quality assessment on the first image to obtain the first signal-to-noise ratio, and by performing quality assessment on the second image to obtain the second signal-to-noise ratio, the second noise distribution can be normalized to a new noise distribution when the slice thickness is 4 based on the above relationship between noise and slice thickness. The signal-to-noise ratio when the slice thickness is 4 can be generated through the new noise distribution.

[0107] refer to Figure 11 The present invention demonstrates the quality assessment results of multiple magnetic resonance images under the signal-to-noise ratio index using the method of the present invention, wherein corresponding scores are obtained, wherein the higher the image signal-to-noise ratio, the higher the score, and the results are verified to be accurate and therefore can be used as a reliable basis for clinical evaluation.

[0108] Table 1 below shows the scanning parameters used when acquiring MRI images of the same site from the same volunteers using three different models of the product.

[0109] Product 1 Product 2 Product 3 coil Coil A Coil B Coil B Size of collected data First dimension Second size Third size Reconstructed data size Fourth size Fourth size Fourth size Number of incentives 1 1 2 Received signal bandwidth First bandwidth Second bandwidth Third bandwidth

[0110] Table 1

[0111] Table 2 below shows the evaluation results of the MRI images acquired based on the three products in Table 1 under three metrics: subjective score, image signal-to-noise ratio (SNR) (unnormalized), and normalized SNR obtained by normalizing the image SNR based on multiple scanning parameters in Table 1. The figures in Table 2 are only used to compare the score sizes under different products.

[0112] Product 1 Product 2 Product 3 Subjective rating 4.8 4.8 5.0 Image signal-to-noise ratio 61 99 118 Normalized signal-to-noise ratio 7.7 8.4 6.6

[0113] Table 2

[0114] Analysis of Table 2 reveals that although Product 3 has a higher subjective score and obtains images with a higher signal-to-noise ratio, it also has a lower normalized signal-to-noise ratio. This indicates that the image quality degradation caused by Product 3 is likely only due to parameter settings, and it has better system performance than the other two products.

[0115] The embodiments of the present invention can also be used for performance analysis of individual products. Based on such analysis, a corresponding analysis report can be obtained, and users can read or retrieve the analysis report based on, for example, the user interface operation described above.

[0116] refer to Figure 12 The flowchart 1200 of another embodiment of the magnetic resonance image quality assessment method of the present invention is shown, wherein the assessment results under the motion artifact index can be obtained.

[0117] Step 205, which processes the magnetic resonance image, may include steps 1201, 1203, 1205, and 1207. Step 209, which determines the image quality assessment result related to the corresponding index based on the image parameters, may include step 1209.

[0118] In step 1201, edge information of the magnetic resonance image is extracted. In this step, the edge information can be obtained by, for example, extracting the MSCN (Mean-Subtracted Contrast-Normalized) coefficients of the magnetic resonance image. The MSCN coefficients can be obtained by the following formula:

[0119]

[0120] Where I is the magnetic resonance image to be evaluated, μ is the signal mean of the magnetic resonance image I, σ is the standard deviation of the signal of the magnetic resonance image I, and C is a constant.

[0121] In step 1203, the distribution information of the edge information in multiple directions is obtained. Specifically, this can be achieved by performing multi-directional filtering on the MSCN coefficients to obtain the multi-directional filtering coefficients as the distribution information of the edge information in multiple directions.

[0122] The distribution information of edge information in multiple directions can be represented by a histogram. In step 1205, a histogram of the distribution information in the above multiple directions is generated.

[0123] In step 1207, a function is fitted to the histograms of the distribution information in multiple directions, and the fitting parameters are determined. The histograms of the distribution information in multiple directions can be fitted separately to obtain multiple sets of fitting parameters corresponding to each direction. These multiple sets of fitting parameters are then fused (e.g., summed, meaned, varianced, etc.) to form the final fitting parameters. In this embodiment of the invention, a Gaussian function is specifically used to fit the histograms of the distribution information in multiple directions.

[0124] In step 1209, the above-mentioned fitting parameters are machine learned by the trained second machine learning model to output the judgment result of whether motion artifacts exist.

[0125] The aforementioned second machine learning model may include a deep learning network or a support vector machine model, and its training dataset may include a first dataset and a second dataset, wherein the magnetic resonance images in the first dataset contain motion artifacts (or motion artifacts are synthesized in them through image processing), and the magnetic resonance images in the second dataset do not contain motion artifacts (or motion artifacts have been removed through image processing).

[0126] As discussed herein, support vector machines (SVMs) achieve this by finding a hypersurface among possible inputs, where the hypersurface is used for binary classification of the input data. A first or second machine learning network in an embodiment of the invention may include multiple ensembled or cascaded support vector machine models for classifying the input data.

[0127] As discussed in this paper, deep learning techniques (also known as deep machine learning, hierarchical learning, or deep structured learning) employ artificial neural networks that learn to process input data. Deep learning methods are characterized by using one or more network architectures to extract or model data of interest. Deep learning methods can be accomplished using one or more layers (e.g., input layers, normalization layers, convolutional layers, output layers, etc., with varying numbers and functions depending on the deep network model). The configuration and number of layers allow deep networks to handle complex information extraction and modeling tasks. Specific parameters (also called “weights” or “biases”) of the network are typically estimated through a so-called learning process (or training process). The learned or trained parameters usually result in (or output) a network corresponding to different levels of layers; therefore, extracting or modeling different aspects of the initial data or the output of a previous layer can often represent the hierarchical structure or cascade of layers. In image processing or reconstruction, this can be characterized as different layers relative to different levels of features in the data. Thus, processing can be layered; that is, earlier or higher-level layers may correspond to extracting “simple” features from the input data, followed by layers that combine these simple features to exhibit more complex features. In practice, each layer (or more specifically, each "neuron" in each layer) can employ one or more linear and / or nonlinear transformations (so-called activation functions) to process the input data into an output data representation. The number of "neurons" can be constant across multiple layers, or it can vary from layer to layer.

[0128] As discussed herein, as part of the initial training of a machine learning process to solve a specific problem, the training dataset for training a machine learning model includes known input values ​​and the expected (target) output value of the final output of the machine learning process. In this way, a deep learning algorithm can process this training dataset (in a supervised or guided manner, or in an unsupervised or unguided manner) until it identifies the mathematical relationship between the known inputs and the expected output and / or identifies and represents the mathematical relationship between the inputs and outputs of each layer. The learning process typically utilizes (partial) input data and creates a network output for that input data, then compares the created network output with the expected output of the dataset, and then iteratively updates the network parameters (weights and / or biases) using the difference between the created and expected outputs. The network parameters can typically be updated using the Stochastic Gradient Descent (SGD) method; however, those skilled in the art will understand that other methods known in the art can also be used to update the network parameters. Similarly, a separate validation dataset can be used to validate the trained network, where both the known input and the desired output are known. The network output can be obtained by feeding the known input to the trained network, and then the network output is compared with the (known) desired output to validate previous training and / or prevent overtraining.

[0129] Once the machine learning model is created or trained, the corresponding quality assessment results can be obtained by simply inputting the image parameters obtained through image processing into the model. This model can be integrated with a module for image processing; in this case, the corresponding quality assessment results can be obtained by simply inputting the magnetic resonance image into the integrated module.

[0130] In some embodiments, the machine learning model trained described above is trained on a training module on an external carrier (e.g., a device other than a magnetic resonance imaging system). In some embodiments, the training system may include a first module for storing the training dataset, a second module for training and / or updating the model, and a communication network for connecting the first and second modules. In some embodiments, the first module includes a data transmission unit and a first storage unit, wherein the first storage unit stores the training dataset, and the data transmission unit receives relevant instructions (e.g., retrieve training dataset) and sends the training dataset according to the instructions. Furthermore, the second module includes a model update unit and a second storage unit, wherein the second storage unit stores the trained model, and the model update unit receives relevant instructions, trains and / or updates the network, etc. In other embodiments, the training dataset may also be stored in the second storage unit of the second module, and the training system may not include the first module. In some embodiments, the communication network may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0131] Once data (e.g., a trained network or model) is generated and / or configured, it can be copied and / or loaded into the MRI system, which can be done in various ways. For example, the model can be loaded via a directed connection or link between the MRI system and the computer, or communication between different components can be accomplished using available wired and / or wireless connections and / or according to any suitable communication (and / or network) standards or protocols. Alternatively, data can be loaded into the MRI system indirectly. For example, data can be stored on a suitable machine-readable medium (e.g., a flash memory card, etc.) and then used to load the data (in the field, such as by a user or authorized person of the system) into the MRI system, or data can be downloaded to an electronic device capable of local communication (e.g., a laptop computer, etc.) and then used in the field (e.g., by a user or authorized person of the system) to upload the data to the MRI system via a direct connection (e.g., a USB connector, etc.).

[0132] refer to Figure 13 The flowchart 1300 of another embodiment of the magnetic resonance image quality assessment method of the present invention is shown, wherein the assessment results under the shadow index can be obtained.

[0133] Step 205, which processes the magnetic resonance image, may include steps 1301 and 1303. Step 209, which determines the image quality assessment result related to the corresponding index based on the image parameters, may include step 1305.

[0134] In step 1301, uniformity correction is performed on the magnetic resonance image to obtain a corrected image. In step 1303, the difference between the magnetic resonance image and the corresponding corrected image is determined. The above correction can be implemented based on existing models for uniformity correction of images. Those skilled in the art will understand that there are various uniformity correction methods in the prior art that can effectively improve the uniformity of magnetic resonance images and reduce shadows in the images. Therefore, in step 1305, the difference between the corrected image (i.e., the corrected image) and the magnetic resonance image to be evaluated is used as an image parameter to further evaluate the image quality of the magnetic resonance image under the shadow index, which can obtain accurate evaluation results. The image quality under the shadow index can be represented by the shadow score.

[0135] refer to Figure 14 The present invention illustrates the quality assessment results of multiple magnetic resonance images under the shadow index using the method of the present invention, wherein shadow severity scores are obtained. The results show that the more severe the shadow in the image, the higher the shadow severity score, indicating that the assessment results are accurate and can therefore be used as a reliable basis for clinical evaluation.

[0136] Table 3 below shows the scanning parameters set when scanning the same volunteer before and after replacing the main component (replacing the original coil C with coil D) using products from the same series.

[0137] Product 4 before coil replacement Product 4 after coil replacement coil Coil C Coil D Size of collected data Fifth size Fifth size Reconstructed data size Fourth size Fourth size Number of incentives 1.5 1.5

[0138] Table 3

[0139] Table 4 below shows the quality assessment results of magnetic resonance images acquired based on the two sets of scanning parameters in Table 1 under different indicators.

[0140] Product 4 before coil replacement Product 4 after coil replacement Subjective rating 4.8 4.3 Image signal-to-noise ratio 85 65

[0141] Table 4

[0142] The comparison shows that the image quality of the acquired images decreased significantly after the coil was replaced, indicating that the coil replacement operation affected the product performance. Based on this comparison, the embodiments of the present invention can also generate corresponding analysis reports or warning messages for user reference.

[0143] An exemplary embodiment of the present invention may also provide a computer-readable storage medium comprising a stored computer program, wherein the computer program, when run, executes the magnetic resonance image quality assessment method of any of the above embodiments.

[0144] refer to Figure 15This illustration shows a magnetic resonance imaging (MRI) system 1500 according to an embodiment of the present invention, which includes an image generation device 1510 and a processor 1530. The image generation device 1510 is used to perform a magnetic resonance scan to generate a magnetic resonance image, and the device 1510 may include, for example... Figure 1 Some or all of the components / units / modules in the system 100 shown, or alternatives to these components / units / modules that have similar functions.

[0145] The processor 1530 is used to execute the magnetic resonance image quality assessment method of any of the above embodiments. The processor 1530 may be located in a computer system capable of communicating with the image generation apparatus 1510, and may communicate with the computer system in the system 100 described above.

[0146] refer to Figure 15 The system 1500 further includes a memory 1550, which can be any memory in the system 100 or a separate memory accessible to the image generating device 1510 and the processor 1530. The memory 1550 stores a pre-established database 1551, which includes association information associated with, for example, magnetic resonance images generated by the image generating device 1510. This association information includes one or more of the following: time information of the image generating device 1510 generating the magnetic resonance image, scan parameters during the magnetic resonance scan, human anatomical information during the magnetic resonance scan, and product information for performing the magnetic resonance scan. The database 1551 also includes image quality evaluation results of the aforementioned magnetic resonance images.

[0147] Continue to refer to Figure 15 The processor 1530 may include a result output module 1531, which is used to receive user-defined association information through a user interface, select the corresponding magnetic resonance image quality assessment result based on the association information, and output the selected image quality assessment result through the user interface.

[0148] The output module 1531 is further configured to receive, via a user interface, an indicator for evaluating image quality, and output the image quality evaluation result under the selected indicator via the user interface.

[0149] In the embodiments of the present invention, image parameters related to the preset indicators are extracted by image processing, and quality assessment is performed based on the image parameters. This can improve the accuracy of quality assessment and generate more detailed assessment results in multiple dimensions (indicators). The reliability of the result data is strong, which facilitates further analysis based on the results to meet a wider range of needs.

[0150] In addition to any modifications previously indicated, those skilled in the art can devise many other variations and alternative arrangements without departing from the spirit and scope of this description, and the appended claims are intended to cover such modifications and arrangements. Therefore, although the information has been described above in a specific and detailed manner in conjunction with what is currently considered to be the most practical and preferred aspects, it will be apparent to those skilled in the art that many modifications can be made without departing from the principles and concepts set forth herein, including but not limited to changes in form, function, mode of operation, and use. Likewise, as used herein, in all respects, examples and embodiments are intended to be illustrative only and should not be construed as restrictive in any way.

[0151] The purpose of providing the above specific embodiments is to enable a more thorough and comprehensive understanding of the disclosure of this invention, but this invention is not limited to these specific embodiments. Those skilled in the art should understand that various modifications, equivalent substitutions, and changes can be made to this invention, and all such modifications and changes should be within the scope of protection of this invention, provided they do not depart from the spirit of this invention.

Claims

1. A method for assessing the quality of magnetic resonance images, the method comprising: Acquire magnetic resonance images; The magnetic resonance image is processed to extract image parameters related to a preset index; as well as, Based on the image parameters, determine the image quality assessment results related to the corresponding indicators. The steps for processing the magnetic resonance image include: Generate a histogram of the intensity of the magnetic resonance image, and fit the histogram of the intensity to a first function to generate a signal distribution; A histogram of the gradient of the magnetic resonance image is generated, and a second function is fitted to the histogram of the gradient to generate a noise distribution.

2. The method according to claim 1, wherein, Also includes: The image quality assessment results of the corresponding magnetic resonance images are selected and output based on the magnetic resonance image association information pre-stored in the database.

3. The method according to claim 2, wherein: The magnetic resonance image association information includes one or more of the following: the time information for generating the magnetic resonance image, the scanning parameters when performing the magnetic resonance scan, the human anatomical information for performing the magnetic resonance scan, and the product information for performing the magnetic resonance scan.

4. The method according to claim 1, wherein, Also includes: The status indication information of the magnetic resonance imaging system is generated based on the image quality assessment results.

5. The method according to claim 1, wherein, It also includes: selecting the preset indicators and outputting image quality evaluation results related to the selected indicators.

6. The method according to claim 5, wherein, The step of selecting the preset index includes: selecting one from at least two image quality analysis modes, wherein the at least two image quality analysis modes include different combinations of indexes.

7. The method according to claim 1, wherein, The preset metrics include one or more of the following: signal-to-noise ratio, contrast-to-noise ratio, motion artifacts, shadow intensity, and subjective rating.

8. The method according to claim 7, wherein, The steps for processing the magnetic resonance image include: The magnetic resonance image is preprocessed; The preprocessed magnetic resonance image is scaled according to multiple scaling ratios to obtain multiple scaled images. Detect feature points in the plurality of scaled images respectively; and, Generate the feature distribution of the feature points.

9. The method according to claim 8, wherein, The step of detecting feature points in the plurality of scale images includes: performing high-frequency information enhancement on the plurality of scale images, and detecting the feature points in the scale images after high-frequency information enhancement; The steps for generating the feature distribution of the feature points include: determining the main features describing the feature points and generating a histogram of the main feature distribution.

10. The method according to claim 8, wherein, The steps for determining the image quality assessment result related to the corresponding index based on the image parameters include: The feature distribution is machine-learned using a first machine learning model that has been trained to output a subjective score for the corresponding magnetic resonance image.

11. The method according to claim 1, wherein, The first function is a Gaussian function, and the second function is a gamma function.

12. The method according to claim 1, wherein, The steps for determining the image quality assessment result related to the corresponding index based on the image parameters include: Based on the signal distribution and noise distribution, at least one of the signal-to-noise ratio and contrast-to-noise ratio is obtained; and, At least one of the signal-to-noise ratio and contrast-to-noise ratio is normalized based on the scanning parameters applied during the magnetic resonance scan.

13. The method according to claim 7, wherein, The steps for processing the magnetic resonance image include: Extract the edge information from the magnetic resonance image; Obtain the distribution information of the edge information in multiple directions; Generate a histogram of the distribution information in the multiple directions; The histograms of the distribution information in the multiple directions are fitted with functions, and the fitting parameters are determined.

14. The method according to claim 13, wherein, The edge information is obtained by extracting the mean-free contrast normalization coefficient of the magnetic resonance image. By performing multi-directional filtering on the mean-reduced contrast normalization coefficients, the multi-directional filtering coefficients are obtained as the distribution information of the edge information in multiple directions; The fitting parameters are determined by fitting a Gaussian function to the histogram of the multi-directional filter coefficients.

15. The method according to claim 13, wherein, The steps for determining the image quality assessment result related to the corresponding index based on the image parameters include: The fitted parameters are machine-learned using a trained second machine learning model to output a judgment result on whether motion artifacts exist.

16. The method according to claim 7, wherein, The steps for processing the magnetic resonance image include: The magnetic resonance image is subjected to homogeneity correction to obtain a corrected image; and, Determine the difference between the magnetic resonance image and the corresponding corrected image; The steps for determining the image quality assessment result related to the corresponding index based on the image parameters include: The shading intensity of the magnetic resonance image is determined based on the difference.

17. The method according to claim 1, wherein, Also includes: An analysis report relating to the performance of the magnetic resonance imaging system is generated based on the image quality assessment results.

18. A computer-readable storage medium comprising a stored computer program, wherein, When the computer program is run, it performs the method for evaluating the quality of magnetic resonance images according to any one of claims 1 to 17.

19. A magnetic resonance imaging system, comprising: An image generating apparatus for performing magnetic resonance scanning to generate magnetic resonance images; as well as A processor for performing the method of magnetic resonance image quality assessment as described in any one of claims 1 to 17.

20. The system according to claim 19, wherein, It also includes a memory that stores a pre-established database, the database including magnetic resonance image association information, the magnetic resonance image association information including one or more of the following: time information of the image generation device generating the magnetic resonance image, scanning parameters when performing the magnetic resonance scan, human anatomical information of the person performing the magnetic resonance scan, and product information of the person performing the magnetic resonance scan.

21. The system according to claim 19, wherein, The processor includes a result output module, which is used to receive magnetic resonance image association information determined by the user through a user interface, and output the image quality assessment result of the corresponding magnetic resonance image through the user interface based on the magnetic resonance image association information.

22. The system according to claim 21, wherein, The result output module is also used to receive the image quality evaluation index determined by the user through the user interface, and output the image quality evaluation result related to the determined index.

Images