Medical image reconstruction method and device, medical imaging apparatus, and storage medium

Abstract

The present application relates to the technical field of image processing, and discloses a medical image reconstruction method and device, a medical imaging device and a storage medium. First, based on the data obtained by receiving coil scanning, real sensitivity spectrum data corresponding to the receiving coil is obtained. Then, the K-space data obtained by scanning is reconstructed by using a low-rank method to generate a phase to be corrected medical image and obtain the to-be-corrected phase information corresponding to the receiving coil. Finally, based on the real sensitivity spectrum data, the phase to be corrected medical image is phase corrected to obtain a target phase medical image with a smaller phase change than the phase to be corrected medical image. By phase correcting the phase to be corrected medical image based on the real sensitivity spectrum data, the problem of introducing artifacts due to inaccurate real sensitivity spectrum data can be avoided, and the quality of the phase to be corrected medical image generated due to the phase arbitrariness of the low-rank method is improved to obtain the target phase medical image.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a medical image reconstruction method, apparatus, medical imaging equipment, and storage medium. Background Technology

[0002] Compressed sensing is a sampling and reconstruction method for signals with sparse characteristics. It achieves high-precision signal reconstruction from the small amount of sampled data obtained by sampling the signal at a rate much lower than that specified by the Nyquist sampling theorem.

[0003] In related technologies, compressed sensing reconstruction methods based on low-rank characteristics are used to obtain phase images. However, the image quality of the phase images obtained by these technologies needs improvement. Therefore, a new medical image reconstruction method is required. Summary of the Invention

[0004] The embodiments described in this specification aim to at least partially solve one of the technical problems in the related art. To this end, the embodiments of this specification propose a medical image reconstruction method, apparatus, medical imaging device, and storage medium.

[0005] This specification provides a medical image reconstruction method, the method comprising:

[0006] Based on the data obtained from the scanning of the receiving coil, the true sensitivity spectrum data corresponding to the receiving coil is obtained;

[0007] The scanned K-space data is reconstructed using a low-rank method to generate a medical image of phase to be corrected and to obtain the phase information to be corrected corresponding to the receiving coil.

[0008] Based on the real sensitivity spectrum data, phase correction is performed on the medical image to be phase corrected to obtain a target phase medical image, wherein the phase change of the target phase medical image is smaller than the phase change of the medical image to be phase corrected.

[0009] In one embodiment, the receiving coil includes multiple receiving coil units, and the step of performing phase correction on the medical image to be corrected based on the true sensitivity spectrum data to obtain a target phase medical image includes:

[0010] The true sensitivity spectrum data corresponding to each receiving coil unit is extracted to obtain the true phase information corresponding to each receiving coil unit.

[0011] The phase correction of the medical image to be corrected is performed using the real phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit, to obtain the target phase medical image.

[0012] In one embodiment, the step of performing phase correction on the medical image to be corrected using the real phase information corresponding to each of the receiving coil units and the phase information to be corrected corresponding to each of the receiving coil units to obtain the target phase medical image includes:

[0013] The phase correction matrix is ​​calculated using the actual phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit to obtain the first correction data;

[0014] The first correction data is applied to the phase medical image to be corrected to obtain the target phase medical image.

[0015] In one implementation, the step of performing phase correction on the medical image to be corrected based on the true sensitivity spectrum data to obtain a target phase medical image includes:

[0016] The phase correction matrix is ​​calculated using the real sensitivity spectrum data and the reconstructed sensitivity spectrum data corresponding to the medical image to be phase corrected, to obtain the second correction data; wherein, the reconstructed sensitivity spectrum data is determined based on K-space data reconstructed using a low-rank method;

[0017] The second correction data is applied to the phase medical image to be corrected to obtain the target phase medical image.

[0018] In one embodiment, the receiving coil includes an array coil, the array coil including multiple receiving coil units, and obtaining the true sensitivity spectrum data corresponding to the receiving coil based on the data obtained by scanning the receiving coil includes:

[0019] The array coils are used to perform a low-resolution scan to obtain the received data corresponding to each receiving coil unit.

[0020] The true sensitivity spectrum data corresponding to each receiving coil unit is obtained by calculating the received data corresponding to each receiving coil unit and the overall received data. The overall received data is obtained by low-resolution scanning of the body coil.

[0021] In one embodiment, the receiving coil includes an array coil, the array coil including multiple receiving coil units, and obtaining the true sensitivity spectrum data corresponding to the receiving coil based on the data obtained by scanning the receiving coil includes:

[0022] The array coils are used to perform a low-resolution scan to obtain the received data corresponding to each receiving coil unit.

[0023] Channel merging processing is performed on the received data corresponding to each of the receiving coil units to determine the overall uniform data.

[0024] The true sensitivity spectrum data corresponding to each receiving coil unit is obtained by calculating based on the received data corresponding to each receiving coil unit and the overall homogenized data.

[0025] This specification provides a phase medical image reconstruction apparatus, the apparatus comprising:

[0026] The sensitivity spectrum determination module is used to obtain the true sensitivity spectrum data corresponding to the receiving coil based on the data obtained by scanning the receiving coil;

[0027] The medical image generation module is used to reconstruct images from scanned K-space data using a low-rank method, generate a medical image to be corrected for phase, and obtain the phase information to be corrected corresponding to the receiving coil.

[0028] A phase correction module is used to perform phase correction on the medical image to be corrected based on the real sensitivity spectrum data to obtain a target phase medical image, wherein the phase change of the target phase medical image is smaller than the phase change of the medical image to be corrected.

[0029] This specification provides a medical imaging device, which includes: a memory, and one or more processors communicatively connected to the memory; the memory stores instructions executable by the one or more processors, which, when executed by the one or more processors, cause the one or more processors to perform the steps of the method described in any of the above embodiments.

[0030] This specification provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any of the above embodiments.

[0031] This specification provides a computer program product that includes instructions that, when executed by a processor of a computer device, enable the computer device to perform the steps of the method described in any of the above embodiments.

[0032] In the above-described embodiment, firstly, based on the data obtained from scanning with the receiving coil, the true sensitivity spectrum data corresponding to the receiving coil is obtained. Then, a low-rank method is used to reconstruct the image from the scanned K-space data, generating a medical image to be phase-corrected and obtaining the phase information to be corrected corresponding to the receiving coil. Finally, based on the true sensitivity spectrum data, phase correction is performed on the medical image to be phase-corrected to obtain a target phase medical image whose phase change is less than that of the medical image to be phase-corrected. By performing phase correction on the medical image to be phase-corrected based on the true sensitivity spectrum data, the problem of artifacts introduced by inaccurate true sensitivity spectrum data can be avoided, and the quality of the medical image to be phase-corrected generated due to the phase arbitrariness of the low-rank method can be improved, thus obtaining the target phase medical image. Attached Figure Description

[0033] Figure 1a A schematic flowchart illustrating the medical image reconstruction method provided in the embodiments of this specification;

[0034] Figure 1b A schematic diagram of a phase image with phase discontinuities provided for embodiments of this specification;

[0035] Figure 1c A schematic diagram of the amplitude image provided for the embodiments of this specification;

[0036] Figure 1d A schematic diagram of a phase image with continuous phase provided for embodiments of this specification;

[0037] Figure 2 A schematic diagram illustrating the process of obtaining a target phase medical image as provided in the embodiments of this specification;

[0038] Figure 3 A schematic flowchart illustrating the process of obtaining a target phase medical image as provided in the embodiments of this specification;

[0039] Figure 4 A schematic diagram illustrating the process of obtaining a target phase medical image as provided in the embodiments of this specification;

[0040] Figure 5 A flowchart illustrating the process of determining reconstructed sensitivity spectrum data provided for the implementation of this specification;

[0041] Figure 6 A schematic diagram of the process for obtaining the true sensitivity spectrum data corresponding to the receiving coil, provided for the implementation of this specification;

[0042] Figure 7 A flowchart illustrating the process of obtaining real sensitivity spectrum data provided for the implementation of this specification;

[0043] Figure 8A schematic flowchart illustrating the medical image reconstruction method provided in the embodiments of this specification;

[0044] Figure 9 A schematic diagram of a phase medical image reconstruction apparatus provided for embodiments of this specification. Detailed Implementation

[0045] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0046] The Nyquist sampling theorem requires that the sampling frequency be at least twice the highest frequency in the analog signal to prevent aliasing. This requirement results in a massive amount of data being collected, increasing the burden on storage and processing. Furthermore, during signal compression, a large amount of redundant information is discarded, leading to resource waste.

[0047] Compressed sensing (CS) is a signal acquisition and processing technique based on applied mathematics that can improve the scanning speed and temporal and spatial resolution of magnetic resonance imaging (MRI).

[0048] Compressed sensing technology overcomes the limitations of the Nyquist sampling theorem. By utilizing the sparsity of signals and optimizing algorithms, it can sample at a rate lower than that required by the Nyquist theorem, achieving high-quality, high-precision signal reconstruction even with a significantly reduced sampling rate. This means that scanning speed can be greatly increased without sacrificing image resolution, thereby shortening scanning time.

[0049] Compressed sensing technology can significantly reduce scan time, typically by several to ten times, thereby increasing scan speed or improving image resolution. Compressed sensing technology has been applied in multiple clinical and research procedures throughout the body, demonstrating its feasibility and clinical reliability.

[0050] Compressed sensing technology, by combining parallel imaging technology, utilizes the sparsity, incoherence, and nonlinear reconstruction algorithms of signals, along with wavelet transform and other methods, to overcome the limitations of the traditional Nyquist sampling theorem. It can achieve accurate signal reconstruction at sampling rates below the Nyquist frequency, while retaining the advantages of parallel imaging technology, thereby further improving the reconstruction results of compressed sensing.

[0051] Compressed sensing technology is based on two key technologies: 1) parallel imaging technology. 2) the basic elements of compressed sensing, namely sparsity, incoherence, and nonlinear reconstruction.

[0052] In parallel imaging, image reconstruction methods can be broadly classified into three categories: image domain-based reconstruction methods and K-space-based reconstruction methods. Image domain-based reconstruction methods are represented by SENSE (Sensitivity Encoding), while K-space-based reconstruction methods are represented by GRAPPA (Generalized Autocalibrating Partially Parallel Acquisitions).

[0053] SENSE requires additional prior knowledge, namely the sensitivity spectrum. Among related techniques, it is possible to use low-rank theory to calculate the sensitivity spectrum through the central region of the image in the K-space region based on ESPIRiT (Extended SENSE with Iterative Reconstructions and k-space Time-domain constraints).

[0054] The sparsity of compressed sensing refers to the sparse nature of the signal to be recovered within a specific domain. In related techniques, compressed sensing primarily utilizes total variation or wavelet transform as the sparse domain. In these techniques, the Hankel matrix of the recovered signal in K-space has the smallest kernel norm, which directly indicates that the matrix is ​​low-rank. Therefore, due to the relatively concentrated singular values, it exhibits sparsity in most orthogonal transform domains.

[0055] The calculation of the low-rank property of a matrix can rely on Singular Value Decomposition (SVD). However, SVD has a problem: the calculated singular vectors are not unique. Different methods and platforms may result in singular vectors with the same singular value having inconsistent phases but consistent magnitudes. This is due to the very definition of SVD.

[0056] Specifically, the singular value decomposition results of the complex matrix A are shown below:

[0057] A=UΣV H

[0058] Where U and V are unitary matrices, and the column vectors of U and V are the left and right singular vectors of the complex matrix A, respectively. H Let I be the conjugate transpose of matrix V. It should be noted that since U and V are unitary matrices, then I = VV. H with I=U H I is the identity matrix.

[0059] Then by introducing a diagonal matrix Its diagonal element l i = iθ , satisfy l i *conj(l i ) = 1, thus ensuring I = L H Where θ can be any value, the above requirements can be satisfied. Construct a new left singular vector. and the new right singular vector At the same time, the singular value matrix Σ remains unchanged. Then the following conclusions hold:

[0060]

[0061] Simultaneously satisfy but and U and V are singular vectors with the same singular value Σ, differing only in their phase e. iθ The phase has changed. Therefore, it can be assumed that the phase becomes arbitrary after SVD decomposition. In low-rank methods, the arbitrariness of the phase leads to discontinuous phases in the results obtained separately.

[0062] Furthermore, the FFT encoding in K-space naturally leads to a low-rank structure of the Hankel matrix, resulting in an inherent implicit sparsity. This prior knowledge is widely used in compressed sensing reconstruction methods. Low-rank reconstruction methods have become a direction in compressed sensing algorithms, achieving high reconstruction quality and high downsampling rates. However, since their solution methods typically involve SVD decomposition, this results in diverse phases of the singular vectors in the reconstruction results. This variable phase property limits the application of low-rank reconstruction methods in algorithms that require phase information.

[0063] Based on this, the embodiments of this specification provide a medical image reconstruction method. First, based on the data obtained from scanning with a receiving coil, the true sensitivity spectrum data corresponding to the receiving coil is obtained. Then, a low-rank method is used to reconstruct the image from the scanned K-space data, generating a medical image to be phase-corrected and obtaining the phase information to be corrected corresponding to the receiving coil. Finally, based on the true sensitivity spectrum data, phase correction is performed on the medical image to be phase-corrected to obtain a target phase medical image with a phase change smaller than that of the medical image to be phase-corrected. By performing phase correction on the medical image to be phase-corrected based on the true sensitivity spectrum data, the problem of artifacts introduced by inaccurate true sensitivity spectrum data can be avoided, and the quality of the medical image to be phase-corrected generated due to the phase arbitrariness of the low-rank method can be improved, thus obtaining the target phase medical image.

[0064] This specification provides a medical image reconstruction method. Please refer to [link / reference]. Figure 1a The medical image reconstruction method may include the following steps:

[0065] S110. Based on the data obtained from the scanning of the receiving coil, obtain the true sensitivity spectrum data corresponding to the receiving coil.

[0066] Specifically, in medical image reconstruction, sensitivity spectrum data is crucial information, affecting the accuracy and quality of the reconstructed image. Before the formal imaging scan, one or more pre-scanning processes are performed to collect sensitivity information from the receiving coil. During the pre-scanning, the receiving coil acquires signals from the subject, obtaining pre-scan data. By analyzing and processing the pre-scan data acquired by the receiving coil, the true sensitivity spectrum data corresponding to the receiving coil is calculated.

[0067] It should be noted that during multiple scans, the movement of the subject can cause a mismatch between the sensitivity spectrum data and the actual image. For example, in abdominal imaging, the subject's respiratory movements cause deviations between the pre-acquired image and the actual image. This deviation leads to inconsistencies between the pre-calibrated true sensitivity spectrum data and the sensitivity spectrum data obtained during actual imaging. Due to this inconsistency, artifacts are introduced when using the true sensitivity spectrum data for image reconstruction, reducing the quality of the final image. However, while the true sensitivity spectrum data is not entirely accurate due to motion, the phase information it contains is still determined based on the actual coil reception data, thus possessing high accuracy. Therefore, although the true sensitivity spectrum data may not be suitable for direct image reconstruction, its accurate phase information can be used for subsequent phase correction to improve the quality of the phase image.

[0068] S120. The K-space data obtained by scanning is reconstructed using a low-rank method to generate a medical image of the phase to be corrected and to obtain the phase information to be corrected corresponding to the receiving coil.

[0069] In some cases, phase information has wide applications in medical imaging. For example, SWI (Susceptibility Weighted Imaging) utilizes the phase information generated in magnetic resonance images by magnetically susceptible contrast agents to enhance the contrast of blood vessels and brain tissue, thereby improving the detection capability of microvessels and iron-containing substances in blood. SWIM (Susceptibility Weighted Imaging and Mapping) further combines magnetic susceptibility imaging and quantitative phase analysis, which can be used to quantitatively assess local magnetic susceptibility and local field inhomogeneities, and can be used for disease diagnosis. Dixon imaging technology, through the processing of phase information, achieves the separation imaging of fat and water, and can be used for the diagnosis of diseases such as fat deposition and body fluid distribution. Therefore, generating accurate phase images is essential.

[0070] Specifically, undersampling is used to undersample the K-space to obtain partial K-space data. Next, a low-rank method is used to reconstruct the image from the scanned K-space data, recovering lost information to generate a phase-corrected medical image. Furthermore, during the image reconstruction process using the low-rank method, the phase information to be corrected corresponding to the receiving coil is obtained. The phase-corrected medical image is represented as a complex matrix, where each element contains both amplitude and phase components. It should be noted that the low-rank method can be a low-rank compressed sensing method, which utilizes the characteristic that a signal is low-rank in a certain transform domain to reconstruct the undersampled data using compressed sensing theory.

[0071] For example, when reconstructing images in both the Phase Encoding (PE) and Slice-Phase Encoding (SPE) directions, each image layer is reconstructed independently, and then the reconstruction results of each layer are combined along the readout dimension (RO direction). This reconstruction process involves Singular Value Decomposition (SVD). However, because the phase obtained by SVD decomposition is arbitrary and each image layer is reconstructed independently, phase consistency is not maintained, leading to phase discontinuities in the readout dimension (RO direction) and the PE direction. See also... Figure 1b When viewing the phase image separated from the reconstructed phase-corrected medical image in the readout dimension (RO direction) and PE direction, Figure 1b Phase discontinuities can be observed in the displayed phase images. Furthermore, since the acquired K-space data is in complex form, the data at each spatial location contains both amplitude and phase information. Therefore, the reconstructed medical image for phase correction can not only extract phase information to determine a phase image, but also extract amplitude information to determine an amplitude image. Please refer to [link to relevant documentation]. Figure 1c , Figure 1c To and Figure 1b The phase image corresponds to the amplitude image.

[0072] In some implementations, the basic formula for image reconstruction is as follows:

[0073]

[0074] Among them, F u F represents undersampled Fast Fourier Transform. u =M·FFT, where M is the acquisition mask (acquired points are 1, unacquired points are 0), · represents the Hadamard product (point-to-point product), and Ψ TFor wavelet transform, λ1 is the regularization coefficient of the wavelet transform, y is the acquired K-space data, and D... i This represents calculating the gradient at the i-th point. For u j The total variation is calculated, where λ² is the regularization coefficient of the total variation, S is the prior information of the sensitivity spectrum data used for the data fidelity term, and x is the compressed sensing result. It should be noted that the compressed sensing result is expressed as a complex matrix.

[0075] Furthermore, the above formula is a basic formula and needs to be adaptively adjusted and modified according to specific application scenarios and data characteristics to adapt to different application scenarios. For example, SPIRiT implicitly utilizes the channel relationship of the receiving coil to reconstruct the sensitivity spectrum data, and the formula can be modified as follows:

[0076]

[0077] Where D is the linear operator, y is the acquired K-space data, x is the compressed sensing result, G is the SPIRiT operator obtained from calibration, I is the unit operator, λ1 is the regularization coefficient, λ2 is the regularization coefficient, and R(x) is the penalty function containing prior knowledge.

[0078] It should be noted that r(x) takes different forms under different conditions. For example, when the K-space is sparse, R(x) = ||x||²; when the weighted K-space is sparse, R(x) = ||Wx||², where W is the weight; when the image domain is sparse, ... Among them, F H This represents the inverse Fourier transform. Represents the gradient operator; when the wavelet transform is sparse, R(x) = ||Ψ T F H x‖1, where Ψ T For wavelet transform; in low-rank sparse cases, Where H is the transformation matrix, rank represents the rank of the matrix, and k is a preset parameter. In some implementations, the above conditions can be combined to determine R(x).

[0079] S130. Based on real sensitivity spectrum data, perform phase correction on the medical image to be corrected to obtain the target phase medical image.

[0080] Among them, the phase change of the target phase medical image is smaller than the phase change of the phase-corrected medical image.

[0081] Specifically, low-rank methods lead to phase variability, affecting the quality of the medical image to be phase-corrected. True sensitivity spectrum data is determined based on actually acquired coil reception data; therefore, the phase information included in true sensitivity spectrum data is reliable and highly accurate, and can be considered a standard. Thus, using true sensitivity spectrum data, phase correction is performed on the medical image to be phase-corrected, making the image phase closer to the phase of the true coil. This results in a target phase medical image with a phase change smaller than that of the image to be corrected. The target phase medical image is represented as a complex matrix, where each element contains both amplitude and phase components.

[0082] In some implementations, the phase jumps between the medical images to be corrected are large, resulting in discontinuities that affect the phase information of the image and thus the accuracy of clinical diagnosis. Therefore, it is necessary to perform phase correction on the medical images to be corrected based on real sensitivity spectrum data to make the phase of the medical images to be corrected continuous and closer to the phase of the real coil, thereby obtaining the target phase medical image for practical application.

[0083] In other embodiments, to further improve the quality of the medical image to be phase-corrected, phase correction can be performed on the medical image to be phase-corrected based on real sensitivity spectrum data, so that the phase of the medical image to be phase-corrected is closer to the phase of the real coil, thereby obtaining a higher quality medical image, namely the target phase medical image.

[0084] For example, please refer to Figure 1d When viewing the phase images separated from the target phase medical image in the readout dimensions (RO direction) and PE direction, Figure 1d Phase continuity can be observed in the displayed phase image. In the above embodiment, firstly, based on the data obtained from scanning the receiving coil, the true sensitivity spectrum data corresponding to the receiving coil is obtained. Then, a low-rank method is used to reconstruct the image from the scanned K-space data, generating a medical image to be phase-corrected and obtaining the phase information to be corrected corresponding to the receiving coil. Finally, based on the true sensitivity spectrum data, phase correction is performed on the medical image to be phase-corrected to obtain a target phase medical image with a phase change smaller than that of the medical image to be phase-corrected. By performing phase correction on the medical image to be phase-corrected based on the true sensitivity spectrum data, the problem of artifacts introduced by inaccurate true sensitivity spectrum data can be avoided, and the quality of the medical image to be phase-corrected generated due to the phase arbitrariness of the low-rank reconstruction method can be improved, thus obtaining the target phase medical image.

[0085] In some implementations, please refer to Figure 2The receiving coil includes multiple receiving coil units. Based on real sensitivity spectrum data, it performs phase correction on the medical image to be corrected to obtain the target phase medical image. This may include the following steps:

[0086] S210. Extract the true sensitivity spectrum data corresponding to each receiving coil unit to obtain the true phase information corresponding to each receiving coil unit.

[0087] S220. Using the real phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit, the phase of the medical image to be corrected is phase corrected to obtain the target phase medical image.

[0088] Specifically, certain imaging techniques (such as STDLR and SHLR) typically require implicit use of the channel relationships of the receiving coils, i.e., reconstructing the sensitivity spectrum data, to obtain accurate image reconstruction results. Therefore, to achieve phase correction, phase information needs to be extracted from the known data for subsequent phase correction. For the true sensitivity spectrum data corresponding to each receiving coil unit, the true phase information corresponding to each receiving coil unit is extracted from the true sensitivity spectrum data corresponding to each receiving coil unit. The true phase information corresponding to each receiving coil unit is determined based on the received data actually acquired by the receiving coil, so the true phase information can be regarded as a standard.

[0089] In the process of reconstructing the K-space data obtained from scanning using a low-rank method to generate a medical image for phase correction, the phase information to be corrected corresponding to each receiving coil unit is determined. Since the phase information to be corrected is calculated, the phase difference between each channel is fixed after one SVD calculation. However, phase arbitrariness can occur in the results of different SVD calculations. To reduce or even eliminate this arbitrariness, it is necessary to correct the phase information to be corrected using standard real phase information. Therefore, by combining the real phase information and the phase information to be corrected of each receiving coil unit, phase correction is performed on the medical image for phase correction, thereby obtaining the final target phase medical image.

[0090] In the above embodiments, the true sensitivity spectrum data corresponding to each receiving coil unit is extracted to obtain the true phase information corresponding to each receiving coil unit. The true phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit are used to perform phase correction on the medical image to be corrected, thereby obtaining a high-quality target phase medical image.

[0091] In some implementations, please refer to Figure 3The phase correction of the medical image to be corrected is performed using the real phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit to obtain the target phase medical image. This can include the following steps:

[0092] S310. Calculate the phase correction matrix using the real phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit to obtain the first correction data.

[0093] S320. Apply the first correction data to the phase of the medical image to be corrected to obtain the target phase medical image.

[0094] Specifically, the true phase information corresponding to each receiving coil unit is obtained based on the actual acquired receiving coil, so the true phase information can be used as a standard. The phase information to be corrected is affected by the randomness during the SVD calculation process, which manifests as differences between the results of different SVD calculations. In one SVD calculation, all receiving coil units are filled at once, so phase inconsistencies occur when there are different index values ​​in the RO (readout) direction. The differences between receiving coil units are fixed. To eliminate the influence of phase randomness during the SVD calculation process, a phase correction matrix is ​​calculated using the true phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit, resulting in the first correction data. The first correction data is applied to the medical image to be phase corrected to eliminate phase differences, resulting in a target phase medical image with a phase change smaller than that of the medical image to be phase corrected.

[0095] In some implementations, the phase of the medical image to be corrected, i.e., the compressed sensing result, is corrected using the formula shown below to obtain the corrected compressed sensing result:

[0096] P = T·∑ γ E γ ·conj(F γ ) / ∑ γ (F γ ·onj(F γ )+R)

[0097] Where T represents the phase-corrected medical image, i.e., the compressed sensing result, R is the regularization matrix, and F... γ E represents the true phase information corresponding to the γth receiving coil unit. γ This represents the phase information to be corrected corresponding to the γ-th receiving coil unit, where γ represents the number of channels (i.e., the number of receiving coil units), and P represents the corrected compressed sensing result. ∑ γ E γ ·conj(F γ ) / ∑ γ (Fγ ·conj(F γ )+R) is the first correction data.

[0098] In the above embodiments, a phase correction matrix is ​​calculated using the real phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit to obtain first correction data. The first correction data is then applied to the phase to be corrected medical image to obtain a high-quality target phase medical image.

[0099] In some implementations, please refer to Figure 4 Phase correction is performed on the medical image to be phase corrected based on real sensitivity spectrum data to obtain the target phase medical image. This may include the following steps:

[0100] S410. Calculate the phase correction matrix using the real sensitivity spectrum data and the reconstructed sensitivity spectrum data corresponding to the medical image to be phase corrected, and obtain the second correction data.

[0101] S420. Apply the second correction data to the phase of the medical image to be corrected to obtain the target phase medical image.

[0102] The reconstructed sensitivity spectrum data is determined based on the K-space data reconstructed using the low-rank method.

[0103] Specifically, sensitivity spectral data varies because different receiving coils have different sensing capabilities for the same signal. This difference manifests as variations in amplitude and phase within the sensitivity spectral data. Certain imaging techniques (such as ESPIRIT) do not use additional sensitivity spectra. Instead, based on the low-rank property of the Hankel arrangement in K-space, they typically require explicit generation of the channel relationships of the receiving coils—that is, reconstructing the sensitivity spectral data—to obtain accurate image reconstruction results. Therefore, the reconstructed sensitivity spectral data can be directly used for phase correction. The phase correction matrix is ​​calculated using the real sensitivity spectral data and the reconstructed sensitivity spectral data corresponding to the medical image to be phase corrected, yielding the second correction data. This second correction data is then applied to the medical image to be phase corrected to eliminate phase differences, resulting in a target phase medical image with a phase change smaller than that of the medical image to be phase corrected.

[0104] In some implementations, the compressed sensing results of the tissue corresponding to the reconstructed sensitivity spectral data of the j-th set are phase-corrected using the formula shown below to obtain the corrected compressed sensing results:

[0105]

[0106] Among them, u j Let S represent the compressed sensing result of the tissue corresponding to the j-th set of reconstructed sensitivity spectrum data, R be the regularization matrix, and S be the reconstructed sensitivity spectrum data. γThis is the actual sensitivity spectrum data corresponding to the γth receiving coil unit. This refers to the reconstructed sensitivity spectrum data corresponding to the γ-th receiving coil unit of the j-th set, where γ represents the number of channels, i.e., the number of receiving coil units, and P... j This represents the corrected compressed sensing result. This is the second corrective data.

[0107] Through phase correction, P values ​​are obtained that are all corrected back to their true phase. j Using the SOS method, each P j The amplitudes are obtained by combining them. Each P is obtained by directly adding them together. j The phase information extracted from the obtained image is used to obtain the phase. Then, the amplitude obtained by the SOS method and the phase obtained by direct addition are combined to obtain the final reconstructed complex matrix.

[0108] In the above embodiments, a phase correction matrix is ​​calculated using real sensitivity spectrum data and reconstructed sensitivity spectrum data corresponding to the medical image to be phase corrected, to obtain second correction data. The second correction data is then applied to the medical image to be phase corrected to obtain a high-quality target phase medical image.

[0109] In some implementations, please refer to Figure 5 The reconstructed sensitivity spectrum data were determined in the following ways:

[0110] S510. Based on the K-space data, determine the K-space Hankel matrix.

[0111] Specifically, the K-space data is automatically calibrated using a local sampling mode. Within the automatic calibration region, the K-space data is locally sampled using a sliding window approach. Each time the sliding window covers a new local region, the actually collected sample points are extracted from that region and rearranged into rows of the Hankel matrix to obtain the K-space Hankel matrix.

[0112] In some implementations, the K-space data is organized into a multi-channel grid, typically with uncollected data padded with zeros. An operator Rr is defined for selecting K-space blocks, and an operator Pr is defined for selecting local sampling modes. Within the Automatic Calibration (ACS) region, K-space data is locally sampled using a sliding window. The window slides across the K-space data, covering a new local region each time it moves. For each local region, operator Rr is applied to extract data blocks from that region, and then operator Pr is applied to select the actual sample points from these data blocks. These local sample points are reordered into a vector, which is then used as a row in the Hankel matrix. This process is repeated throughout the entire ACS region until all local samples are included in the Hankel matrix.

[0113] S520. The Hankel matrix in the K-space is calculated and processed using the low-rank method to obtain the reconstructed sensitivity spectrum data.

[0114] Specifically, K-space data blocks are selected using a sliding window, and these blocks overlap in the row direction. Each data block captures information from multiple receiving coil units, so that each row of the K-space Hankel matrix contains mixed information from multiple receiving coils. Low-rank methods are used to approximate the K-space Hankel matrix and process the data to obtain the reconstructed sensitivity spectrum data corresponding to each receiving coil unit.

[0115] In some implementations, SVD decomposition is performed using the Hankel matrix of the ACS region in K-space, and the singular value decomposition results are thresholded to retain the main information in the data. The thresholded singular vectors are then zero-padded and arranged before a Fast Fourier Transform (FFT) is performed. The FFT-transformed data is then subjected to another SVD decomposition, the resulting singular vectors are arranged, and the resulting singular values ​​are thresholded and weighted. Finally, the singular vectors are multiplied by the corresponding processed singular values ​​to obtain the final reconstructed sensitivity spectrum data.

[0116] In some implementations, when the imaging field of view is smaller than the object being imaged, the sensitivity spectrum may not accurately represent the entire image area, especially at the object edges or outside the field of view. In such cases, multiple sensitivity spectra may be needed to represent different parts of the image separately. Different tissue types may have different magnetic susceptibility and chemical environments, resulting in different sensitivity responses to the coils. For example, signals from fat and water may produce different sensitivity responses, thus requiring two sets of reconstructed sensitivity spectrum data to represent the signals from fat and water respectively. This can be understood as multiple receiving coil units corresponding to one set of reconstructed sensitivity spectrum data for the fat signal. For water signals, the same set of reconstructed sensitivity spectrum data corresponds to multiple receiving coil units. in, This represents the reconstructed sensitivity spectrum data corresponding to the γth receiving coil unit of the first set. This represents the reconstructed sensitivity spectrum data corresponding to the γth receiving coil unit of the second set, where γ represents the number of channels, i.e., the number of receiving coil units.

[0117] It should be noted that when multiple sets of reconstructed sensitivity spectrum data exist, the compressed sensing result of the tissue corresponding to each set of reconstructed sensitivity spectrum data is determined by the following formula. The compressed sensing result can extract amplitude and phase information, which are then used to determine the amplitude image and phase image, respectively:

[0118]

[0119] Among them, F u F represents undersampled Fast Fourier Transform. i =M·FFT, where M is the acquisition mask (acquired points are 1, unacquired points are 0), · represents the Hadamard product (point-to-point product), and Ψ T For wavelet transform, λ1 is the regularization coefficient of the wavelet transform, y is the acquired K-space data, and D... i This represents calculating the gradient at the i-th point. For u j Find the total variation, where λ² is the regularization coefficient of the total variation. U represents the reconstructed sensitivity spectrum data of the j-th set. j This represents the compressed sensing result of the tissue corresponding to the reconstructed sensitivity spectrum data of the j-th set.

[0120] In the above implementation, the K-space Hankel matrix is ​​determined based on K-space data, and the K-space Hankel matrix is ​​calculated and processed using the low-rank method to obtain reconstructed sensitivity spectrum data, providing data support for subsequent phase correction of medical images to be corrected.

[0121] In some implementations, please refer to Figure 6 The receiving coil includes an array coil, which in turn includes multiple receiving coil units. Based on the data obtained from scanning the receiving coil, the true sensitivity spectrum data corresponding to the receiving coil is obtained, which may include the following steps:

[0122] S610: Perform low-resolution scanning using the array coils to obtain the received data corresponding to each receiving coil unit.

[0123] S620. Calculate the true sensitivity spectrum data corresponding to each receiving coil unit based on the received data corresponding to each receiving coil unit and the overall received data.

[0124] The overall received data is obtained through low-resolution scanning using a body coil.

[0125] In some cases, since acquiring the true sensitivity spectrum data of each receiving coil unit requires additional pre-scanning processing, a low-resolution scan and a large field of view (FOV) are used to save time and improve efficiency.

[0126] Specifically, a low-resolution pre-scan of the subject is performed using an array coil, which consists of multiple receiving coil units, each corresponding to one channel. During the pre-scan, each receiving coil unit independently acquires the magnetic resonance signal from the subject, obtaining its own received data. A second low-resolution pre-scan is then performed using a volume coil to obtain the overall received data. The volume coil has a large coverage area and uniform signal, providing a good baseline for calculating the sensitivity spectrum data. Different receiving coil units have varying sensitivities to the received signal due to their different spatial distributions, resulting in different sensitivity spectrum data for the magnetic resonance signal. Therefore, calculations need to be performed for each receiving coil unit, processing its acquired received data with the overall received data to obtain the true sensitivity spectrum data corresponding to that receiving coil unit. The volume coil can be an orthogonal body coil.

[0127] For example, the true sensitivity spectrum data of the receiving coil unit is determined by the following formula:

[0128]

[0129] Among them, S γ C represents the true sensitivity spectrum data corresponding to the γth receiving coil unit. γ This represents the received data corresponding to the γ-th receiving coil unit under low-resolution scanning, and Q represents the overall received data corresponding to the body coil under low-resolution scanning. Additionally, γ can also be understood as the number of channels.

[0130] It should be noted that because two pre-scans are required, the subject (patient) may exhibit involuntary movements during the scanning process, which could affect image quality. Secondly, in pursuit of rapid scanning, the pre-scan sequence typically uses a fast gradient echo (GRE) sequence, while the actual compressed sensing acquisition sequence can be of many types, such as TSE, SE, etc. The data acquired between different codes are mismatched. Therefore, compressed sensing results obtained using the true sensitivity spectrum data based on the pre-scan will exhibit artifacts.

[0131] In the above embodiments, low-resolution scanning is performed using array coils to obtain the received data corresponding to each receiving coil unit. Based on the received data corresponding to each receiving coil unit and the overall received data, calculations are performed to obtain the true sensitivity spectrum data corresponding to each receiving coil unit, providing data support for subsequent phase correction of medical images to be corrected.

[0132] In some implementations, please refer to Figure 7The receiving coil includes an array coil, which in turn includes multiple receiving coil units. Based on the data obtained from scanning the receiving coil, the true sensitivity spectrum data corresponding to the receiving coil is obtained, which may include the following steps:

[0133] S710: Low-resolution scanning is performed using the array coils to obtain the received data corresponding to each receiving coil unit.

[0134] S720: Perform channel merging processing based on the received data corresponding to each receiving coil unit to determine the overall uniform data.

[0135] S730. Calculate the true sensitivity spectrum data corresponding to each receiving coil unit based on the received data and the overall homogenization data.

[0136] In some cases, because acquiring the true sensitivity spectrum data for each receiving coil unit requires additional pre-scanning processing, a low-resolution scan with a large field of view (FOV) is used to save time and improve efficiency. However, to maximize time savings, the true sensitivity spectrum data can be obtained based on a single pre-scan.

[0137] Specifically, a low-resolution pre-scan of the subject is performed using an array coil, which consists of multiple receiving coil units, each corresponding to a channel. During the pre-scan, each receiving coil unit independently acquires the magnetic resonance signal from the subject, obtaining its own received data. The received data corresponding to each receiving coil unit are then combined, performing channel merging processing to remove channel dimensions, resulting in an ideal, uniformly homogenized data set without coil modulation. Different receiving coil units exhibit varying sensitivities to the received signal due to their different spatial distributions, leading to different sensitivity spectra for the magnetic resonance signal. Therefore, calculations need to be performed on the received data of each receiving coil unit and the uniformly homogenized data to obtain the true sensitivity spectrum data corresponding to that receiving coil unit.

[0138] In some implementations, the SOS (Sum of Squares) method is used, which involves multiplying the data point-to-point with its conjugate points, summing the results along the channel dimension, and then taking the square root to determine the overall homogenized data. For example, the SOS method is used to obtain an overall homogenized data without channel dimensions by taking the square root of the sum of the squares of the moduli of each point along the channel dimension of the multi-channel image. The true sensitivity spectrum data of the receiving coil unit is then determined using the following formula:

[0139]

[0140] Among them, S γC represents the true sensitivity spectrum data corresponding to the γth receiving coil unit. γ This represents the received data corresponding to the γth receiving coil unit under low-resolution scanning. This is for overall data homogenization.

[0141] In the above implementation, a low-resolution scan is performed using an array coil to obtain the received data corresponding to each receiving coil unit. Channel merging is then performed based on the received data corresponding to each receiving coil unit to determine the overall homogenized data. Calculations are then performed based on the received data corresponding to each receiving coil unit and the overall homogenized data to obtain the true sensitivity spectrum data corresponding to each receiving coil unit, providing data support for subsequent phase correction of medical images to be phase corrected.

[0142] This specification also provides a medical image reconstruction method, wherein the receiving coil includes an array coil, and the array coil includes multiple receiving coil units. For example, please refer to... Figure 8 The medical image reconstruction method may include the following steps:

[0143] S802. Low-resolution scanning is performed using the array coils to obtain the received data corresponding to each receiving coil unit.

[0144] S804. Calculate the true sensitivity spectrum data corresponding to each receiving coil unit based on the received data corresponding to each receiving coil unit and the overall received data.

[0145] The overall received data is obtained through low-resolution scanning using a body coil.

[0146] S806. Extract the true sensitivity spectrum data corresponding to each receiving coil unit to obtain the true phase information corresponding to each receiving coil unit.

[0147] S808. The low-rank method is used to reconstruct the K-space data obtained by scanning, generate a medical image of phase to be corrected, and obtain the phase information to be corrected corresponding to the receiving coil.

[0148] S810. Calculate the phase correction matrix using the real phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit to obtain the first correction data.

[0149] S812. Apply the first correction data to the phase medical image to be corrected to obtain the target phase medical image.

[0150] Among them, the phase change of the target phase medical image is smaller than the phase change of the phase-corrected medical image.

[0151] This specification provides a phase medical image reconstruction device 900. Please refer to [link / reference]. Figure 9 The phase medical image reconstruction device 900 includes: a sensitivity spectrum determination module 910, a medical image generation module 920, and a phase correction module 930.

[0152] The sensitivity spectrum determination module 910 is used to obtain the true sensitivity spectrum data corresponding to the receiving coil based on the data obtained by scanning the receiving coil;

[0153] The medical image generation module 920 is used to reconstruct the image from the scanned K-space data using a low-rank method, generate a phase-to-corrected medical image, and obtain the phase-to-corrected information corresponding to the receiving coil.

[0154] The phase correction module 930 is used to perform phase correction on the medical image to be corrected based on the real sensitivity spectrum data to obtain a target phase medical image, wherein the phase change of the target phase medical image is smaller than the phase change of the medical image to be corrected.

[0155] For a detailed description of the phase medical image reconstruction device, please refer to the description of the medical image reconstruction method above, which will not be repeated here.

[0156] In some embodiments, a medical imaging device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method steps described above.

[0157] This specification provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method in any of the above embodiments.

[0158] One embodiment of this specification provides a computer program product including instructions that, when executed by a processor of a computer device, enable the computer device to perform the steps of the method described in any of the above embodiments.

[0159] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

Claims

1. A method for medical image reconstruction, characterized in that, The method includes: Based on the data obtained from the scanning of the receiving coil, the true sensitivity spectrum data corresponding to the receiving coil is obtained, wherein the receiving coil includes multiple receiving coil units; The scanned K-space data is reconstructed using a low-rank method to generate a medical image of phase to be corrected and to obtain the phase information to be corrected corresponding to the receiving coil. The true sensitivity spectrum data corresponding to each receiving coil unit is extracted to obtain the true phase information corresponding to each receiving coil unit. Phase correction is performed on the medical image to be corrected using the true phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit to obtain a target phase medical image, wherein the phase change of the target phase medical image is less than the phase change of the medical image to be corrected; or The phase correction matrix is ​​calculated using the real sensitivity spectrum data and the reconstructed sensitivity spectrum data corresponding to the medical image to be phase corrected, to obtain the second correction data; wherein, the reconstructed sensitivity spectrum data is determined based on K-space data reconstructed using a low-rank method; The second correction data is applied to the phase medical image to be corrected to obtain the target phase medical image.

2. The method according to claim 1, characterized in that, The step of performing phase correction on the medical image to be corrected using the real phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit to obtain the target phase medical image includes: The phase correction matrix is ​​calculated using the actual phase information corresponding to each receiving coil unit and the phase information to be corrected corresponding to each receiving coil unit to obtain the first correction data; The first correction data is applied to the phase medical image to be corrected to obtain the target phase medical image.

3. The method according to claim 1, characterized in that, The receiving coil includes an array coil, which comprises multiple receiving coil units. Obtaining the true sensitivity spectrum data corresponding to the receiving coil based on the data obtained from the receiving coil scanning includes: The array coils are used to perform a low-resolution scan to obtain the received data corresponding to each receiving coil unit. The true sensitivity spectrum data corresponding to each receiving coil unit is obtained by calculating the received data corresponding to each receiving coil unit and the overall received data. The overall received data is obtained by low-resolution scanning of the body coil.

4. The method according to claim 1, characterized in that, The receiving coil includes an array coil, which comprises multiple receiving coil units. Obtaining the true sensitivity spectrum data corresponding to the receiving coil based on the data obtained from the receiving coil scanning includes: The array coils are used to perform a low-resolution scan to obtain the received data corresponding to each receiving coil unit. Channel merging processing is performed on the received data corresponding to each of the receiving coil units to determine the overall uniform data. The true sensitivity spectrum data corresponding to each receiving coil unit is obtained by calculating based on the received data corresponding to each receiving coil unit and the overall homogenized data.

5. A phase medical image reconstruction device, characterized in that, The device includes: The sensitivity spectrum determination module is used to obtain the true sensitivity spectrum data corresponding to the receiving coil based on the data obtained by scanning the receiving coil, wherein the receiving coil includes multiple receiving coil units; The medical image generation module is used to reconstruct the image from the scanned K-space data using a low-rank method, generate a medical image to be corrected for phase, and obtain the phase information to be corrected corresponding to the receiving coil. A phase correction module is used to extract the true sensitivity spectrum data corresponding to each receiving coil unit to obtain the true phase information corresponding to each receiving coil unit; to perform phase correction on the medical image to be corrected using the true phase information and the phase information to be corrected corresponding to each receiving coil unit to obtain a target phase medical image, wherein the phase change of the target phase medical image is less than the phase change of the medical image to be corrected; or to calculate a phase correction matrix using the true sensitivity spectrum data and the reconstructed sensitivity spectrum data corresponding to the medical image to be corrected to obtain second correction data; wherein the reconstructed sensitivity spectrum data is determined based on K-space data reconstructed using a low-rank method; and to apply the second correction data to the medical image to be corrected to obtain the target phase medical image.

6. A medical imaging device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images