Non-contrast agent enhanced magnetic resonance imaging and flow velocity quantification method and system

Abstract

The application provides a non-contrast agent enhanced magnetic resonance imaging and flow velocity quantitative measurement method and system, and belongs to the technical field of blood vessel imaging. The method improves the signal-to-noise ratio and image definition of blood vessel imaging through an optimized low-speed encoding acquisition scheme, combines an image reconstruction algorithm and a background phase correction technology, can clearly present the blood vessel network of hands and feet, and includes fine small blood vessel structures. Meanwhile, through a multi-stage arteriovenous separation technology, high-accuracy arteriovenous differentiation is realized, and a more reliable basis is provided for the visualization of blood vessel structures. The application adopts the above-mentioned non-contrast agent enhanced magnetic resonance imaging and flow velocity quantitative measurement method and system, avoids the risk that may be caused by a contrast agent, is suitable for a wider patient group, and improves the safety and universality of imaging.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of blood vessel imaging, in particular to a non-contrast agent enhanced magnetic resonance imaging and flow velocity quantitative measurement method and system. BACKGROUND

[0002] At present, the assessment of hand and foot vascular abnormalities mainly relies on imaging examination. Digital subtraction angiography (DSA) is still considered as the main means for detecting peripheral blood vessels of extremities, which has high spatial resolution, but it is an invasive examination, and there are risks of puncture injury, contrast agent allergy and radiation exposure, which is difficult to apply to routine screening and long-term follow-up.

[0003] Contrast-enhanced magnetic resonance angiography can realize examination without radiation, which obtains high-contrast images in the arterial phase by injecting gadolinium contrast agent. However, the peripheral blood vessels of extremities have small diameter and short arteriovenous transit time, and the imaging is easily affected by venous contamination, especially when the acquisition timing control is not proper or the contrast agent distribution is uneven. In addition, gadolinium contrast agent may cause nephrogenic systemic fibrosis in patients with abnormal renal function, which limits its application in certain population.

[0004] To avoid the risk of contrast agent, non-contrast agent enhanced magnetic resonance angiography techniques have been gradually applied in clinic, including two-dimensional time of flight (2D TOF) and quasi-static blood flow sensitive balance technology (QISS) based on blood flow inflow enhancement, and flow-related diffusion gradient technology (FBI), flow-sensitive dephasing technology (FSD) and enhanced accelerated arterial spin labeling (eAccASL) relying on blood flow hemodynamic characteristics. However, these techniques have obvious limitations in imaging of peripheral blood vessels of extremities (such as fingers and distal toes): due to small diameter, slow blood flow and weak pulsatility of peripheral blood vessels, it is difficult for existing methods to effectively suppress venous signals, resulting in low image contrast and poor arteriovenous differentiation, and the imaging quality is further deteriorated in the case of vascular stenosis or occlusion.

[0005] Three-dimensional phase-contrast magnetic resonance angiography (3D PC-MRA), as a typical non-contrast-enhanced method, generates vascular signals by creating phase differences in blood flow through flow-encoded gradients. However, traditional methods have inherent limitations in low-velocity blood flow regions: low blood signal intensity and a tendency for signal loss, especially in peripheral small blood vessels such as those in the fingers and toes. Furthermore, the method requires manually preset velocity encoding values ​​(VENC), but individual differences in blood flow velocity make parameter selection difficult, easily leading to weakened arterial signals or insufficient venous suppression. In addition, its velocity / acceleration-dependent arteriovenous separation mechanism is ineffective in suppressing venous contamination under conditions of low flow velocity and weak pulsation in peripheral arteries. Summary of the Invention

[0006] The purpose of this invention is to provide a non-contrast-enhanced magnetic resonance imaging method and system for quantitative blood flow measurement, which can achieve high-quality imaging in low-velocity, small-vessel regions, has the ability to separate arteries and veins, and can accurately quantify blood flow velocity to meet the clinical needs for peripheral vascular imaging.

[0007] To achieve the above objectives, the present invention provides a method for non-contrast-enhanced magnetic resonance imaging and quantitative flow rate measurement, comprising the following steps:

[0008] Step S1: Set up a three-dimensional phase contrast vascular imaging scanning sequence for the peripheral blood vessels of the limbs, and configure the readout direction to be consistent with the main course of the blood vessels;

[0009] Step S2: Perform dual flow rate encoding acquisition using two different low-speed encoding values;

[0010] Step S3: Apply low-rank constraints and variation regularization to the undersampled data for joint reconstruction;

[0011] Step S4: Perform background field correction and unwinding processing on the reconstructed phase data;

[0012] Step S5: Classify arteries and veins based on blood flow direction determination and vascular topology analysis.

[0013] Preferably, in step S1, the reading direction during hand scanning is from wrist to fingers; the reading direction during foot scanning is from ankle to toes; and the phase encoding direction is from right to left.

[0014] Preferably, in step S2, the two different low-speed coding values ​​are: an ultra-low speed coding value of ≤5cm / s and a reference speed coding value of ≥10cm / s.

[0015] Preferably, in step S3, low-rank constraints and variation regularization are applied to the undersampled data for joint reconstruction, as follows:

[0016] ;

[0017] wherein, represents an under-sampling operator, represents a Fourier transform, represents a coil sensitivity, vector represents actually acquired spatial data, represents an operator, extracts the b-th data block in all velocity encoded images and converts it into a Casorati matrix, and locally applies a low-rank constraint by penalizing the kernel norm of the Casorati matrix. represents a TV regularization constraint along the VENC direction, represents a weight of the local low-rank constraint, represents a weight of the variation regularization constraint, represents an image to be reconstructed spatially corresponding to the image, represents a reconstructed image.

[0018] Preferably, the step S4 comprises:

[0019] Step S41, applying Maxwell bias field correction to the reconstructed phase data;

[0020] Step S42, Gaussian filtering background phase estimation based on a blood vessel mask;

[0021] Step S43, frequency domain Laplace filtering unwrapping.

[0022] Preferably, the step S5 comprises:

[0023] Step S51, performing arteriovenous primary labeling based on the phase sign direction;

[0024] Step S52, extracting a blood vessel centerline by a three-dimensional refinement algorithm and establishing a topological connection relationship of the blood vessel branches;

[0025] Step S53, for each blood vessel branch, verifying the arteriovenous attribute based on its upstream and downstream connection relationship and spatial continuity, and when there is an attribute conflict at a branch node, correcting the attribute conflict by using a majority voting mechanism based on the length weighting of the blood vessel;

[0026] Step S54, for a phase inversion region caused by local magnetic field inhomogeneity, automatically correcting the phase inversion region in combination with the consistency of the blood flow direction of adjacent blood vessel segments;

[0027] Step S55, outputting three-dimensional models of an arterial tree and a venous tree with complete topological connection relationship.

[0028] Preferably, in step S53, a majority voting mechanism based on the length of the blood vessels is included, comprising:

[0029] Step S531, voxel-centerline association: according to the Euclidean distance between the voxel and the blood vessel skeleton and the connected domain relationship, the phase voxel in the velocity map is matched with the blood vessel centerline;

[0030] Step S532, single blood vessel majority decision: according to the majority principle of the positive and negative phases of the voxels in each blood vessel, the arterial and venous properties are determined;

[0031] Step S533, abnormal voxel correction: abnormal voxels that do not conform to the majority decision in a single blood vessel are uniformly corrected;

[0032] Step S534, set weighted voting:

[0033] For the current blood vessel branch, find the three-dimensional connected longest upstream and downstream blood vessel set;

[0034] The length of each blood vessel segment in the set is taken as the weight to calculate the overall arterial and venous property tendency value;

[0035] If the current blood vessel property does not conform to the set tendency value, the phase symbol is reversed;

[0036] Step S535, output the corrected velocity map.

[0037] The application also provides a non-contrast agent enhanced magnetic resonance imaging and flow velocity quantitative measurement system, comprising:

[0038] A low velocity encoding acquisition unit is configured to set a three-dimensional phase contrast blood vessel imaging scanning sequence for the peripheral blood vessels of the limbs, configure the readout direction to be consistent with the main running direction of the blood vessels, and perform double flow velocity encoding acquisition by using two different low velocity encoding values;

[0039] An image reconstruction and background phase correction module includes an image reconstruction unit and a background phase correction unit;

[0040] The image reconstruction unit is configured to jointly reconstruct the under-sampled data by applying low-rank constraint and variation regularization;

[0041] The background phase correction unit is configured to perform background field correction and unwrapping processing on the reconstructed phase data;

[0042] An arterial and venous separation module is configured to realize arterial and venous classification based on blood flow direction determination and blood vessel topology analysis.

[0043] Therefore, the non-contrast agent enhanced magnetic resonance imaging and flow velocity quantitative measurement method and system have the following beneficial technical effects:

[0044] (1) The present application significantly improves the signal-to-noise ratio and image clarity of vascular imaging by optimizing the multi-phase low-speed encoding acquisition scheme, image reconstruction algorithm, and precise background phase correction and unwrapping processing, and can more completely and clearly present the hand and foot vascular network, including the fine small blood vessel structure. At the same time, through the multi-stage arteriovenous separation technology, high-accuracy arteriovenous differentiation is realized, which provides a more reliable basis for the visualization of vascular structure.

[0045] (2) The present application breaks through the limitations of the prior art in low-speed and small blood vessel area imaging, is suitable for imaging of complex blood vessel areas such as extremity periphery, and can provide high-quality vascular images, which can maintain good imaging effect at both macrovascular and microvascular levels. This full-scale imaging capability provides more comprehensive imaging support for vascular research and clinical evaluation.

[0046] (3) The present application adopts non-contrast agent enhanced imaging technology, which avoids the risks and limitations that may be brought by the use of contrast agents, such as the use contraindication of patients with abnormal kidney function. This non-invasive and non-radiation imaging method is suitable for a wider patient group, improving the safety and universality of imaging. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 For comparison of the hand vascular image obtained by the present application with the foot vascular image obtained by the conventional three-dimensional phase contrast method, Figure 1 a in which is the maximum intensity projection image of the hand blood vessels of subject 1 by the conventional three-dimensional phase contrast method (velocity encoding = 5 cm / s), Figure 1 b in which is the maximum intensity projection image of the hand blood vessels of subject 1 by the non-contrast agent enhanced magnetic resonance imaging and flow velocity quantitative measurement method of the present application, Figure 1 c in which is the maximum intensity projection image of the hand blood vessels of subject 1 by the conventional three-dimensional phase contrast method (velocity encoding = 10 cm / s), Figure 1 d in which is the maximum intensity projection image of the hand blood vessels of subject 2 by the conventional three-dimensional phase contrast method (velocity encoding = 5 cm / s), Figure 1 e in which is the maximum intensity projection image of the hand blood vessels of subject 2 by the non-contrast agent enhanced magnetic resonance imaging and flow velocity quantitative measurement method of the present application, Figure 1 f in which is the maximum intensity projection image of the hand blood vessels of subject 2 by the conventional three-dimensional phase contrast method (velocity encoding = 10 cm / s);

[0048] Figure 2 For the hand vascular flow velocity image and arteriovenous separation image obtained by the present application;

[0049] Figure 3 For the foot vascular flow velocity image and arteriovenous separation image obtained by the present application;

[0050] Figure 4 A structural schematic diagram of a non-contrast agent enhanced magnetic resonance imaging and flow velocity quantitative measurement system. DETAILED DESCRIPTION

[0051] The technical solutions of the present application are further described below by means of the accompanying drawings and examples.

[0052] Unless otherwise defined, the technical terms or scientific terms used in the present application shall have the usual meanings understood by those skilled in the art to which the present application belongs.

[0053] Example 1

[0054] The non-contrast agent enhanced magnetic resonance imaging and flow velocity quantitative measurement method comprises the following steps:

[0055] Step S1, a three-dimensional phase contrast angiography scanning sequence is set for the peripheral blood vessels of the limbs, and the readout direction is configured to be consistent with the main running direction of the blood vessels.

[0056] When scanning the hand, the readout direction is from the wrist to the fingers;

[0057] When scanning the foot, the readout direction is from the ankle to the toes;

[0058] The phase encoding direction is the right-to-left direction.

[0059] In this embodiment, hand vascular imaging was performed on 20 healthy subjects, and foot vascular imaging was performed on 3 subjects, and the image data was analyzed.

[0060] Step S2, double flow velocity encoding acquisition is performed using two different low velocity encoding values.

[0061] The two different low velocity encoding values are: an ultra-low velocity encoding value of ≤5 cm / s and a reference velocity encoding value of ≥10 cm / s.

[0062] Table 1 Comparison of signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of vascular images under different low velocity encoding values

[0063] ;

[0064] The super-low velocity encoding value is used to improve the blood vessel contrast, improve the image signal-to-noise ratio, especially the slow blood vessels of fingertips, so as to obtain a more fine blood vessel structure image; however, the super-low velocity encoding value needs to apply a super-high amplitude bipolar gradient, which is easy to cause gradient nonlinearity, eddy current effect and significant phase artifact, and further affect the accuracy of blood flow phase. Therefore, in the quantitative analysis of blood flow velocity, high velocity encoding data also need to be selected for combination to obtain more reliable blood flow phase information. This not only helps to improve the accuracy of blood flow velocity measurement, but also facilitates subsequent background phase correction and phase unwrapping processing, so as to generate a more accurate velocity atlas.

[0065] In the embodiment, two velocity encoding values of 1cm / s and 13cm / s are set. The multi-phase low velocity encoding design can significantly improve the detection sensitivity of low-speed blood flow at the extremities, wherein the super-low velocity encoding of 1cm / s can obtain more complete microvascular blood flow information, and the higher velocity encoding of 13cm / s provides a stable reference for blood flow phase quantification.

[0066] In step S3, the low-rank constraint and the variation regularization are combined to reconstruct the undersampled data, the structural correlation between different velocity encoding data is fully utilized, and the image quality is effectively improved and the noise and artifact are suppressed. Specifically, the following is used:

[0067] ;

[0068] wherein, indicates an undersampling operator, indicates a Fourier transform, indicates a coil sensitivity, and the vector indicates the actual acquired spatial data, indicates an operator, extracts the bth data block from all velocity encoding images, and converts it into a Casorati matrix, and the low-rank constraint is locally applied by punishing the kernel norm of the Casorati matrix. indicates a TV regularization constraint along the VENC direction, indicates a weight of the local low-rank constraint, indicates a weight of the variation regularization constraint, indicates an image corresponding to the to-be-reconstructed spatial data, indicates a reconstructed image.

[0069] In step S4, the reconstructed phase data is subjected to background field correction and unwrapping processing.

[0070] To address the phase aliasing and phase wrapping problems that are prone to occur in low-speed encoding, first, the Maxwell bias field correction is applied to the reconstructed phase data to eliminate the background phase offset caused by the accompanying gradient field.

[0071] Subsequently, based on the local bias correction method, a blood vessel mask guided Gaussian filter estimation strategy is used for background phase correction. The specific process is as follows:

[0072] Because there is significant blood flow phase information in the blood vessel region, which may interfere with the estimation of the background phase offset, first, the segmented blood vessel structure and the phase difference data are masked to shield the blood vessel phase and accurately lock the background tissue region. Then, a three-dimensional Gaussian kernel is used to smooth filter the masked background phase difference data to estimate its background phase distribution, and the estimated value is subtracted from the original phase data to achieve phase error correction. Through this method, the interference of non-blood vessel parts can be excluded, and the accuracy of background correction can be improved.

[0073] The specific parameters are as follows: the standard deviation of the Gaussian kernel is [5, 5, 5] (the same standard deviation in the XYZ three directions). The filter size is 2ceil (2σ) + 1 = 21, ceil represents the ceiling function; (21 21 21, taking the upper integer of 2σ (standard deviation) in each direction to ensure that the kernel size is odd).

[0074] The phase unwrapping based on fast Fourier transform (FFT-based Unwrapping) is used. First, the unwrapped phase data after removing the background is converted to the frequency domain space through three-dimensional fast Fourier transform, then a frequency domain filter constructed by a discrete Laplace operator (the normalized coefficient is 6) is used for convolution operation, wherein the frequency domain kernel threshold 1e-10 is set for numerical stability to eliminate singular value influence, and finally the inverse Fourier transform is used to reconstruct the unwrapped phase field. This method effectively avoids the error propagation problem of the path integral method by solving the Poisson equation (Poisson equation) in the global frequency domain, and is particularly suitable for processing large-size three-dimensional phase data. While maintaining the accuracy of 1mm isotropic voxels, it realizes the robust elimination of phase discontinuity. represents the Laplace operator (the coefficient is 6 in the three-dimensional discrete form), represents the spatial second-order differential operation, represents the real unwrapped phase to be solved (continuous and no jump), and represents the input wrapped phase (the value range is [-π, π]).

[0075] Finally, the unwrapped phase and the segmented blood vessels are masked to obtain the preliminary blood flow velocity map.

[0076] Step S5, based on blood flow direction determination and blood vessel topology analysis to realize arteriovenous classification.​

[0077] Step S51: Perform primary arterial and venous marking based on phase sign direction, defining blood flow from wrist to fingers and from ankle to toes in the hand as positive (artery) and reverse blood flow as negative (venous).

[0078] Step S52: Extract the centerline of the blood vessel using a three-dimensional thinning algorithm and establish the topological connection relationship of the blood vessel branches;

[0079] Step S53: For each blood vessel branch, verify the arterial / venous attributes based on its upstream and downstream connectivity and spatial continuity. When there are attribute conflicts in the branch nodes, a majority voting mechanism based on blood vessel length weighting is used for correction.

[0080] Step S531, Voxel-Centerline Association: First, calculate the distance from each phase voxel to all centerline points in the vascular skeleton based on the Euclidean distance metric to determine its nearest neighbor skeleton point; then, combine MRA vascular segmentation data and verify whether the voxel and the candidate skeleton point belong to the same vascular branch through connected component analysis. Only when the dual conditions of "minimum distance" and "same connected component" are met simultaneously, the voxel is assigned to the corresponding vascular centerline, and finally a complete voxel-centerline mapping relationship is established.

[0081] Step S532, Single vessel majority determination: Traverse vessel branches in descending order of voxel count, and determine arterial and venous attributes based on the majority principle of positive and negative phase of voxels in each vessel.

[0082] Step S533, Abnormal Voxel Correction: Perform unified correction on abnormal voxels in a single blood vessel that do not conform to the majority of judgments;

[0083] Step S534, Weighted Voting:

[0084] For the current vascular branch, find the longest set of upstream and downstream vessels that are connected in three dimensions;

[0085] The overall arteriovenous attribute tendency value is calculated using the length of each blood vessel segment in the set as the weight.

[0086] If the current vessel properties do not match the set tendency value, then reverse its phase sign;

[0087] Step S535: Output the corrected velocity spectrum.

[0088] Step S54: For the phase reversal region caused by local magnetic field inhomogeneity, automatic correction is performed in combination with the consistency of blood flow direction of adjacent blood vessel segments.

[0089] Step S55: Output a three-dimensional model of the arterial tree and vein tree with complete topological connections.

[0090] The results obtained using the above method are as follows:Figures 1-3 are shown.

[0091] As Figure 1 , the maximum intensity projections of two healthy subjects' hands are shown. The traditional single velocity encoding method has a defect: different VENC needs to be used for different people, which is generally set to 120% of the maximum target flow rate, but due to the difference in hand blood flow rate between individuals, this value is generally difficult to accurately estimate. This is specifically reflected in the figure, where one subject's velocity encoding = 5cm / s image quality is better, while the other subject's velocity encoding = 10cm / s image quality is better. In contrast, the multi-period low velocity encoding method proposed in the present application does not need to adjust the parameters for individuals, and in both subjects, the optimal imaging effect is achieved: not only does it retain the continuous structure of large blood vessels such as the palmar arch, but it also significantly outperforms traditional methods in displaying the tortuous superficial veins and tiny blood vessels at the end of the fingers, and the background noise suppression effect is stable.

[0092] Figure 2 and Figure 3 are the sagittal and coronal views of the whole processed blood flow velocity map of the hand and foot, respectively, with red representing arteries and blue representing veins. The deeper the color, the faster the blood flow.

[0093] As shown in Tables 2 and 3, compared with the traditional phase contrast method (velocity encoding value = 5cm / s), the number of blood vessels of the proposed method increased by 52.8% (15329 vs 10031), the number of blood vessel branches increased by 331.7% (449 vs 104), the CNR of the finger region increased by 1.72 times (37.0 vs 21.5), and the SNR of the wrist and palm increased by about 45%. Compared with the traditional method (velocity encoding value = 10cm / s), the advantages are more pronounced: the number of blood vessels increased by 108.8% (15329 vs 7341), the number of blood vessel branches increased by 388.0% (449 vs.92), the CNR of the fingers increased by 4.3 times (37.0 vs 8.6), and the SNR of all parts increased by more than 70%.

[0094] Table 2 Image quality evaluation

[0095] ;

[0096] Table 3 Blood vessel feature evaluation

[0097] ;

[0098] Example Two

[0099] As Figure 4The low velocity encoding acquisition unit is used for setting a three-dimensional phase contrast angiography scanning sequence for the peripheral blood vessels of the limbs, configuring a readout direction to be consistent with a main running direction of the blood vessels, and performing double flow velocity encoding acquisition by using two different low velocity encoding values.

[0100] The image reconstruction and background phase correction module comprises an image reconstruction unit and a background phase correction unit.

[0101] The image reconstruction unit is used for jointly reconstructing the undersampled data by using a low rank constraint and a variation regularization.

[0102] The background phase correction unit is used for performing background field correction and unwrapping processing on the reconstructed phase data.

[0103] The arteriovenous separation module is used for realizing arteriovenous classification based on blood flow direction determination and blood vessel topology analysis.

[0104] It should be noted that the contents not described in detail in the present application are all prior art and are well known to those skilled in the art.

[0105] Therefore, the non-contrast agent enhanced magnetic resonance imaging and flow velocity quantitative measurement method and system can realize high-quality imaging in a low-velocity and small-vessel region, have arteriovenous separation capability, and can accurately quantify blood flow velocity, so as to meet the needs of clinical peripheral vessel imaging.

[0106] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application but not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that: it can still modify or equivalently replace the technical solutions of the present application, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. Method for non-contrast agent enhanced magnetic resonance angiography and flow quantification, characterized in that, The method comprises the following steps: Step S1, setting a three-dimensional phase contrast angiography scanning sequence for peripheral blood vessels of limbs, and configuring a readout direction to be consistent with a main running direction of the blood vessels; Step S2, performing double flow encoding acquisition by using two different low velocity encoding values; Step S3, jointly reconstructing the undersampled data by applying low rank constraint and variation regularization; Step S4, performing background field correction and unwrapping processing on the reconstructed phase data; Step S5, realizing arteriovenous classification based on blood flow direction determination and blood vessel topology analysis.

2. The non-contrast enhanced magnetic resonance angiography and flow quantification method of claim 1, wherein, In step S1, the readout direction is from the wrist to the fingers during hand scanning, and the readout direction is from the ankle to the toes during foot scanning; the phase encoding direction is the right-to-left direction.

3. The non-contrast enhanced magnetic resonance angiography and flow quantification method of claim 1, wherein, In step S2, the two different low velocity encoding values are: an ultralow velocity encoding value of ≤5 cm / s and a reference velocity encoding value of ≥10 cm / s.

4. The non-contrast enhanced magnetic resonance angiography and flow quantification method of claim 1, wherein, In step S3, the undersampled data is jointly reconstructed by applying low rank constraint and variation regularization, and the specific process is as follows: ； wherein, denotes an under-sampling operator, denotes a Fourier transform, denotes a coil sensitivity, vector denotes the actually acquired spatial data, denotes an operator, denotes a TV regularization constraint along the VENC direction, denotes a weight for the local low-rank constraint, denotes a weight for the variation regularization constraint, denotes the image to be reconstructed, spatially corresponding image, denotes the reconstructed image.

5. The non-contrast enhanced magnetic resonance angiography and flow quantification method of claim 1, wherein, Step S4 comprises: Step S41, applying Maxwell bias field correction to the reconstructed phase data; Step S42, performing Gaussian filtering background phase estimation based on a blood vessel mask; Step S43, performing frequency domain Laplace filtering unwrapping.

6. The non-contrast enhanced magnetic resonance angiography and flow quantification method of claim 1, wherein, Step S5 comprises: Step S51, performing arteriovenous primary labeling based on the phase sign direction; Step S52, extracting a blood vessel centerline by a three-dimensional refinement algorithm, and establishing a topology connection relationship of blood vessel branches; Step S53, for each blood vessel branch, verifying the arteriovenous attribute based on the upstream and downstream connection relationship and spatial continuity, and when there is an attribute conflict in a branch node, a majority voting mechanism based on blood vessel length weighting is used for correction; Step S54, for a phase inversion area caused by local magnetic field inhomogeneity, automatic correction is performed in combination with the consistency of the blood flow direction of adjacent blood vessel segments; Step S55, outputting three-dimensional models of an arterial tree and a venous tree with complete topology connection relationship.

7. The non-contrast enhanced magnetic resonance angiography and flow quantification method of claim 6, wherein, In step S53, the majority voting mechanism based on blood vessel length weighting comprises: Step S531, voxel-centerline association: according to the Euclidean distance and connected domain relationship between a voxel and a blood vessel skeleton, phase voxels in a velocity atlas are paired with a blood vessel centerline; Step S532, single blood vessel majority determination: traversing blood vessel branches in descending order of voxel quantity, the arteriovenous attribute is determined according to the majority principle of the positive and negative phases of the voxels in each blood vessel; Step S533, abnormal voxel correction: abnormal voxels that do not conform to the majority determination in a single blood vessel are uniformly corrected; Step S534, set weighted voting: For the current blood vessel branch, find a set composed of three-dimensionally connected longest upstream and downstream blood vessels; Taking the length of each blood vessel segment in the set as a weight, an overall arteriovenous attribute tendency value is calculated; If the current blood vessel attribute does not conform to the set tendency value, the phase sign of the current blood vessel attribute is reversed; Step S535, output the corrected velocity atlas.

8. A non-contrast agent enhanced magnetic resonance angiography and flow quantification system, characterized in that, Comprise: A low velocity encoding acquisition unit is configured to set a three-dimensional phase contrast angiography scanning sequence for peripheral blood vessels of limbs, configure a readout direction to be consistent with a main running direction of the blood vessels, and perform double flow encoding acquisition by using two different low velocity encoding values. The image reconstruction and background phase correction module comprises an image reconstruction unit and a background phase correction unit. The image reconstruction unit is configured to jointly reconstruct the undersampled data by applying a low-rank constraint and a variation regularization. The background phase correction unit is configured to perform background field correction and unwrapping processing on the reconstructed phase data. The artery-vein separation module is configured to realize artery-vein classification based on blood flow direction determination and vessel topology analysis.

Images