Medical imaging method and system, non-transitory computer readable storage medium

Abstract

The application provides a medical imaging method and system, and a non-transitory computer readable storage medium. The medical imaging method comprises identifying an image quality type of a medical image based on a trained learning network, and generating a corresponding control signal for controlling a medical imaging system based on the identified image quality type.

Description

Technical Field

[0001] This invention relates to medical imaging technology, and more specifically to a medical imaging method and system, as well as a non-transitory computer-readable storage medium. Background Technology

[0002] Magnetic resonance imaging (MRI), a medical imaging modality, can acquire images of the human body without using X-rays or other ionizing radiation. MRI utilizes a magnet with a strong magnetic field to generate a static magnetic field B0. When the area to be imaged is positioned within the static magnetic field B0, the nuclear spins associated with the hydrogen nuclei in the tissue become polarized, resulting in a longitudinal magnetization vector on a macroscopic scale. When a radio frequency field B1 is applied, intersecting the direction of the static magnetic field B0, the direction of proton rotation changes, resulting in a transverse magnetization vector on a macroscopic scale. After the radio frequency field B1 is removed, the transverse magnetization vector decays in a spiral pattern until it returns to zero. During this decay process, a free-inductance decay signal is generated, which can be acquired as an MRI signal. Based on this acquired signal, an image of the tissue area to be imaged can be reconstructed.

[0003] During the imaging process, due to hardware problems such as equipment issues or human movement, the reconstructed images often suffer from image quality problems (e.g., artifacts). For example, Nyquist (N / 2) artifacts are caused by short eddies due to phase difference, acceleration artifacts are caused by accelerated acquisition, and motion artifacts are caused by voluntary or involuntary human movement during acquisition. These artifacts can degrade image quality and may fail to accurately represent the actual situation of lesions, making clinical diagnosis and analysis difficult.

[0004] In practice, when artifacts in medical images affect doctors' diagnoses, it is necessary to rely on on-site operators or engineers to judge the type of artifact based on their experience, and then determine the cause of the artifact based on the type of artifact and take relevant actions. However, sometimes it is not possible to judge the type of artifact in a timely and accurate manner, which brings many difficulties to clinical practice. Summary of the Invention

[0005] This invention provides a medical imaging method and system, as well as a non-transitory computer-readable storage medium.

[0006] An exemplary embodiment of the present invention provides a medical imaging method, the method comprising a trained learning network for identifying image quality types of medical images and generating corresponding control signals for controlling a medical imaging system based on the identified image quality types.

[0007] An exemplary embodiment of the present invention also provides a non-transitory computer-readable storage medium for storing a computer program that, when executed by a computer, causes the computer to perform the instructions described above for the medical imaging method.

[0008] An exemplary embodiment of the present invention also provides a medical imaging system, the system including an identification module and a control module. The identification module is used to identify the image quality type of medical images generated by the medical imaging system based on a trained learning network, and the control module is used to generate corresponding control signals for controlling the medical imaging system based on the identified image quality type.

[0009] Other features and aspects will become clear from the following detailed description, accompanying drawings, and claims. Attached Figure Description

[0010] The invention can be better understood by describing exemplary embodiments of the invention in conjunction with the accompanying drawings, in which:

[0011] Figure 1 This is a flowchart of a medical imaging method according to some embodiments of the present invention;

[0012] Figure 2 This is a schematic diagram of a medical image that includes Nyquist artifacts;

[0013] Figure 3 This is a schematic diagram of a medical image that includes acceleration artifacts;

[0014] Figure 4 This is a schematic diagram of a learning network according to some embodiments of the present invention;

[0015] Figure 5 yes Figure 1 The flowchart shown illustrates a medical imaging method that generates corresponding control signals for controlling the medical imaging system based on the identified image quality type.

[0016] Figure 6 A schematic diagram of a magnetic resonance imaging system according to some embodiments of the present invention is shown;

[0017] Figure 7 A schematic diagram of a medical imaging system according to some embodiments of the present invention is shown; and

[0018] Figure 8 A schematic diagram of a medical imaging system according to other embodiments of the present invention is shown. Detailed Implementation

[0019] 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.

[0020] 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 to which this invention pertains. The terms "first," "second," and similar terms used in the patent application description 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 element or object preceding "comprising" or "including" encompasses the element or object listed following "comprising" or "including" and its 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.

[0021] In some embodiments, during the imaging process, different hardware problems of the device or different factors such as human movement may cause different artifacts in the reconstructed image. The medical imaging methods and systems in some embodiments of the present invention are based on deep learning technology, which can accurately determine the image quality type (e.g., artifacts) in the image and automatically generate control signals for controlling the magnetic resonance imaging (MRI) system based on the image quality type, such as to calibrate the MRI system or issue warning signals. It should be noted that, from the perspective of those skilled in the art or related fields, such description should not be construed as limiting the present invention only to MRI systems. In fact, the medical imaging methods and systems described herein can be reasonably applied to other imaging fields in the medical and non-medical fields, such as CT systems, PET systems, SPECT systems, X-ray machines, or any combination thereof.

[0022] As discussed in this paper, deep learning techniques (also known as deep machine learning, hierarchical learning, or deep structured learning) employ artificial neural networks for learning. 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 processing layers (e.g., input layers, output layers, convolutional layers, normalization layers, sampling layers, etc., with varying numbers and functions depending on the deep learning network model). The configuration and number of layers allow deep learning 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 parameters learned or trained 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 into features exhibiting higher complexity. 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.

[0023] As discussed herein, as part of the initial training of a deep learning process for solving a specific problem, the training dataset includes known input values ​​(e.g., sample images or pixel matrices of images after coordinate transformation) and the expected (target) output values ​​of the final output of the deep learning process (e.g., an image or recognition judgment result). In this way, the 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. Stochastic gradient descent (SGD) methods are typically used to update the network parameters; 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 learning 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 learning network, and then the network output is compared with the (known) desired output to validate previous training and / or prevent overtraining.

[0024] Figure 1 A flowchart of a medical imaging method 100 according to some embodiments of the present invention is shown. Figure 2 A schematic diagram of a medical image including Nyquist artifacts is shown. Figure 3 A schematic diagram of a medical image including acceleration artifacts is shown. (e.g.) Figure 1-3 As shown, the medical imaging method 100 in some embodiments of the present invention includes steps 110 and 120.

[0025] In step 110, the image quality type of the medical image is identified based on the trained learning network.

[0026] In some embodiments, the image quality type described above may include one or more artifact types, such as Nyquist artifacts caused by phase-difference short eddies (e.g., Figure 2 As shown), due to acceleration artifacts caused by accelerated acquisition (such as... Figure 3 The image quality issues mentioned above include motion artifacts caused by voluntary or involuntary human movement during the acquisition process (not shown in the figure), and truncation artifacts caused by missing or truncated data (not shown in the figure). In some embodiments, the above-mentioned image quality problems may also include one or more non-artifact types, such as image quality problems caused by signal-to-noise ratio, resolution, contrast, etc.

[0027] In some embodiments, identifying image quality types in medical images includes analyzing the degree of matching between artifacts (or non-artifacts) in the medical image and at least one artifact type (or at least one non-artifact type). The degree of matching includes labels and their probabilities, where the probability is the probability that an artifact in the image to be judged is a specific type of artifact. For example, the medical image to be judged is input into a learning network for recognition, and the output is a predefined label representing each artifact and the probability of the identified artifact. For example, the output might be: Label 1 - Nyquist artifact 95%; Label 2 - Motion artifact 0%; Label 3 - Acceleration artifact 0%; Label 4 - Truncation artifact 0%, etc. In some embodiments, identifying image quality types in medical images also includes outputting artifact types whose degree of matching with the medical image is greater than a preset value or whose degree of matching is within a preset ranking. For example, only the following might be output: Label 1 - Nyquist artifact 95%, or Nyquist artifact 95%.

[0028] Although the above lists the methods for outputting artifact types, those skilled in the art should know that artifact types can be output in other ways, and are not limited to the methods described above. For example, label codes or specific artifact types or any combination thereof can be output.

[0029] Furthermore, although only four types of artifacts are listed in the above embodiments, those skilled in the art should know that there are far more than four types of artifacts in actual imaging processes. Moreover, the medical imaging method in the embodiments of the present invention is not limited to recognizing only the above four types of artifacts. Various image quality types can be input into the training phase of the learning network to enable the learning network to recognize different image quality types.

[0030] The learning network is completed through the preparation of training data, the selection and construction of the learning network model, and the training, testing, and optimization of the learning network.

[0031] In some embodiments, the learning network is trained on a dataset of sample images (known input) and their corresponding image quality types (expected output). Specifically, the training of the learning network includes steps one to three.

[0032] Step 1: Acquire multiple medical images with image quality issues (e.g., artifacts and / or non-artifact quality issues) as a sample image set. In some embodiments, the sample images are obtained after preprocessing medical images reconstructed from a medical imaging system, where preprocessing includes normalization. In other embodiments, unnormalized medical images can also be used as the sample image set, and the sample image set can be input into a learning network, whereby the sample images are normalized based on the normalization layer in the learning network. In some embodiments, the sample images are medical images obtained under multiple specific imaging conditions of a medical imaging system. These specific imaging conditions may include, for example: a hardware configuration or scanning procedure that facilitates the acquisition of Nyquist artifacts, another hardware configuration or scanning procedure that facilitates the acquisition of acceleration artifacts, scanning under motion conditions of the imaging site (e.g., the human heart) that facilitates the acquisition of motion artifacts, or a combination of multiple imaging conditions that includes two or more different types of artifacts.

[0033] Step 2: Obtain the image quality type (which can be represented by labels) for each image in the sample image set as an image quality type set. In some embodiments, the image quality type can be determined in advance based on experience. In some embodiments, the image quality type corresponding to the sample image can be represented by labels, for example, label 1 represents Nyquist artifacts, label 2 represents motion artifacts, label 3 represents acceleration artifacts, label 4 represents truncation artifacts, etc. Furthermore, label 0 can be set to represent non-artifact types, or different labels can be set to represent different non-artifact quality types. However, those skilled in the art will understand that the above image quality types are not limited to the above representation methods, and any suitable method can be used. The method of labeling and the content it represents can also be changed.

[0034] Step 3: Using the sample image set as input and the image quality type set as output, train the learning network to obtain the trained learning network. In some embodiments, the trained learning network can be updated based on a new sample image set and its corresponding image quality type set, wherein the image quality type set can be obtained in advance based on experience, or it can be obtained based on the trained learning network described above.

[0035] In some embodiments, the learning network is trained on ResNet (Residual Network) or VGGNet (Visual Geometry Group Network) or other well-known models. Since the number of processing layers in ResNet can be set to be large (up to 1000 or more), the classification based on this network structure (e.g., artifact type determination) performs better accordingly. Furthermore, ResNet is easier to optimize based on more training data.

[0036] Figure 4 A learning network 150 according to some embodiments of the present invention is illustrated. For example... Figure 4 As shown, the learning network 150 includes an input layer 151, a processing layer (or hidden layer) 152, and an output layer 153. In some embodiments, such as Figure 4 As shown, the processing layer 152 includes a first convolutional layer 155, a first pooling layer 156, and a fully connected layer 157. The first convolutional layer 155 convolves the input image (or pixels) to obtain a feature map of the first convolutional layer. The first pooling layer 156 pools (downsamples) the feature map of the first convolutional layer to compress the feature map and extract the main features, thus obtaining a feature map of the first pooling layer. The fully connected layer 157 can output a judgment result based on the feature map of the first pooling layer.

[0037] Although Figure 4 Only one example of a convolutional layer is shown. In other examples, the number of convolutional layers can be arbitrary. The number of convolutional layers can be adaptively adjusted according to the size of the input data in the learning network (e.g., the number of pixels in an image). For example, a second convolutional layer and a second pooling layer (not shown in the figure) may be added between the first pooling layer 156 and the fully connected layer 157, or a second convolutional layer and a second pooling layer, and a third convolutional layer and a third pooling layer (not shown in the figure) may be added between the first pooling layer 156 and the fully connected layer 157, and so on.

[0038] Although Figure 4 Only the connection between the convolutional layer and the input layer, the connection between the pooling layer and the convolutional layer, and the connection between the fully connected layer and the pooling layer are shown. In other examples, any number of processing layers of any type can be set between any two of the above layers. For example, a normalization layer can be set between the convolutional layer and the input layer to normalize the input image (or pixels), or an activation layer can be set between the fully connected layer and the pooling layer to perform non-linear mapping on the feature map of the pooling layer using the Rectified Linear Unit (ReLU) activation function.

[0039] In some embodiments, each layer includes a plurality of neurons 160, and the number of neurons in each layer can be the same or different as needed. By inputting a sample dataset (known input) and a set of image quality types (desired output) into the learning network, and by setting the number of processing layers and the number of neurons in each processing layer, and estimating (or adjusting or calibrating) the weights and / or biases of the learning network, the mathematical relationship between the known input and the desired output and / or the mathematical relationship between the input and output of each layer is identified and represented. The learning process typically utilizes (partial) input data and creates a network output for that input data. Then, the network output created based on the known input is compared with the desired output of the dataset; the difference is the loss function. The network parameters (weights and / or biases) can be iteratively updated using the loss function, continuously reducing the loss function and thus training a neural network model with higher accuracy. In some embodiments, many functions can be used as the loss function, including but not limited to mean squared error, cross-entropy error, etc. Once the learning network is created or trained, simply inputting the medical image to be judged into the learning network will allow us to obtain information such as the image quality type, or the label and matching degree related to that image quality type.

[0040] In one embodiment, while the configuration of the learning network 150 is guided by prior knowledge of the estimation problem, dimensions of the inputs, outputs, etc., the learning itself is treated as a "black box" and relies primarily on, or specifically on, achieving the best approximation of the desired output data based on the input data. In various alternative implementations, certain aspects and / or features of the data, imaging geometry, reconstruction algorithms, etc., can be leveraged to give explicit meaning to certain data representations in the learning network 150. This can help accelerate training because it creates the opportunity to train (or pre-train) or define certain layers in the learning network 150 separately.

[0041] Please continue to refer to this. Figure 1 In step 120, based on the identified image quality type, a corresponding control signal for controlling the medical imaging system is generated.

[0042] Figure 5 A flowchart illustrating the process of generating corresponding control signals for controlling the medical imaging system based on the identified image quality type (step 120) in a medical imaging method according to some embodiments of the present invention is shown. Figure 5 As shown, by inputting the medical image to be judged into the learning network, the artifact type of the medical image can be obtained.

[0043] When the medical image is identified as an acceleration artifact in the magnetic resonance imaging (MRI) image, a first control signal is generated to control the medical imaging system to issue a warning signal adjusting the acceleration factor. When the medical image is identified as a motion artifact in the MRI image, a second control signal is generated to control the medical imaging system to issue a warning signal regarding motion of the detected object.

[0044] In some embodiments, the warning signal can be transmitted via the display unit of a medical imaging system (such as...). Figure 6 The display unit 260 in the illustrated MRI system shows that its graphical user interface can display alarm indicators, such as those represented by colors, symbols, etc. The graphical user interface can display, for example, artifact types and corresponding correction methods, adjustments to certain scan coefficients in the medical imaging system, or other relevant information. In other embodiments, the warning signal can be issued by an alarm device of the medical imaging system, which can be, for example, a horn, an alarm, or any other known type of alarm device.

[0045] When the medical image is identified as a Nyquist artifact in a magnetic resonance imaging (MRI) image, a third control signal is generated to control the medical imaging system to initiate a calibration mode. In some embodiments, when the medical image is identified as a Nyquist artifact, a threshold comparison is further performed on the matching degree (e.g., probability) output by the learning network. If the matching degree is higher than a preset threshold (e.g., 80%), a third control signal is generated to control the medical imaging system (to stop scanning and) initiate a calibration mode. In some embodiments, the calibration mode includes adjustments or compensation to the gradient system. Those skilled in the art should understand that the above description is not limited to threshold comparisons of matching degree only for Nyquist artifacts; the same or similar comparisons can be performed for other artifact types or image quality types. Furthermore, those skilled in the art should understand that embodiments of the present invention are not limited to automatically initiating a calibration mode when a Nyquist artifact is identified. Alternatively, when a Nyquist artifact is identified, the medical imaging system may perform human-computer interaction based on alarms, inquiries, or other signals, and initiate or activate the calibration mode based on the interaction results.

[0046] In some embodiments, the control signal can be set to perform different functions based on different artifact types, and is not limited to the two functions mentioned above. It can also have any other functions, such as self-testing of the medical imaging system, historical data analysis of imaging images, etc.

[0047] In some embodiments, prior to step 110 (identifying the type of image quality in the medical image based on the trained learning network), the medical image generated by the medical imaging system is received based on user instructions. The user instructions can be received via the operation console unit of the medical imaging system (e.g., Figure 6 The operator console unit 250 in the illustrated MRI system provides input, and further, the controller unit (or other control device or computer) can receive user commands sent by the operator console unit. In some embodiments, user commands can also be transmitted via buttons or keys (such as buttons on the display unit of the medical imaging system) Figure 6 The operation console unit 260 of the MRI system shown is used as an input. Furthermore, the user commands sent by the display unit can be received through the controller unit (or other control device or computer). For example, the user can select to perform the above steps 110 and 120 on the image based on a prediction of the medical image (e.g., believing that it will affect the diagnosis). In this case, the image is sent to the module for identifying the image quality type before performing step 110, based on the user's operation command.

[0048] In some embodiments, a learning network can be trained using sample image sets of different types of artifacts and their corresponding artifact type sets. This learning network not only has low requirements for the amount of training dataset, but can also accurately identify the types of artifacts in medical images obtained by medical imaging systems. However, this learning network also has some problems. For example, if a medical image that is mistakenly identified as having artifacts (due to image quality problems caused by other reasons, such as image quality problems caused by signal-to-noise ratio, contrast, resolution, etc.) is input into the learning network, although the matching degree of the learning network output will be low, there will still be cases of misjudgment. Therefore, further, during the training and / or update phase of the learning network, some images without artifacts but with other quality problems can also be input. The corresponding image quality type can be set to, for example, label 0 - no artifacts. With the above settings, when a user finds that the medical image quality is poor but believes it is an artifact, the medical imaging method provided in this embodiment of the invention can be activated. The learning network can identify and output a control signal of "no artifacts or no artifacts" to control the medical imaging system to issue a warning signal (or, for example, a label) to the doctor or activate the corresponding calibration mode. On the one hand, this makes it easier for doctors to investigate other causes of image quality degradation or improve such image quality problems. On the other hand, it can also better deal with situations where doctors accidentally activate or misoperate the system.

[0049] Furthermore, during the training and / or update phase of the learning network, the input includes sample images of different non-artifact types, and the corresponding image quality types can be set as different labels, such as label 11 - signal-to-noise ratio problem, label 12 - contrast problem, label 13 - resolution problem, etc. With the above settings, when the user finds that the medical image quality is poor, the medical imaging method provided in this embodiment of the invention can be activated to identify the image quality problem without prior judgment. In addition, the corresponding control signals can be output by the learning network to control the medical imaging system to issue warning signals (or labels) to the doctor or activate different calibration modes (e.g., adjust the scanning time or scanning parameters, etc.).

[0050] The deep learning-based medical imaging method proposed in this invention can more accurately and timely determine the image quality type of medical images, such as different artifacts, without relying on the experience judgment of on-site engineers. In addition, based on the determined image quality type, control signals can be generated to control the medical imaging system, so as to help users understand the objective reasons for the formation of image quality problems in a timely manner, and further adjust or calibrate the medical imaging system, which can solve image quality problems more quickly.

[0051] Figure 6 A schematic diagram of an MRI system 200 according to some embodiments of the present invention is shown. Figure 6 As shown, the MRI system 200 includes a scanner 210, a controller unit 220, and a data processing unit 230. The MRI system 200 described above is only an example; in other embodiments, the MRI system 200 can have various variations, as long as it can acquire image data from the object being examined.

[0052] The scanner 210 can be used to acquire data of the object 216 being inspected. The controller unit 220 is coupled to the scanner 210 to control the operation of the scanner 210. The scanner 210 may include a main magnet 211, an RF transmitting coil 212, an RF generator 213, a gradient coil system 217, a gradient coil driver 218, and an RF receiving coil 219.

[0053] The main magnet 211 typically includes, for example, a toroidal superconducting magnet mounted within a toroidal vacuum container. This toroidal superconducting magnet defines a cylindrical space surrounding the object under test 216 and generates a constant static magnetic field, such as a static magnetic field B0, along the Z-direction of the cylindrical space. The MRI system 200 utilizes the generated static magnetic field B0 to transmit a magnetic field pulse signal to the object under test 216 placed in the imaging space, thereby ordering the precession of protons within the object under test 216 and generating a longitudinal magnetization vector.

[0054] Radio frequency (RF) generator 213 generates RF pulses, which may include RF excitation pulses. These RF excitation pulses are amplified (e.g., by an RF power amplifier (not shown)) and applied to RF transmitting coil 212. This causes RF transmitting coil 212 to emit an RF magnetic field B1 orthogonal to the static magnetic field B0 towards the object under test 216, exciting the atomic nuclei within the object under test 216 and transforming the longitudinal magnetization vector into a transverse magnetization vector. As the RF excitation pulse ends, the transverse magnetization vector of the object under test 216 gradually returns to zero, generating a free-induction decay signal, i.e., a magnetic resonance signal that can be acquired.

[0055] The radio frequency transmitting coil 212 can be a body coil, which can be connected to a transmit / receive (T / R) switch (not shown). By controlling the transmit / receive switch, the body coil can be switched between transmit and receive modes. In the receive mode, the body coil can be used to receive magnetic resonance signals from the object under test 216.

[0056] The gradient coil system 217 generates a gradient magnetic field in the imaging space to provide three-dimensional position information for the aforementioned magnetic resonance signal. This magnetic resonance signal can be received by the radio frequency receiving coil 219 or by the body coil in receiving mode. The data processing unit 230 can process the received magnetic resonance signal to obtain the desired image or image data.

[0057] Specifically, the gradient coil system 217 may include three gradient coils, each of which generates a gradient magnetic field tilted to one of three mutually perpendicular spatial axes (e.g., the X-axis, Y-axis, and Z-axis), and generates gradient fields in each of the slice selection direction, phase encoding direction, and frequency encoding direction according to the imaging conditions. More specifically, the gradient coil system 217 applies a gradient field in the slice selection direction of the object under test 216 to select a slice; and the radio frequency transmission coil 212 transmits a radio frequency excitation pulse to the selected slice of the object under test 216 and excites the slice. The gradient coil system 217 also applies a gradient field in the phase encoding direction of the object under test 216 to perform phase encoding of the magnetic resonance signal of the excited slice. The gradient coil system 217 then applies a gradient field in the frequency encoding direction of the object under test 216 to perform frequency encoding of the magnetic resonance signal of the excited slice.

[0058] The gradient coil driver 218 is used to provide appropriate power signals to the three gradient coils respectively in response to the sequence control signals issued by the controller unit 230.

[0059] The scanner 210 may further include a data acquisition unit 214 for acquiring magnetic resonance signals received by the radio frequency surface coil 219 or the volume coil. The data acquisition unit 214 may include, for example, a radio frequency preamplifier (not shown), a phase detector (not shown), and an analog-to-digital converter (not shown). The radio frequency preamplifier amplifies the magnetic resonance signals received by the radio frequency surface coil 219 or the volume coil, the phase detector performs phase detection on the amplified magnetic resonance signals, and the analog-to-digital converter converts the phase-detected magnetic resonance signals from analog signals to digital signals. The digitized magnetic resonance signals can be processed by the data processing unit 230 through calculations, reconstruction, etc., to obtain medical images, such as the medical images requiring image quality type identification described in the embodiments of the present invention.

[0060] The data processing unit 230 may include a computer and a storage medium on which a predetermined data processing program to be executed by the computer is recorded. The data processing unit 230 may be connected to the controller unit 220 and perform data processing based on control signals received from the controller unit 220. The data processing unit 230 may also be connected to the data acquisition unit 214 to receive magnetic resonance signals output by the data acquisition unit 214 in order to perform the aforementioned data processing.

[0061] The controller unit 220 may include a computer and a storage medium for storing a program executable by the computer. When the computer executes the program, it can cause multiple components of the scanner 210 to perform operations corresponding to the imaging sequence described above. The data processing unit 230 can also perform predetermined data processing. When the computer executes the program, it can also execute the medical imaging method described above to identify the image quality type of the medical image and output control signals for controlling any component or module of the MRI system 200. For example, the control signals can appropriately calibrate the MRI system 200 according to the identified image quality type; for example, for one image quality problem, gradient-related parameters can be adjusted, and for another image quality problem, radio frequency-related parameters can be adjusted.

[0062] The storage media of the controller unit 220 and the data processing unit 230 may include, for example, ROM, floppy disk, hard disk, optical disk, magneto-optical disk, CD-ROM, or non-volatile memory card.

[0063] Controller unit 220 can be configured and / or arranged for use in different ways. For example, in some implementations, a single controller unit 220 may be used; in other implementations, multiple controller units 220 are configured to work together (e.g., based on a distributed processing configuration) or individually, each controller unit 220 being configured to handle specific aspects and / or functions, and / or process data for generating models that are only used for a specific medical imaging system 200. In some implementations, controller unit 220 may be local (e.g., co-located with one or more medical imaging systems 200, such as within the same facility and / or the same local network); in other implementations, controller unit 220 may be remote and therefore accessible only via a remote connection (e.g., via the Internet or other available remote access technologies). In a particular implementation, controller unit 220 may be configured in a cloud-like manner and may be accessed and / or used in a manner substantially similar to that used for accessing and using other cloud-based systems.

[0064] The MRI system 200 also includes a stage 240 for placing the object 216 to be examined thereon. The object 216 can be moved into or out of the imaging space by moving the stage 240 based on control signals from the controller unit 220.

[0065] The MRI system 200 also includes an operation console unit 250 connected to the controller unit 220. The operation console unit 250 can send acquired operation signals to the controller unit 220 to control the operating status of components such as the stage 240 and the scanner 210. These operation signals may include, for example, scanning protocols and parameters selected manually or automatically. The scanning protocol may include the aforementioned imaging sequence. Furthermore, the operation console unit 250 can send acquired operation signals to the controller unit 220 to control the data processing unit 230 to obtain the desired image.

[0066] The operation console unit 250 may include a user input device, such as a keyboard, mouse, voice-activated controller or any other suitable input device, in the form of an operator interface, through which the operator can input operation signals / control signals to the controller unit 220.

[0067] The MRI system 200 may also include a display unit 260, which can be connected to the operation control unit 250 to display the operation interface, and can also be connected to the data processing unit 230 to display images. The display unit 260 can also display the identified image quality type and corresponding alarm signals.

[0068] In some embodiments, system 200 may be connected to one or more display units, cloud networks, printers, workstations and / or similar devices located locally or remotely via one or more configurable wired and / or wireless networks, such as the Internet and / or virtual private networks.

[0069] Figure 7 A schematic diagram of a medical imaging system 300 according to some embodiments of the present invention is shown. Figure 7 As shown, the medical imaging system 300 includes an identification module 310 and a control module 320.

[0070] The recognition module 310 is used to identify the image quality type of medical images generated by the medical imaging system based on a trained learning network. In some embodiments, the recognition module 310 is configured with a network such as... Figure 6 The controller unit 220 in the MRI system 200 shown is connected to or is part of the controller unit 220.

[0071] In some embodiments, the image quality type may include one or more artifact types, such as Nyquist artifacts, acceleration artifacts, motion artifacts, etc. In some embodiments, the image quality problem may also include one or more non-artifact types, such as image quality problems caused by signal-to-noise ratio, resolution, contrast, etc. In some embodiments, the recognition module 310 may output artifacts (or non-artifacts) in the medical image and the degree of matching between the artifacts (or non-artifacts) and at least one artifact type (or non-artifact type). In some embodiments, the degree of matching includes labels and their probabilities. For example, when a medical image of the artifact type to be determined is input into a learning network, the learning network can output the artifact type in the image to be determined, and the output is the predefined label represented by each artifact and the probability of the identified artifact. In some embodiments, the recognition module 310 is further configured to output artifact types whose degree of matching with the medical image is greater than a preset value or whose degree of matching is within a preset ranking.

[0072] The training network described above is trained on an external carrier (e.g., a device other than a medical imaging system). In some embodiments, the training system trains the learning network based on a dataset of sample images and their corresponding image quality types, using ResNet, VGGNet, or other known models. 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 network for connecting the first and second modules. In some embodiments, the first module includes a first processing unit and a first storage unit, wherein the first storage unit stores the training dataset, and the first processing unit receives relevant instructions (e.g., retrieve training dataset) and sends the training dataset according to the instructions. Furthermore, the second module includes a second processing unit and a second storage unit, wherein the second storage unit stores the training model, and the second processing unit receives relevant instructions, trains and / or updates the learning 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 network may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0073] Once data (e.g., a trained learning network) is generated and / or configured, it can be copied and / or loaded into the MRI system 200, which can be done in various ways. For example, it can be loaded via a directional connection or link between the MRI system 200 and the controller unit 220. In this regard, 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 or additionally, data can be loaded into the MRI system 200 indirectly. For example, data can be stored on a suitable machine-readable medium (e.g., a flash memory card, etc.) and then loaded into the MRI system 200 (in the field, such as by the system's user or authorized personnel) using that medium, 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 the system's user or authorized personnel) to upload the data to the MRI system 200 via a direct connection (e.g., a USB connector, etc.).

[0074] The control module 320 is used to generate corresponding control signals for controlling the medical imaging system based on the identified image quality type. In some embodiments, the control module 320 is used in conjunction with, for example, Figure 6 The controller unit 220 in the MRI system 200 shown is connected to or is part of the controller unit 220 to control the medical imaging system to issue warning signals or perform calibration based on the type of image quality.

[0075] Image quality types include one or more artifact types. In some embodiments, control module 320 includes a first control unit (not shown) for generating a first control signal based on the acceleration artifact type in the magnetic resonance image output by the recognition module, to control the medical imaging system to issue a warning signal to adjust the acceleration factor.

[0076] The control module 320 includes a second control unit (not shown in the figure), which generates a second control signal based on motion artifacts in the magnetic resonance image output by the recognition module, so as to control the medical imaging system to issue a warning signal about the motion of the detected object.

[0077] In some embodiments, the warning signal can be transmitted via the display unit of a medical imaging system (such as...). Figure 6 The MRI system (display unit 260) shown in the diagram displays the signal, or an alarm is triggered by the medical imaging system.

[0078] The control module 320 includes a third control unit (not shown in the figure), which generates a third control signal based on the Nyquist artifact in the magnetic resonance image output by the recognition module to control the medical imaging system to start the calibration mode.

[0079] In some embodiments, the control module further includes a matching degree comparison unit connected to the third control unit, which is used to compare the matching degree output by the recognition module 310 with a preset threshold. The third control unit generates the third control signal based on the Nyquist artifact output by the recognition module 310 and the comparison result. Specifically, when the matching degree output by the recognition module exceeds the preset threshold, the third control unit generates the third control signal.

[0080] In some embodiments, the medical imaging system 300 further includes a receiving module 340 for receiving the medical images generated by the medical imaging system based on user instructions. Further, the receiving module 340 is configured to... Figure 6 The operator console unit 250 in the MRI system 200 shown is connected to receive user commands. Optionally, the receiving module 340 is connected to the controller unit 220 (e.g., Figure 6 (As shown) is connected to receive commands sent by the user through the operator console unit 250 via the controller unit 220.

[0081] Figure 8 A schematic diagram of a medical imaging system 400 according to other embodiments of the present invention is shown. For example... Figure 8 As shown, with Figure 7Unlike the medical imaging system 300 shown in some embodiments, the medical imaging system 400 further includes a training module 430. The training module 430 trains the learning network based on a dataset of sample images and their corresponding image quality types, using a ResNet, VGGNet, or other known models. In some embodiments, the training dataset is stored in the storage medium of the controller unit 220 of the MRI system 200, and the training module 430 can use the training dataset to train and / or update the learning network.

[0082] The present invention may also provide a non-transitory computer-readable storage medium for storing an instruction set and / or a computer program that, when executed by a computer, causes the computer to perform the above-described method for acquiring a truncated portion of the predicted image. The computer executing the instruction set and / or the computer program may be a computer of an MRI system or other devices / modules of an MRI system. In one embodiment, the instruction set and / or the computer program may be programmed into the processor / controller of the computer.

[0083] Specifically, when this instruction set and / or computer program is executed by the computer, it causes the computer to:

[0084] A training-based learning network identifies image quality types in medical images; and

[0085] Based on the identified image quality type, corresponding control signals are generated for controlling the medical imaging system.

[0086] As described above, instructions can be combined into a single instruction for execution, or any instruction can be split into multiple instructions for execution. Furthermore, the execution order of instructions is not limited to that described above.

[0087] In some embodiments, the medical image generated by the medical imaging system is received based on user instructions before identifying the type of image quality in the medical image.

[0088] As used herein, the term "computer" can include any processor-based or microprocessor-based system, including systems that use microcontrollers, reduced instruction set computers (RISC), application-specific integrated circuits (ASICs), logic circuits, and any other circuitry or processors capable of performing the functions described herein. The examples above are merely illustrative and are not intended to limit the definition and / or meaning of the term "computer" in any way.

[0089] The instruction set may include various commands that instruct a computer or processor, acting as a processor, to perform specific operations, such as methods and processes according to various embodiments. The instruction set may take the form of a software program, which may be part of one or more tangible, non-transitory computer-readable media. The software may take various forms, such as system software or application software. Furthermore, the software may take the form of a collection of independent programs or modules, a program module within a larger program, or part of a program module. The software may also include modular programming in the form of object-oriented programming. Input data processing by the processor may be in response to operator commands, previous processing results, or requests made by another processor.

[0090] Some exemplary embodiments have been described above; however, it should be understood that various modifications can be made. For example, suitable results may be achieved if the described techniques are performed in a different order and / or if components in the described system, architecture, device, or circuit are combined in a different manner and / or replaced or supplemented by other components or their equivalents. Accordingly, other embodiments also fall within the scope of the claims.

Claims

1. A medical imaging method, comprising: A training-based learning network identifies the image quality type of medical images; as well as Based on the identified image quality type, corresponding control signals are generated for controlling the medical imaging system. The image quality type includes one or more artifact types; Specifically, when the medical image is identified as an acceleration artifact in a magnetic resonance imaging (MRI) image, a first control signal is generated to control the medical imaging system to issue a warning signal to adjust the acceleration factor. The image quality type includes one or more non-artifact types.

2. The method as described in claim 1, wherein, Prior to identifying the image quality type in the medical image, the process also includes receiving the medical image generated by the medical imaging system based on user instructions.

3. The method as described in claim 1, wherein, The learning network is trained on a dataset of sample images and their corresponding image quality types.

4. The method of claim 1, wherein, Identifying the image quality type of a medical image includes analyzing the degree to which artifacts in the medical image match one or more of the artifact types.

5. The method of claim 4, wherein, The image quality type for identifying medical images also includes artifact types where the output matches the medical image to a degree greater than a preset value or the degree of matching is within a preset ranking.

6. The method of claim 1, wherein, When the medical image is identified as a motion artifact in the magnetic resonance image, a second control signal is generated to control the medical imaging system to issue a warning signal regarding the motion of the detected object.

7. The method of claim 1, wherein, When the medical image is identified as a Nyquist artifact in a magnetic resonance imaging (MRI) image, a third control signal is generated to control the medical imaging system to activate calibration mode.

8. The method of claim 4, wherein, When the medical image is identified as a Nyquist artifact in a magnetic resonance imaging image, and the matching degree exceeds a preset threshold, a third control signal is generated to control the medical imaging system to start the calibration mode.

9. A non-transitory computer-readable storage medium for storing a computer program that, when executed by a computer, causes the computer to perform the medical imaging method according to any one of claims 1-8.

10. A medical imaging system comprising: A recognition module is used to identify the image quality type of medical images generated by the medical imaging system based on a trained learning network. as well as The control module is used to generate corresponding control signals for controlling the medical imaging system based on the identified image quality type. The image quality type includes one or more artifact types; The control module includes a first control unit, which generates a first control signal based on the type of acceleration artifact in the magnetic resonance image output by the identification module, to control the medical imaging system to issue a warning signal for adjusting the acceleration factor. The image quality type includes one or more non-artifact types.

11. The system of claim 10, wherein, Further includes: The training module is used to train the learning network based on a dataset of sample images and their corresponding image quality types.

12. The system of claim 10, wherein, Further includes: A receiving module is used to receive the medical images generated by the medical imaging system based on user instructions.

13. The system of claim 10, wherein, The recognition module outputs the artifacts in the medical image, as well as the degree of matching between the artifacts and one or more artifact types.

14. The system of claim 10, wherein, The control module includes a second control unit, which generates a second control signal based on motion artifacts in the magnetic resonance image output by the recognition module, to control the medical imaging system to issue a warning signal regarding the motion of the detected object.

15. The system of claim 10, wherein, The control module includes a third control unit, which generates a third control signal based on the Nyquist artifact in the magnetic resonance image output by the recognition module, to control the medical imaging system to start the calibration mode.

16. The system of claim 15, wherein, The control module further includes a matching degree comparison unit connected to the third control unit, which is used to compare the matching degree output by the recognition module with a preset threshold. The third control unit generates the third control signal based on the Nyquist artifact output by the recognition module and the comparison result.

Images