MR mammography

By using radial or helical k-space trajectory acquisition and deep learning algorithm reconstruction techniques in MR imaging, the problems of distinguishing healthy tissue from cancerous abnormalities and insufficient identification of microcalcifications in MR breast imaging have been solved, achieving high-resolution breast photography and tissue characterization, and improving the accuracy of breast cancer screening.

CN116724246BActive Publication Date: 2026-11-10KONINKLIJKE PHILIPS NV
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202280010167.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2021-01-14
Filing Date
2022-01-12
Publication Date
2026-11-10
Estimated Expiration
2042-01-12

AI Technical Summary

Technical Problem

Existing MR breast imaging methods are ineffective in distinguishing between healthy tissue and cancerous abnormalities, and cannot identify microcalcifications, leading to false positives and an increase in breast biopsies. Meanwhile, conventional mammography has insufficient resolution in dense breast tissue.

Method used

MR signals were acquired using radial or helical k-space trajectories, and high-resolution water, fat, B0, and T2* images were reconstructed using deep learning algorithms. High-resolution mammograms were synthesized to identify microcalcifications and improve tissue characterization.

Benefits of technology

It achieves high-resolution imaging of breast tissue, enabling the identification of microcalcifications smaller than 1 mm in diameter, thus improving the accuracy and efficiency of breast cancer screening and reducing false positive results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116724246B_ABST
    Figure CN116724246B_ABST
Patent Text Reader

Abstract

This invention relates to a method for MR imaging of an object (i.e., at least a portion of a female breast). The method of the invention includes the following steps: a) subjecting the object (10) to an imaging sequence comprising an RF excitation pulse and a switched magnetic field gradient, wherein multiple echo signals are generated at different echo times (TE1, TE2, TE3) after each RF excitation pulse; b) acquiring the echo signals along a set of radial or helical k-space trajectories covering a given k-space region, wherein each echo signal generated after the RF excitation pulse is assigned a different orientation of the radial or helical trajectory in k-space; c) reconstructing a single echo image for each echo time (TE1, TE2, TE3) based on the acquired echo signals; d) deriving high-resolution water, fat, BO, and / or apparent lateral relaxation time (T2*) maps based on the single echo images using a deep learning algorithm; and d) synthesizing a high-resolution mammogram based on the water, fat, BO, and / or T2* maps. Furthermore, this invention relates to an MR device (1) and a computer program for the MR device (1).
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of magnetic resonance (MR) imaging. It relates to a method for performing MR imaging on an object placed within the examination volume of an MR device. The invention also relates to an MR device and a computer program to be run on the MR device. Background Technology

[0002] Image-forming MR methods, which utilize the interaction between magnetic fields and nuclear spins to form two-dimensional or three-dimensional images, are now widely used, especially in the field of medical diagnostics, because they are superior to other imaging methods in many ways for imaging soft tissues, do not require ionizing radiation, and are generally non-invasive.

[0003] Typically, according to the MR method, the patient's body is positioned in a strong, uniform magnetic field B0, the direction of which simultaneously defines the measurement-related axis of the coordinate system (usually the z-axis). The magnetic field B0 generates different energy levels for individual nuclear spins depending on the magnetic field strength, which can be excited (spin resonance) by applying an electromagnetic alternating field (RF field) of a defined frequency (so-called Larmor frequency or MR frequency). From a macroscopic perspective, the distribution of individual nuclear spins produces global magnetization, which can be deviated from equilibrium by applying an electromagnetic pulse (RF pulse) of appropriate frequency, with the corresponding magnetic field B1 extending perpendicular to the z-axis, causing the magnetization to precess about the z-axis. Precession describes the surface of a cone, the aperture angle of which is called the flip angle. The magnitude of the flip angle depends on the strength and duration of the applied electromagnetic pulse. In the case of a so-called 90° pulse, the magnetization deflects from the z-axis to the transverse plane (flip angle 90°).

[0004] After the RF pulse terminates, the magnetization relaxes back to its original equilibrium state, where the magnetization in the z-direction is re-established with a first time constant T1 (spin lattice or longitudinal relaxation time), and the magnetization in the direction perpendicular to the z-direction relaxes with a second, shorter time constant T2 (spin-spin or transverse relaxation time). The transverse magnetization and its changes can be detected by means of a receiving RF coil arranged and oriented within the examination volume of the MR device in such a way that changes in magnetization are measured in the direction perpendicular to the z-axis.

[0005] To achieve spatial resolution within the body, time-varying magnetic field gradients extending along three principal axes are superimposed on a uniform magnetic field B0, resulting in a linear spatial dependence of the spin resonant frequency. The signal picked up in the receiving coil then contains components of different frequencies, which can be correlated to different locations within the body. The signal data obtained via the receiving coil corresponds to the spatial frequency domain and is referred to as k-space data. k-space data typically comprises multiple lines acquired with different phase codes. Each line is digitized by collecting multiple samples. A set of k-space data is converted into an MR image using a Fourier transform.

[0006] Breast cancer is one of the most common cancers among women. Recent studies have found that MR imaging can locate small breast lesions that mammography sometimes misses. MR imaging can also help detect breast cancer in women with breast implants and in younger women who tend to have denser breast tissue. Conventional (X-ray based) mammography may be less effective in these cases. Because MR imaging does not use ionizing radiation, it can be used to screen women, especially those aged 40 or younger, and increase the number of annual screenings for women at high risk of breast cancer.

[0007] Despite its significant advantages over conventional mammography, breast MRI also has potential limitations. For example, it is not always able to distinguish between healthy tissue and cancerous abnormalities, which can lead to unnecessary breast biopsies. This is often referred to as a "false positive" test result. Another drawback of breast MRI is that it has historically failed to identify tiny calcium deposits (microcalcifications) that can indicate breast cancer.

[0008] Regardless of the high sensitivity of MR imaging and its particularly compelling results in dense breast tissue, routine mammography remains the primary breast screening tool due to its higher resolution and generally higher specificity.

[0009] WINKELMANN et al. (JOURNAL OF MAGNETIC RESONANCE IMAGING, 24:939-944, 2006) described a technique using simultaneous MR imaging and T2* mapping with radial multigradient echo sequences. Multiple undersampled images were reconstructed, each for a different echo time. These images were simultaneously used to compute high-resolution images and T2* maps.

[0010] KHOULI et al (PROCEEDINGS OF THE INTERNATIONAL SOCIETY FOR MAGNETICRESONANCE IN MEDICINE,18 THAnnual Meeting (page 2489, 2010) describes a method for detecting breast microcalcifications via 3 Tesla MR imaging. 3D SPGRE ultrashort echo time-series with radial reconstruction is used in conjunction with susceptibility-weighted imaging.

[0011] EP 3 531 154 A1 discloses an MR imaging method in which water / fat separation and B0 and T2* mapping are performed using a multi-gradient echo sequence with radial or helical acquisition. An image with a specified contrast is synthesized based on the acquired echo signal data, the B0 map, and / or the T2* map. Summary of the Invention

[0012] It is easy to see from the foregoing that an improved MR-based mammography method is needed.

[0013] According to the present invention, a method for performing MR imaging on an object positioned within the examination volume of an MR device is disclosed, the method comprising the following steps:

[0014] a) subjecting the object to an imaging sequence comprising RF excitation pulses and switched magnetic field gradients, wherein multiple echo signals are generated at different echo times following each RF excitation pulse.

[0015] b) Acquire the echo signals along a set of radial or helical k-space trajectories covering a given k-space region, wherein each echo signal generated after the RF excitation pulse is assigned a different orientation of the radial or helical trajectory in the k-space.

[0016] c) Reconstruct a single echo image for each echo time based on the acquired echo signals;

[0017] d) Use deep learning algorithms to derive high-resolution water, fat, B0, and / or apparent lateral relaxation time (T2*) maps from the single echo image; and

[0018] d) Synthesize high-resolution mammograms based on the water, fat, B0, and / or T2* maps.

[0019] The high-resolution intermediate image is reconstructed based on a combination of echo signals acquired at the different echo times (TE1, TE2, TE3) from the entire k-space region covered by the set of radial or spiral k-space trajectories, wherein the derivation of the high-resolution water, fat, B0 and / or T2* images is additionally based on the high-resolution intermediate image.

[0020] According to the present invention, multiple echo signals are acquired at different echo times. Echo signals are acquired along radial or helical k-space trajectories (including known acquisition schemes such as KOOSH spheres, Archimedes spirals, helical leaf sequences, etc.). Radial or helical k-space trajectories are superior to conventional Cartesian k-space trajectories due to their inherent motion robustness. Using radial or helical k-space sampling, the center of the k-space is oversampled and continuously updated. This redundancy can be advantageously utilized to detect and correct for motion, B0, and T2* effects.

[0021] Each radial or helical trajectory in the echo signal generated after the RF excitation pulse is assigned a different orientation in k-space. This implies performing "echo-time distributed" sampling of k-space. A given radial or helical k-space trajectory is uniquely associated with a specific echo time. Multiple acquisitions of the same radial or helical k-space trajectory with different echo times should be avoided to achieve very high acquisition speeds.

[0022] A single echo image is reconstructed for each echo time based on echo signals acquired from a covered k-space region. A k-space weighted image contrast (KWIC) filter can be advantageously used for reconstructing single echo images (see Song et al., Magn. Reson. Med., 44, 825-832, 2000). For example, each frame in a single echo image can be reconstructed from undersampled k-space data attributed to the corresponding echo time. The acquired echo signals attributed to a specific echo time are typically undersampled, at least at the periphery of k-space. Therefore, compressed sensing (CS) can be advantageously used to reconstruct single echo images from undersampled signal data. The (KWIC-filtered) echo signals attributed to a specific echo time are irregularly distributed in k-space. CS theory is well known to have great potential for significant signal data reduction. In CS theory, by applying a suitable regularization algorithm, a signal dataset with a sparse representation in the transform domain can be recovered from undersampled measurements. As a mathematical framework for signal sampling and reconstruction, CS specifies the conditions under which signal datasets can be accurately or at least reconstructed with high image quality, even when the k-space sampling density is much lower than the Nyquist criterion, and it also provides methods for such reconstruction.

[0023] Furthermore, a single high-resolution intermediate image is reconstructed from the entire echo signal (i.e., including all echo signals acquired for different echo times) and from the entire k-space region covered by a set of radial or spiral k-space trajectories.

[0024] As a next step, a deep learning algorithm is used to derive high-resolution water and / or fat and / or B0 and / or apparent lateral relaxation time (T2*) maps from a single echo image. The use of deep learning schemes to compute “super-resolution” MR images from lower-resolution images is well-known in the art. The Dixon super-resolution deep learning reconstruction algorithm according to the present invention is implemented using appropriate training data (including single echo images as input) and associated high-resolution (super-resolution) water, fat, B0, and T2* maps as the output of an artificial neural network. To improve the quality of the high-resolution water, fat, B0, and T2* maps, the deep learning-based reconstruction is additionally based on high-resolution intermediate images reconstructed from a combination of echo signals acquired for all different echo times, as described above.

[0025] High-resolution tissue classification maps and / or calcification maps can be derived from water, fat, B0, and / or T2* maps. Furthermore, other maps, such as fat classification maps, fat fraction maps, etc., can be derived. T2* is primarily generated by inhomogeneities in the main magnetic field B0. These inhomogeneities are a result of inherent inhomogeneities in the main magnetic field itself and field distortions caused by the magnetic susceptibility generated by the tissue. Therefore, according to the present invention, high-resolution T2* maps derived from a single echo image (ideally combined with water, fat, and B0 maps) can be used to derive (high-resolution) tissue classification maps. In particular, microcalcifications that may be associated with cancer in breast tissue can cause field distortions due to characteristic magnetic susceptibility in the local environment. The present invention leverages this by specifically using T2* maps to derive high-resolution calcification maps with particularly high diagnostic value. Combined with super-resolution Dixon deep learning reconstruction, microcalcifications smaller than 1 mm in diameter can be identified.

[0026] Finally, a high-resolution mammogram is synthesized based on the water, fat, B0, and / or T2* maps. The synthesized mammogram is calculated by assigning a Henlein unit value to each pixel or voxel. The Henlein unit value assigned to each image location provides the X-ray radiation attenuation properties of the tissue, making it possible to calculate a real mammogram with the exact same contrast characteristics as radiologists are accustomed to in routine mammographic screening. The assignment of Henlein units to image locations is performed based on high-resolution water, fat, B0, and / or T2* maps. Thus, Dixon imaging, which provides water / fat separation, is known in the art to be effective for the purpose of generating a synthesized X-ray image. Any additional contrast information for the same region is useful for correctly determining the Henlein units for each image location. Therefore, the synthesis of the mammogram is ideally also based on the derived tissue classification and / or calcification maps.

[0027] To obtain a uniform k-space distribution of the echo signal, a radial or spiral k-space trajectory (located in k...) is used. x / k yThe rotation angle of the plane can be increased by the golden angle during acquisition, which corresponds to 180° multiplied by the golden ratio. To avoid large gradient increments, small golden angle increments may be preferred. Another possibility is to acquire echo signals in a continuous sequence in k-space along a radial or helical k-space trajectory. This has the advantage of correcting phase errors caused by the system from even and odd echo sampling. Another option is to acquire the echo signals in k-space during acquisition. z The k-space trajectory of the individual echo is distributed in k-directions by increasing in direction. z In the direction (i.e., in the direction perpendicular to the plane of rotation of the radial or helical k-space trajectory). A uniform distribution (as opposed to a distribution on a regular grid in k-space) supports the use of deep learning algorithms to derive high-resolution water, fat, B0, and / or apparent lateral relaxation time (T2*) maps from a single echo image. To further support deep learning-based derivation of high-resolution maps, shifts in the echo along the radial or helical k-space trajectory can be applied. Furthermore, the acquisition bandwidth (i.e., the duration and / or sampling frequency of the corresponding echo acquisition) can vary between different captures in the imaging sequence and / or between acquisitions of different echo signals.

[0028] In a preferred embodiment, the first echo signal following each RF excitation is generated at an ultrashort echo time (UTE) to extend the range of accessible echo time values. Known partial echo techniques can also be applied for this purpose. Echo shifting can be alternatively or additionally employed to improve echo time coverage and optimize T2* mapping. Ultrashort echo times are particularly well-suited for identifying calcifications characterized by very short apparent lateral relaxation times. The result is improved fitting and T2* mapping. A 3D radial k-space acquisition scheme (“KOOSH sphere”) can be used to sample a given k-space region in conjunction with UTE techniques. Due to MR hardware limitations (transmit / receive switching dead time), each center-outward radial k-space trajectory sampled at the ultrashort echo time will have gaps at the center of the k-space. To fill these gaps, the known PETRA scheme can be employed. The PETRA scheme fills the gaps by performing individual (Cartesian) sampling of the central k-space region.

[0029] In another preferred embodiment, the motion of the object occurring during acquisition is derived from at least one of the echo signals. This can be used to correct for motion detected in the step of reconstructing a single echo image. For example, one or more echo signals representing radial or helical k-space samples can be used as an inherent navigator for detecting motion or respiratory states. This can be advantageously combined with the method of the present invention to reduce motion-induced artifacts. Specifically, detected motion (e.g., detected displacement of the imaging object) is attributed to one of several discrete motion states experienced by the examined patient or the examined body part, for example, during respiration. A B0 map can then be derived for each motion state based on the echo signal assigned to the corresponding motion state. In this way, incorporating inherent motion and motion-induced B0 determination and correction achieves more accurate T2* mapping. By addressing motion-induced B0 variations, the accuracy of tissue mapping in breast imaging can be significantly improved. In particular, respiratory-induced B0 variations have a significant impact on breast tissue mapping. The intermediate information generated by motion and motion-induced B0 detection can be provided to deep learning algorithms to determine accurate tissue classification maps. This improved fitting allows the technique to also be combined with a susceptibility-weighted imaging strategy (SWI), thereby enabling the determination of, for example, more accurate phase images.

[0030] As a result, this invention provides a high-resolution MR mammography and tissue characterization technique for detecting calcifications. Tissue characterization provides additional information to the calcification map. Based on the same acquisition, tissue oxygenation levels, such as fat classification by fat fraction, can be provided within the same scan. It provides tumor-related diagnostic and management information and is well-suited for breast screening and general non-contrast-enhanced breast MR imaging.

[0031] The method of the present invention, as described herein, can be performed using an MR device comprising: at least one main magnet coil, a plurality of gradient coils, at least one body RF coil, a control unit, and a reconstruction unit. The at least one main magnet coil is used to generate a uniform and stable magnetic field B0 within an examination volume. The plurality of gradient coils are used to generate switched magnetic field gradients in different spatial directions within the examination volume. The at least one body RF coil is used to generate RF pulses within the examination volume and / or to receive MR signals from the body of a patient positioned within the examination volume. The control unit is used to control the temporal succession of the RF pulses and the switched magnetic field gradients. The reconstruction unit is used to reconstruct MR images based on the received MR signals. The method of the present invention can be implemented by correspondingly programming the reconstruction unit and / or control unit of the MR device.

[0032] The method of the present invention can be advantageously performed on most MR devices currently used in clinical practice. For this purpose, only a computer program controlling the MR device is needed to cause the MR device to perform the above-described method steps of the present invention. This computer program may reside on a data carrier or may exist on a data network for download and installation in the control unit of the MR device. Attached Figure Description

[0033] The accompanying drawings disclose preferred embodiments of the invention. However, it should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of the invention. In the drawings:

[0034] Figure 1 An MR apparatus for performing the method of the present invention is shown;

[0035] Figure 2 A schematic (simplified) radial k-space sampling scheme according to an embodiment of the present invention is shown;

[0036] Figure 3 The method of the present invention is illustrated in the form of a flowchart;

[0037] Figure 4 An MR mammogram synthesized using the method of the present invention is shown. Detailed Implementation

[0038] refer to Figure 1 The diagram illustrates an MR apparatus 1. The apparatus includes a superconducting or resistive main magnet coil 2, which creates a substantially uniform, time-constant main magnetic field B0 along the z-axis passing through the examination volume. The apparatus also includes a set of (first-order, second-order, and, where applicable, third-order) shimming coils 2', wherein the current flowing through each individual shimming coil in the set 2' is controllable for the purpose of minimizing B0 deviations within the examination volume.

[0039] The magnetic resonance generation and manipulation system applies a series of RF pulses and switched magnetic field gradients to reverse or excite nuclear magnetic spin, induce magnetic resonance, refocus magnetic resonance, manipulate magnetic resonance, encode magnetic resonance spatially and otherwise, saturate spin, etc., thereby performing MR imaging.

[0040] More specifically, gradient amplifier 3 applies current pulses or waveforms along the x, y, and z axes of the examination volume to selected coils in the whole-body gradient coils 4, 5, and 6. A digital RF frequency transmitter 7 transmits RF pulses or pulse packets to the body RF coil 9 via transmit / receive switch 8, thereby transmitting RF pulses into the examination volume. A typical MR imaging sequence consists of packets of short-duration RF pulse segments that, together with any applied magnetic field gradient, enable selective manipulation of the MRI signal. RF pulses are used to saturate, excite, reverse, refocus, or manipulate the resonance and select portions of the body 10 located within the examination volume. The MR signal is also picked up by the body RF coil 9.

[0041] To generate MR images of a limited area of ​​body 10 or to accelerate scanning by means of parallel imaging, a set of local array RF coils 11, 12, 13 are placed adjacent to the area selected for imaging. The array coils 11, 12, 13 are capable of receiving MR signals generated by the RF emission of the body coils.

[0042] The resulting MR signal is picked up by the body RF coil 9 and / or the array RF coils 11, 12, 13 and demodulated by a receiver 14, which preferably includes a preamplifier (not shown). The receiver 14 is connected to the RF coils 9, 11, 12 and 13 via a transmit / receive switch 8.

[0043] The host computer 15 controls the shimming coil 2', gradient pulse amplifier 3, and transmitter 7 to generate any of a plurality of MR imaging sequences, such as echo plane imaging (EPI), echo volume imaging, gradient and spin echo imaging, fast spin echo imaging, etc. For a selected sequence, receiver 14 rapidly and continuously receives one or more MR signals after each RF excitation pulse. Data acquisition system 16 performs analog-to-digital conversion of the received signals and converts each MR data sample into a digital format suitable for further processing. In modern MR devices, data acquisition system 16 is a separate computer dedicated to acquiring raw image data.

[0044] Ultimately, the raw digital image data is reconstructed into an image representation by reconstruction processor 17, which applies Fourier transform or other suitable reconstruction algorithms. MR images can represent planar slices across the patient, arrays of parallel planar slices, three-dimensional volumes, etc. The images are then stored in an image memory, accessible in which slices, projections, or other parts of the image representation can be converted into appropriate formats for visualization, for example via video monitor 18, which provides a human-readable display of the resulting MR images.

[0045] The host computer 15 is programmed to perform the methods of the present invention described above and below.

[0046] In embodiments of the invention, an imaging sequence is applied, comprising an RF excitation pulse and a magnetic field gradient that switches in the readout and phase encoding directions x and y and the slice selection direction z. Multiple sets of echo signals are acquired in multiple repetitions (shots) of the sequence using different gradient waveforms in the x / y and / or z directions to fully cover the desired region of k-space via a radial sampling mode. The timing and amplitude of the readout gradient in the x / y directions are selected to provide different echo times TE1, TE2, ..., TE... N Preferably, the imaging sequence is a UTE sequence, such that the echo signal immediately following each RF excitation is generated with an ultra-short echo time.

[0047] A basic UTE pulse sequence with a half-excitation pulse and radial imaging from the center of k-space can be employed. The magnetic field gradient is switched in the z-direction for phase encoding to achieve a Cartesian k-space sampling mode in that direction (based on the known "star stack" sampling scheme for 3D image reconstruction). Shift k-space techniques for super-resolution imaging can be advantageously applied. z Slice-coded sampling (see van Reeth et al., Concepts in Magnetic Resonance Part A, Vol. 40A(6), 306-325, 2012). Radial k-space sampling is used, with oversampling at the center of the k-space. Multiple echo signals with variable echo times can be acquired by moving or delaying echo acquisition and / or by acquiring multiple consecutive echoes. Reverse echo acquisition can be applied. "Echo time-distributed" sampling of the k-space is performed. This means that the echo times are distributed throughout the k-space, with oversampling at the center of the k-space. This is in Figure 2 The diagram is shown in the image. Figure 2 The orientation of the radial k-space trajectory for radial imaging is shown, with four radial acquisitions for each shot at three different echo times TE1, TE2, and TE3 and the UTE sequence used. The acquisition sequence of the radial k-space trajectory is indicated by numbers 1, 2, ... 12. This method achieves very fast acquisition, and the acquired echo signal data has an optimal distribution in k-space. Echo signal data is acquired in the manner described. Figure 3 The process is executed in step 31 of the flowchart.

[0048] As the next step ( Figure 3 In step 32), a single echo image is reconstructed for each echo time based on the echo signal acquired from the central portion of the covered k-space region. A k-space weighted image contrast (KWIC) filter can be used to reconstruct the single echo image.

[0049] Furthermore, in step 33, a single high-resolution intermediate image is reconstructed based on the entire echo signal (i.e., including all echo signals acquired for different echo times) and the entire k-space region covered by the set of radial or spiral k-space trajectories.

[0050] In step 34, the motion that occurred during acquisition is derived from at least one of the echo signals. One or more of the radial k-space samples can be used as an inherent navigator for detecting motion or respiratory states. Specifically, the detected motion is attributed to one of multiple discrete motion states. A low-resolution B0 map for each motion state is then derived from the echo signals assigned to the corresponding motion states.

[0051] In step 35, a deep learning algorithm is used based on a single echo image (and the shifted k) z (Slice encoding) derives high-resolution water, fat, B0, and T2* maps. To improve the quality of the high-resolution water, fat, B0, and T2* maps, the deep learning-based reconstruction is also based on the high-resolution intermediate images reconstructed in step 33. Furthermore, intermediate information generated by motion and B0 detection caused by motion in step 34 are provided to the deep learning algorithm to determine accurate tissue classification maps. By addressing motion-induced B0 variations, the accuracy of the T2* mapping is significantly improved.

[0052] In step 36, high-resolution tissue classification maps and calcification maps are derived based on water, fat, B0 and / or T2* maps.

[0053] Finally, in step 37, a high-resolution mammogram is synthesized based on water, fat, B0, and / or T2* maps. The synthesized mammogram is calculated by assigning Henle unit values ​​to each pixel or voxel and calculating pixel or voxel intensity based on the resulting X-ray radiation attenuation. The assignment of Henle units to image locations is performed based on high-resolution water, fat, B0, and / or T2* maps. The synthesis of the mammogram is also based on the derived tissue classification and / or calcification maps to correctly determine the Henle units for each image location.

[0054] Optional additional steps (not depicted) may be included, such as post-processing steps to remove system defects (e.g., deblurring) and / or deep learning reconstruction using deterministic graphs.

[0055] The possible image output for the user is high-resolution water, fat, and T2* maps reconstructed by deep learning, and especially synthetic mammograms, such as... Figure 4 As shown. The exported tissue classification and calcification maps can also be displayed to the user to support diagnosis.

Claims

1. A method for performing MR imaging on an object (10) positioned in an examination volume of an MR device (1), the method comprising the following steps: a) subjecting the object (10) to an imaging sequence comprising RF excitation pulses and switched magnetic field gradients, wherein multiple echo signals are generated at different echo times (TE1, TE2, TE3) after each RF excitation pulse. b) Acquire the echo signals along a set of radial or helical k-space trajectories covering a given k-space region, wherein each echo signal generated after the RF excitation pulse is assigned a different orientation of the radial or helical trajectory in the k-space. c) Reconstruct single echo images for each echo time (TE1, TE2, TE3) based on the acquired echo signals; d) Use deep learning algorithms to derive high-resolution water, fat, B0, and / or apparent lateral relaxation times (T2) from the single echo image. (Figure); and e) Based on the water, fat, B0 and / or T2 Images are used to synthesize high-resolution mammograms. The high-resolution intermediate image is reconstructed based on a combination of echo signals acquired at different echo times (TE1, TE2, TE3) from the entire k-space region covered by the set of radial or spiral k-space trajectories. This reconstruction includes high-resolution water, fat, B0, and / or T2 data. The derivation of the figure is based on both the single echo image and the high-resolution intermediate image.

2. The method according to claim 1, wherein, According to the water, fat, B0 and / or T2 The diagram is used to derive tissue classification and / or calcification diagrams.

3. The method according to claim 1 or 2, wherein, The first echo signal after each RF excitation is generated at the ultrashort echo time (UTE).

4. The method according to claim 1 or 2, wherein, Echo signals are acquired separately from one or more missing portions of the k-space region.

5. The method according to claim 1 or 2, wherein, In addition, at least one of the following is additionally based on the high-resolution intermediate image: the synthesis of the high-resolution mammogram and the derivation of the tissue classification map and / or the calcification map.

6. The method according to claim 1 or 2, wherein, The set of radial or spiral k-space trajectories oversamples the central portion of the k-space.

7. The method according to claim 1 or 2, wherein, The rotation angle of the radial or helical k-space trajectory increases during acquisition, resulting in a uniform k-space distribution of the echo signal.

8. The method according to claim 1 or 2, wherein, The k-space trajectory of the individual echo in k z Distributed in direction.

9. The method according to claim 1 or 2, wherein, The motion of the object (10) that occurred during the acquisition of the echo signals is derived from at least one of the echo signals, and In the step of reconstructing the single echo image, the detected motion is corrected, and / or The detected motion is attributed to one of a plurality of motion states, wherein a B0 diagram is derived for each motion state based on the echo signal assigned to the corresponding motion state.

10. The method according to claim 9, wherein, For the aforementioned high-resolution water, fat, B0 and / or T2 The derivation of the image, the synthesis of the high-resolution mammogram, and / or the derivation of the tissue classification image and / or the calcification image are additionally based on the BO image derived for different motion states.

11. The method according to claim 1 or 2, wherein, A k-space weighted image contrast (KWIC) filter was used to reconstruct the single echo image.

12. An MR device, comprising: At least one main magnet coil (2) is used to generate a uniform and stable magnetic field B0 within the inspection volume; A plurality of gradient coils (4, 5, 6) for generating switched magnetic field gradients in different spatial directions within the examination volume; at least one RF coil (9) for generating RF pulses within the examination volume and / or for receiving MR signals from an object (10) positioned within the examination volume; a control unit (15) for controlling the temporal succession of the RF pulses and the switched magnetic field gradients; and a reconstruction unit (17) for reconstructing an MR image based on the received MR signals, wherein the MR device (1) is arranged to perform the method according to any one of claims 1-11.

13. A computer program to run on an MR device, the computer program comprising instructions for performing the method according to any one of claims 1-11.

14. A method for obtaining high-resolution water, fat, B0 and / or T2. The graph method includes the following steps: Receive multiple echo signals generated at different echo times (TE1, TE2, TE3) after each RF excitation pulse of an imaging sequence performed on an object, wherein the echo signals are acquired along a set of radial or helical k-space trajectories covering a given k-space region, wherein each of the echo signals generated after the RF excitation pulse is assigned a different orientation of the radial or helical trajectory in k-space. The single echo image for each echo time (TE1, TE2, TE3) is reconstructed based on the acquired echo signals, wherein a high-resolution intermediate image is reconstructed based on a combination of the echo signals acquired at the different echo times (TE1, TE2, TE3) from the entire k-space region covered by the set of radial or spiral k-space trajectories. High-resolution water, fat, B0, and / or T2 values ​​are derived from both the single echo image and the high-resolution intermediate image. picture.

Citation Information

Patent Citations

  • Dixon mr imaging using a multi-gradient-echo sequence

    EP3531154A1

  • Dual-echo dixon-type water / fat separation mr imaging

    CN110720047A

  • Dixon mr imaging using a multi-gradient-echo sequence

    CN111758041A