Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.
In the description of the present invention, it should be understood that the orientation or positional relationship referred to in the description of the orientation, such as the upper, lower, front, rear, left, right, etc., is based on the orientation or positional relationship shown in the drawings, and is only for convenience of description and simplification of description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention.
In the description of the present invention, the meaning of a plurality of means is one or more, the meaning of a plurality of means is two or more, and larger, smaller, larger, etc. are understood as excluding the number, and larger, smaller, inner, etc. are understood as including the number. If the first and second are described for the purpose of distinguishing technical features, they are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated or implicitly indicating the precedence of the technical features indicated.
In the description of the present invention, unless otherwise explicitly limited, terms such as arrangement, installation, connection and the like should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above terms in the present invention in combination with the specific contents of the technical solutions.
Fig. 1 is a flowchart of an artifact removing method according to an embodiment of the present invention. The artifact removing method as shown in fig. 1 at least comprises the following steps: s110: acquiring at least two magnetic resonance signals from different receivers; s120: acquiring weight information of the electromagnetic interference signals according to the at least two magnetic resonance signals; s130: optimizing the magnetic resonance signal according to the weight information of the electromagnetic interference signal and a deep learning algorithm to obtain a magnetic resonance imaging signal; s140 generates magnetic resonance imaging from the magnetic resonance imaging signals.
According to an embodiment of the first aspect of the present invention, there is provided an artifact removing method applied to a magnetic resonance imaging apparatus having at least two receivers, including: acquiring at least two magnetic resonance signals from different receivers, the magnetic resonance signals including a magnetic resonance imaging signal and an electromagnetic interference signal; acquiring weight information of the electromagnetic interference signals according to the at least two magnetic resonance signals; optimizing the magnetic resonance signal according to the weight information of the electromagnetic interference signal and a deep learning algorithm to obtain a magnetic resonance imaging signal; magnetic resonance imaging is generated from the magnetic resonance imaging signals.
In some embodiments, obtaining weight information of the electromagnetic interference signal from the at least two magnetic resonance signals comprises: a blind source separation algorithm is applied to separate the magnetic resonance imaging signal and the electromagnetic interference signal.
In some embodiments, the sources of electromagnetic interference signals EMI include from inside and outside. Sources of electromagnetic interference signals EMI may include power lines or electrical gradient, among others. The electromagnetic interference signal EMI is independent of the encoding gradients of the magnetic resonance imaging MRI. The electromagnetic interference signal EMI may also be related to the sequences and parameters used to collect the data. I.e. the parameters involved in the electromagnetic interference signal EMI, may comprise the main magnetic field or bandwidth and may be sent altered after a long scan has been performed.
Fig. 2 is a flowchart of an artifact removing method according to another embodiment of the present invention. The artifact removing method as shown in fig. 2 at least comprises the following steps: s210: acquiring magnetic resonance signals from at least two receivers; s220: generating a mixing matrix from the magnetic resonance signals; s230: the mixing matrix is resolved to acquire magnetic resonance imaging signals and electromagnetic interference signals.
In some embodiments, the magnetic resonance MR signals may be spatially encoded by magnetic gradient coils, i.e. the received magnetic resonance MR signals are modulated by coil sensitivity maps (coil sensitivity maps).
In some embodiments, the imaging of magnetic resonance imaging MRI is affected by the position, orientation and physical characteristics of the radio frequency coil. Electromagnetic interference signals EMI from external sources, i.e. far field sources, may be provided with differently directed coils and multiple receivers in the vicinity of the subject.
In some embodiments, multiple array coils may be used to detect the near field source spatial distribution. The human body is an electrically conductive structure and the body part outside the magnet will be coupled with the environment. Therefore, there is a need to detect electromagnetic interference signal EMI distributions using multiple coils at different positions and orientations, which can further improve the magnetic resonance MR scan results.
In some embodiments, three directions, e.g., X, Y, Z directions, are selected to position the plurality of array coils. Under the ultra-low magnetic field of 0.001-0.1T, the wavelength of the magnetic resonance MR signal is long enough, and the wavelength of the electromagnetic interference signal EMI which can interact with the magnetic resonance imaging MRI acquisition is also long. Thus, the number of sensors required to characterize the electromagnetic interference signal EMI may be small. While in medium and high field MR signals are shorter in wavelength, more array coils need to be used. At high fields, the noise comes primarily from the thermal noise of the sample. In low or ultra-low fields, the noise mainly includes electrical noise generated by the coil and the receiving chain, such as radio frequency pre-amplification, analog-to-digital conversion, demodulation, and the like.
In some embodiments, the artifact removal method can achieve artifact removal for ultra-low magnetic fields, i.e., 0.001T to 0.1T, low fields, i.e., 0.1T to 0.5T, mid-fields, i.e., 0.5T to 3T, and high fields, 3T and beyond magnetic resonance imaging MRI systems by applying radio frequency electronics with multiple receivers.
In some embodiments, the blind source separation algorithm comprises: acquiring magnetic resonance signals from at least two receivers; generating a mixing matrix from the magnetic resonance signals; the mixing matrix is resolved to acquire magnetic resonance imaging signals and electromagnetic interference signals.
In some embodiments, the electromagnetic interference signal EMI signal and the magnetic resonance MR signal are separated by using Blind Signal Separation (BSS). The magnetic resonance MR images are modulated by Coil Sensitivity Maps (CSM) and then fourier transformed into k-space. Modulation of the coil sensitivity map in image space is equivalent to convolution in k-space.
In some embodiments, the electromagnetic interference signal EMI signal is assumed to be statistically invariant over the entire k-space. The magnetic resonance MR signals and the electromagnetic interference signals EMI detected by the receiver can be linearly modeled as a Blind Source Separation (BSS) problem, which in turn separates the magnetic resonance MR signals and the electromagnetic interference signals EMI signals.
In some embodiments, the number of measurements should be greater than the number of sources when performing blind source separation. It can be derived that: where M is N measurements, s is M sources, and W is an N x M mixing matrix that linearly combines the sources.
In some embodiments, when N > M, the mixing matrix W and M sources s may be calculated from the measured values M by Independent Component Analysis (ICA). Where s contains at most M-1 sources associated with the magnetic resonance MR signals and 1 source associated with the electromagnetic interference signals EMI.
In some embodiments, the mixing matrix may be predicted by independent component analysis. And decomposing the measured mixed signal into sub-components which can be added to obtain a corresponding unmixing matrix and a corresponding mixing matrix.
In some embodiments, blind source separation may be performed by a fixed-point algorithm, for example, by a fixed-point algorithm based on projection pursuit (projection pursuit), an information maximization algorithm (infomax) based on entropy maximization (entropy maximization), or a fast independent component analysis (fast ICA), among other algorithms.
In some embodiments, the mixing matrix may also be predicted by Principal Component Analysis (PCA). The principal component analysis may be performed by eigenvalue decomposition of a data covariance matrix or singular value decomposition of a data matrix.
In some embodiments, the k-space of the magnetic resonance MR data acquired by the plurality of receiver elements may also be reconstructed as a matrix or tensor and decomposed using singular value decomposition.
In some embodiments, electromagnetic interference EMI and noise analysis may also be performed using the Welch method. Due to the different power spectral distribution characteristics of the MR signal and the EMI signal. Unparameterized classical spectral estimation was performed by the Welch method. The analysis is completed by dividing the signal into partially overlapping data segments, windowing the data segments separately to compute a periodogram, and finally averaging.
In some embodiments, electromagnetic interference EMI and noise analysis may also be performed using an autoregressive moving average (ARMA) model. The problems of poor variance performance and low resolution of classical spectrum estimation can be improved by applying the autoregressive moving average model. Assuming that a source signal s is output by a linear time-invariant system H (z) with a zero point and a pole point through an input sequence, parameters of H (z) are estimated by s or an autocorrelation function r thereof, and then a power spectrum is estimated by H (z), a modified Yule-Walker method can be adopted. After the power spectral density of each source is derived, it is classified, and if the power spectral density of the source is higher than a preset threshold within a preset frequency band, the source is identified as electromagnetic interference (EMI). The magnetic resonance MR signals can be identified by peak detection, i.e. it is checked whether the peak corresponding to the center of k-space is above a threshold.
In some embodiments, the source may be identified in image space. EMI typically appears as a band of bright noise in image space that overlaps with a background or object area.
In some embodiments, the coil sensitivity information may be integrated into an independent component analysis algorithm.
In some embodiments, the deep learning algorithm includes applying a neural network for deep learning; the architecture of the neural network comprises: one or more of a convolutional neural network model, a generative confrontation network model, a recurrent neural network model, a long-short term memory network model, an auto-encoder network model, a deep belief network model, a deep residual network model, a gate cycle unit network model, or an echo state network model.
In some embodiments, a convolutional neural network is applied as an example. The input is complex image data, for example, 256 × 256 × 2 × 2 complex image data, the length of the second to last dimension is 2 to represent 2 receivers, the length of the last dimension is 2 to represent the real part channel and the imaginary part channel, respectively, the output of the convolutional neural network model is 256 × 256 × 2 × 2 residual image data, corresponding to the real part and the imaginary part of the residual image data of 2 receivers. The convolutional neural network model comprises five convolutional layers and activation functions, specifically comprises a 9 × 9 convolutional layer, a ReLu activation layer, a 7 × 7 convolutional layer and a ReLu activation layer; the number of output channels of each convolution layer is 128, 64, 32 and 4, respectively. When training the model, an ADAM or SGD optimizer may be used to minimize a loss function that selects the Mean Squared Error (MSE).
In some embodiments, the output of the neural network may be a clean magnetic resonance MR signal, which in turn is connected to another network responsible for reconstructing the magnetic resonance MR image, i.e. inputs the magnetic resonance MR signal and outputs a magnetic resonance MR map.
In some embodiments, a neural network may be designed that does not perform the image reconstruction process, i.e., the reconstructed image is directly output by a single neural network. In addition, EMI cancellation processes may occur in a hybrid space between raw data k-space and image space, e.g., kx-y space, x-kySpace, kx-y-z space and kx-ky-z, etc.
In some embodiments, the output of the neural network may be an electromagnetic interference signal EMI and a clean magnetic resonance MR signal may be obtained by subtracting the electromagnetic interference signal EMI output from the contaminated magnetic resonance MR signal. The training mode of the neural network can acquire training data as fast as possible.
In some embodiments, the training may be performed by acquiring clean magnetic resonance MR signals by a magnetic resonance system with a good shielding setup.
In some embodiments, the electromagnetic interference signal EMI may be obtained by simulation. The electromagnetic interference signal EMI may be generated by randomizing the phase, center frequency and bandwidth of the electromagnetic interference signal EMI. Noise such as additive white gaussian can also be added to simulate real conditions.
In some embodiments, the electromagnetic interference signal EMI may be acquired by turning off the NMR excitation radio frequency pulse or removing the magnetic resonance MR signal source from the radio frequency coil. The detected electromagnetic interference signal EMI may also be related to the sequence and parameters of the collected data. For different sequences and different parameter settings, matching parameters may be required to obtain the electromagnetic interference signal EMI.
In some embodiments, electromagnetic interference signal EMI training data may be acquired simultaneously during a patient MRI scan. The electromagnetic interference signal EMI may be acquired by each NMR excitation repetition period at a time when the magnetic resonance MR signal is expected to be zero.
In some embodiments, magnetic resonance MR data contaminated by electromagnetic interference signals EMI may be obtained by adding simulated or acquired electromagnetic interference signals EMI or noise signals to the clean magnetic resonance MR signals. Further machine learning can be performed.
In some embodiments, the position and orientation of the array coils may be optimized for targeting.
In some embodiments, pure magnetic resonance MR signals can be obtained by simulation. Magnetic resonance imaging MRI data can be generated by modulating a complex image with the coil sensitivities of a map of the array coils, and then converted to the frequency domain by fourier transformation.
In some embodiments, the artifact removal method further comprises: and optimizing the magnetic resonance imaging signals according to the space-time multi-dimensional coupling linear matrix model.
Fig. 3 is a flowchart of an artifact removing method according to another embodiment of the present invention. The artifact removing method as shown in fig. 3 at least comprises the following steps: s310: grouping receivers; s320: setting different groups of receivers to have different signal sensitivities; s330: acquiring a magnetic resonance signal; s340: acquiring parameter information of a magnetic resonance imaging signal and an electromagnetic interference signal according to the signal sensitivity of the magnetic resonance signal and the receiver;
s350: optimizing magnetic resonance imaging signals based on parameter information of the magnetic resonance imaging signals and electromagnetic interference signals
In some embodiments, the spatio-temporal multidimensional coupled linear matrix model comprises: grouping receivers; setting different groups of receivers to have different signal sensitivities; acquiring a magnetic resonance signal; acquiring parameter information of a magnetic resonance imaging signal and an electromagnetic interference signal according to the signal sensitivity of the magnetic resonance signal and the receiver; the magnetic resonance imaging signal is optimized based on the parameter information of the magnetic resonance imaging signal and the electromagnetic interference signal.
In some embodiments, training data containing no magnetic resonance MR signals, but only electromagnetic interference signals EMI signals, is first acquired. May be obtained from a particular subject, a group of subjects, phantoms, or without any subject or phantom, and may also be acquired at a time during which the magnetic resonance MR signal is expected to be zero for each NMR excitation repetition period during a particular patient magnetic resonance imaging MRI scan. Assuming a total of C receiver elements, wherein the 1 st to K-th receiver elements are adjusted to have a certain sensitivity to both the magnetic resonance MR and the electromagnetic interference signal EMI signals, the K +1 th to C-th receiver elements are adjusted to have a sensitivity to the magnetic resonance MR signal close to 0 and the sensitivity to the electromagnetic interference signal EMI is adjusted to be maximal, i.e. only the electromagnetic interference signal EMI is received. For each receiver element i, the training data collected, which only contains electromagnetic interference signals EMI, is ni, where i is 1,2,3, …, C. Using the data for the last C-K receiver elements to represent the first K receivers, the following equation can be established:
[nK+1,nK+2,…,nC][wK+1,wK+2,…,wC]T=[n1,n2,…,nK]
by simplifying the formula, we can obtain:
NEMIW=Nmix
wherein N isEMI、W、NmixThe matrix sizes are Rx (C-K), (C-K) xK, RxK, respectively, and R is the number of repeated measurements.
It can be derived that: w is NEMI+NmixWhere, + represents Mole-Pentos in the broadest sense.
Suppose the data collected by the front K receivers and the rear C-K receivers during actual scanning is XmixAnd XEMThe data YMR after the first K receiver elements eliminate the EMI signal can be expressed as:
YMR=Xmix-XEMIW
in some embodiments, YMRThe data Of the K receivers can be combined by using a statistical square error method (RSS). Alternatively, the merging is performed before image reconstruction by Singular Value Decomposition (SVD).
According to an embodiment of the second aspect of the present invention, there is provided a magnetic resonance imaging system, comprising: a signal acquisition module for acquiring at least two magnetic resonance signals from different receivers, the magnetic resonance signals including a magnetic resonance imaging signal and an electromagnetic interference signal; the weight calculation module is used for acquiring weight information of the electromagnetic interference signals according to the at least two magnetic resonance signals; the imaging optimization module is used for optimizing the magnetic resonance signals according to the weight information of the electromagnetic interference signals and a deep learning algorithm so as to acquire the magnetic resonance imaging signals; and the imaging module is used for generating magnetic resonance imaging according to the magnetic resonance imaging signal.
In some embodiments, the magnetic resonance imaging system further comprises: and the signal receiving module is used for receiving radio frequency signals with different polarization directions and sensitivities.
In some embodiments, the radio frequency coil may be designed or placed to have different purposes. Some coils may be designed to minimize their sensitivity to magnetic resonance MR signals or to detect only electromagnetic interference signals EMI. This can be achieved by placing the coil away from the imaging region or in an orientation with minimal sensitivity to the magnetic resonance MR signals. Some coils are designed to detect magnetic resonance MR signals with minimal EMI contamination of the electromagnetic interference signals. Such optimization may be used to maximize the efficiency of eliminating the electromagnetic interference signal EMI, while at the same time, it may minimize thermal noise amplification during the electromagnetic interference signal EMI elimination process.
Fig. 4 is an image space image. Fig. 5 is a corresponding k-space image according to fig. 4. In fig. 4, the spatial images of the pure magnetic resonance MR signal, the electromagnetic interference signal EMI, the noise signal, and the magnetic resonance MR signal under measurement which is contaminated by the electromagnetic interference and the noise are sequentially shown from left to right. In fig. 5, a pure magnetic resonance MR signal, an electromagnetic interference signal EMI, a noise signal, and a magnetic resonance MR signal k-space image contaminated by electromagnetic interference and noise are sequentially shown from left to right. As shown in fig. 4 and 5, when interfered by electromagnetic interference signals EMI and/or noise signals, both the image of the image space and the image of the k-space are affected.
Fig. 6 is an image obtained by an artifact removing method according to another embodiment of the present invention. As can be seen from the image shown in fig. 6, the upper image is a magnetic resonance imaging MRI with an electromagnetic interference signal EMI, and the lower image is an image obtained after the electromagnetic interference signal EMI is removed by using an artifact removal method. By comparison, the image obtained after electromagnetic interference signal EMI elimination by using an artifact elimination method has higher signal-to-noise ratio and is clearer in imaging.
Fig. 7 is an image obtained by an artifact removing method according to another embodiment of the present invention. As can be seen from the image shown in fig. 7, the upper image is a human brain MRI with an electromagnetic interference signal EMI, and the lower image is a human brain image obtained by performing electromagnetic interference signal EMI elimination using an artifact elimination method. By comparison, the image obtained after the electromagnetic interference signal EMI is eliminated by using the artifact elimination method is clearer, and noise caused by the electromagnetic interference signal EMI can be eliminated.
According to an embodiment of the third aspect of the present invention, there is provided a magnetic resonance imaging terminal, characterized by comprising at least two receivers, a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing when executing the program: the artifact removal method of the first aspect.
According to an embodiment of a fourth aspect of the present invention, there is provided a computer-readable storage medium for a computer-readable storage, the storage medium storing one or more programs, the one or more programs being executable by one or more processors to implement the artifact removal method of the first aspect.
The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.
The above-described embodiments of the apparatus are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, i.e. may be located in one place, or may also be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.
One of ordinary skill in the art will appreciate that all or some of the steps, systems, and methods disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as is well known to those of ordinary skill in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, Digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can accessed by a computer. In addition, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media as known to those skilled in the art.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an illustrative embodiment," "an example," "a specific example," or "some examples" or the like mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the invention have been shown and described, it will be understood by those of ordinary skill in the art that: various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.