A method, system, and storage medium for magnetic resonance quantification of endometrial peristalsis

Abstract

This invention discloses a method, system, and storage medium for quantifying endometrial peristalsis using magnetic resonance imaging (MRI). The method includes: acquiring multiple frames of continuously acquired MRI dynamic images of the uterine region on the same two-dimensional imaging plane, along with their temporal sequence information; mapping the temporal information of each frame to a virtual third coordinate, establishing a mapping relationship between the virtual third coordinate and the actual acquisition time, and generating derived data capable of multi-planar reconstruction; determining a sampling cutoff line passing through the target anatomical site of the uterus to be observed on a reference image, and determining the spatiotemporal reconstruction plane by the sampling cutoff line and the direction of the virtual third coordinate; obtaining a two-dimensional static spatiotemporal section image through multi-planar reconstruction or equivalent resampling; determining the peristaltic waveform in the image and outputting quantization parameters. This invention transforms dynamic cinematic information into a repeatable and verifiable static image representation, reducing the subjective differences in dynamic observation by the naked eye.

Description

Technical Field

[0001] This invention relates to the field of magnetic resonance medical image processing and dynamic medical image quantitative analysis technology, specifically to a magnetic resonance quantification method, system, and storage medium for endometrial peristalsis. Background Technology

[0002] Endometrial peristalsis refers to the slow, periodic, wave-like movements of the endometrium and its adjacent myometrium. Its direction, frequency, and rhythm are closely related to processes such as embryo implantation and menstrual physiology. Therefore, objectively quantifying endometrial peristalsis has significant clinical value. Currently, in clinical practice, dynamic magnetic resonance imaging (MRI) is commonly used to continuously acquire images of the uterine region on the same two-dimensional imaging plane, forming a dynamic cine sequence. Radiologists then repeatedly play this cine sequence, relying on visual judgment to determine the presence, frequency, direction, and regularity of peristalsis.

[0003] However, this method of interpreting dynamic films by observing them with the naked eye has the most significant technical problem: the peristaltic information in the dynamic films is fleeting, and doctors cannot see the complete waveform evolution throughout the entire observation period in the same frame. They can only estimate the start and end, quantity, spacing, and changes at different time points of the waveform based on memory and subjective impressions. This results in the observation results being highly dependent on the observer's subjective experience, with poor repeatability between different observers and between different interpretations by the same observer. Furthermore, the dynamic observation results are difficult to save as static quantitative records that can be repeatedly measured and verified, thus making it impossible to achieve objective and repeatable quantification of endometrial peristalsis.

[0004] Although conventional medical imaging software has the ability to perform multiplanar reconstruction (MPR) on three-dimensional volume data, the third dimension of a continuous two-dimensional dynamic sequence is essentially time rather than space, and existing software cannot directly convert dynamic time series into static measurable spatiotemporal representations. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, and storage medium for quantifying endometrial peristalsis using magnetic resonance imaging, in order to solve the technical problems in the prior art that rely on visual observation of dynamic magnetic resonance films to interpret endometrial peristalsis, resulting in large subjective differences, inability to objectively quantify, and difficulty in saving and verifying dynamic information.

[0006] To achieve the above objectives, the first aspect of the present invention provides a method for quantifying endometrial peristalsis using magnetic resonance imaging, comprising the following steps: Acquire multiple frames of dynamic magnetic resonance images of the uterine region continuously acquired on the same two-dimensional imaging plane, as well as acquisition time information or frame number used to determine the temporal order of each frame; The multi-frame magnetic resonance dynamic images are sorted according to the acquisition time information or frame number, the time information of each frame is mapped to a virtual third coordinate, and the two-dimensional pixel array of each frame is set as the data layer of virtual three-dimensional image data according to the virtual third coordinate; at the same time, the mapping relationship between the virtual third coordinate and the corresponding real acquisition time is established to generate derivative data that can be reconstructed into multiple planes. Based on the target anatomical site of the uterus to be observed, a sampling cut-off line passing through the target anatomical site is determined on the reference image of the two-dimensional imaging plane, and the spatiotemporal reconstruction plane is determined by the sampling cut-off line and the virtual third coordinate direction. Based on the spatiotemporal reconstruction plane, the derived body data is reconstructed in multiple planes or equivalently resampled to obtain a two-dimensional static spatiotemporal cross-section. One coordinate corresponds to the actual acquisition time through the mapping relationship, and the other coordinate corresponds to the anatomical spatial position on the sampling section. The endometrial peristaltic waveform is determined in the two-dimensional static spatiotemporal cross-section, and at least one of the following quantitative parameters is output: number of peristalsis, peristaltic frequency, propagation direction, spacing between adjacent waves, rhythmicity, and waveform height or peristaltic amplitude.

[0007] Optionally, the virtual third coordinate satisfies z_i=z_0+(i-1)·Δz, and stores the correspondence between the virtual interlayer spacing Δz and the actual frame interval Δt; or establishes a one-to-one correspondence table between the virtual third coordinate z_i and the real acquisition time t_i based on the actual acquisition time of each frame, thereby converting the time axis of the two-dimensional static spatiotemporal cross-section from virtual coordinates to real time; Where z_0 represents the virtual coordinate reference value corresponding to the first frame of the magnetic resonance dynamic image, i represents the i-th frame of the magnetic resonance dynamic image, and Δz represents the virtual interlayer spacing.

[0008] Optionally, the magnetic resonance dynamic image is obtained by acquiring a single-excitation fast spin echo dynamic sequence using a sagittal half-Fourier acquisition; the real time corresponding to each frame is determined based on the DICOM acquisition time information and / or frame sequence information, and a one-to-one correspondence is established with the virtual third coordinate.

[0009] Optionally, the derived data is a derived DICOM sequence; a corresponding derived DICOM instance is generated for each time frame, the SliceLocation and / or ImagePositionPatient are updated according to the virtual third coordinate, and an ImageOrientationPatient adapted to the virtual 3D image data is set so that software with 3D multi-plane reconstruction function can read each time frame according to the virtual third coordinate direction; the derived DICOM sequence has a new SeriesInstanceUID, and each derived DICOM instance has a different SOPInstanceUID from each other.

[0010] Optionally, the derived data is MHD, NIfTI, NRRD, or a three-dimensional voxel array capable of recording; the multi-frame magnetic resonance dynamic images are written in chronological order into a three-dimensional array with dimensions [N,H,W] or equivalent arrangements thereof, and the voxel spacing in the virtual third coordinate direction is saved together with the real time calibration relationship.

[0011] Optionally, the sampling cutoff line is manually determined by the operator based on the target anatomical site, which includes the uterine body, fundus, or the endometrial region to be observed; the position and / or angle of the sampling cutoff line can be adjusted and fixedly applied to all time frames in a single spatiotemporal reconstruction.

[0012] Optionally, on the same set of magnetic resonance dynamic images, at least two different target anatomical sites are set along the anatomical sequence from the fundus to the cervix, and sampling cut-off lines are set at each target anatomical site to generate corresponding two-dimensional static spatiotemporal cross-sections; in each two-dimensional static spatiotemporal cross-section, a consistent waveform feature point definition is used to determine the waveform feature time and / or phase belonging to the same endometrial peristaltic event.

[0013] Optionally, the two-dimensional static spatiotemporal cross-section is displayed with the horizontal axis representing time and the vertical axis representing the spatial coordinates on the sampling cut-off line; the spatial direction of the vertical axis is determined by the position, angle and image orientation information of the sampling cut-off line in the original two-dimensional imaging plane.

[0014] Optionally, the number of peristalsis Nw is the number of mutually separated peristaltic waveforms identified in the two-dimensional static spatiotemporal cross-section within a preset observation time Tobs; the peristaltic frequency f is calculated according to f=60×Nw / Tobs, where Tobs is in seconds and f is in times / minute.

[0015] Optionally, the adjacent wave spacing is obtained by measuring the distance between corresponding feature points of adjacent peristaltic waveforms in the time axis direction in the two-dimensional static spatiotemporal cross-section, and converted into adjacent wave time intervals based on the mapping between the virtual third coordinate and real time; the rhythmicity is characterized by the consistency of multiple adjacent wave spacings or adjacent wave time intervals, and the consistency includes at least one of the mean, standard deviation, coefficient of variation or range.

[0016] Optionally, the waveform height or creep amplitude is obtained by measuring the displacement height of the crest of the creep waveform relative to the baseline, adjacent trough, or preset reference position in the direction of the sampling cut-off spatial axis, and converted into length according to the pixel spacing of the sampling cut-off spatial axis or the spatial calibration relationship after multi-plane reconstruction.

[0017] Optionally, the anatomical propagation direction is determined based on the sequence and / or phase difference of the waveform characteristic times of the same endometrial peristaltic event corresponding to the at least two different target anatomical sites; when the waveform characteristic time of the target anatomical site near the fundus is earlier than that of the target anatomical site near the cervix, it is determined to be propagation from the fundus to the cervix; when the waveform characteristic time of the target anatomical site near the cervix is ​​earlier than that of the target anatomical site near the fundus, it is determined to be propagation from the cervix to the fundus; when the time difference between each target anatomical site is lower than a preset time resolution threshold or the sequence of times does not satisfy the unidirectional propagation relationship, at least one of the following markers is output: undetermined direction, synchronous propagation, or complex propagation.

[0018] Optionally, before constructing the virtual three-dimensional image data, at least one of the following is performed on the multi-frame magnetic resonance dynamic images: rigid registration, affine registration, non-rigid registration, region cropping, intensity normalization, or motion artifact correction targeting the uterine region; after processing, the same sampling cutoff line corresponds to the same target anatomical location in each time frame.

[0019] Optionally, the step of determining the endometrial peristaltic waveform in the two-dimensional static spatiotemporal cross-section is completed by manual marking or by automatic or semi-automatic identification using edge detection, ridge tracing, local gradient, dynamic programming, cross-correlation, phase analysis, or a learning model.

[0020] A second aspect of the present invention provides a magnetic resonance imaging (MRI) system for quantifying endometrial peristalsis, used to implement the above-described MRI method for quantifying endometrial peristalsis, comprising: The dynamic image acquisition module is used to acquire multiple frames of continuous magnetic resonance dynamic images of the uterine region on the same two-dimensional imaging plane, as well as acquisition time information or frame number for determining the time sequence of each frame. The time-dimensional stacking module is used to sort the multi-frame magnetic resonance dynamic images according to the acquisition time information or frame number, map the time information of each frame to a virtual third coordinate, and set the two-dimensional pixel array of each frame as the data layer of virtual three-dimensional image data according to the virtual third coordinate. The derived body data generation module is used to establish the mapping relationship between the virtual third coordinate and the corresponding real acquisition time, and to generate derived body data that can be reconstructed into multiple planes. The target site and sampling cutoff line determination module is used to determine the sampling cutoff line passing through the target anatomical site on the reference image based on the target anatomical site of the uterus to be observed, and the spatiotemporal reconstruction plane is jointly determined by the sampling cutoff line and the virtual third coordinate direction. The spatiotemporal reconstruction module is used to perform multi-plane reconstruction or equivalent resampling on the derived body data based on the spatiotemporal reconstruction plane to obtain a two-dimensional static spatiotemporal cross-section, where one coordinate corresponds to the actual acquisition time through the mapping relationship, and the other coordinate corresponds to the anatomical spatial position on the sampling section line. The quantization module is used to determine the endometrial peristaltic waveform in the two-dimensional static spatiotemporal cross-section and output at least one of the following quantization parameters: number of peristalsis, peristaltic frequency, propagation direction, spacing between adjacent waves, rhythmicity, and waveform height or peristaltic amplitude.

[0021] Optionally, the endometrial peristalsis magnetic resonance quantification system is deployed on a magnetic resonance post-processing workstation, a PACS post-processing server, or a standalone medical image processing platform. The interface of the endometrial peristalsis magnetic resonance quantification system simultaneously displays the original dynamic magnetic resonance reference frame, the target anatomical site and sampling cut-off positioning window, a two-dimensional static spatiotemporal section, and the quantification results. After the sampling cut-off is moved or rotated, the corresponding two-dimensional static spatiotemporal section is regenerated and the quantification results are updated. The endometrial peristalsis magnetic resonance quantification system saves the sampling cut-off coordinates, direction, real time mapping of each time frame, and manual or automatic marker points to form a traceable record.

[0022] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor performs the above-described method for quantifying endometrial peristalsis by magnetic resonance imaging.

[0023] A fourth aspect of the present invention provides a computer program product, including a computer program that, when run on a processor, causes the processor to execute the above-described endometrial peristalsis magnetic resonance quantification method.

[0024] Compared with the prior art, the present invention has the following beneficial effects: This invention encodes the temporal dimension of dynamic magnetic resonance images into a virtual third coordinate and stacks them into derived data capable of multi-planar reconstruction. The spatiotemporal reconstruction plane is then determined by the sampling cutoff line passing through the target anatomical site and the direction of the virtual third coordinate. After multi-planar reconstruction or equivalent resampling, the dynamic information within the entire observation time window is compressed into a single two-dimensional static spatiotemporal cross-section. Thus, the peristaltic waveform, which previously required repeated playback of a dynamic film for observation, is transformed into a geometric structure in a static image that can be directly marked, measured, retained, and verified. This allows quantitative parameters such as the number of peristalsis, frequency, adjacent wave spacing, rhythmicity, waveform height, or peristaltic amplitude to be objectively determined on the same static image, significantly reducing subjective differences caused by relying solely on dynamic visual observation. This achieves repeatable and verifiable objective quantification of endometrial peristalsis. Simultaneously, because the mapping relationship between the virtual third coordinate and the actual acquisition time is preserved, the lateral geometric distance in the static image can be accurately recalculated to the real time, ensuring the physical accuracy of the quantification results. Attached Figure Description

[0025] Figure 1 A schematic flowchart of the magnetic resonance quantification method for endometrial peristalsis provided by the present invention; Figure 2 A schematic diagram of the module structure of the endometrial peristalsis magnetic resonance quantification system provided by the present invention; Figure 3 This is a schematic diagram of the sagittal MR reference image of the uterus and the MPR positioning line in an embodiment of the present invention; Figure 4 This is a static spatiotemporal cross-sectional view formed by spatiotemporal stacking and recombination of dynamic sequences in an embodiment of the present invention; Figure 5 This is a schematic diagram of waveform parameter measurement on a static spatiotemporal diagram in an embodiment of the present invention; Figure 6 The left figure shows a comparison of the characteristic time of multiple target sites in the same dynamic MR data in the embodiments of the present invention, as well as the same peristaltic event (right figure). Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of protection of this invention.

[0027] like Figure 1 As shown, this invention provides a method for quantifying endometrial peristalsis using magnetic resonance imaging based on spatiotemporal stacking, comprising the following steps: Step S1: Acquire multiple frames of dynamic magnetic resonance images of the uterine region continuously acquired on the same two-dimensional imaging plane, as well as acquisition time information or frame number used to determine the temporal order of each frame.

[0028] Step S2: Sort the multiple frames of magnetic resonance dynamic images according to the acquisition time information or frame number, map the time information of each frame to a virtual third coordinate, and set the two-dimensional pixel array of each frame as the data layer of virtual three-dimensional image data according to the virtual third coordinate; at the same time, establish the mapping relationship between the virtual third coordinate and the corresponding real acquisition time, and generate the derived body data that can be reconstructed into multiple planes.

[0029] Step S3: Based on the target anatomical site of the uterus to be observed, determine the sampling cut-off line passing through the target anatomical site on the reference image of the two-dimensional imaging plane, and determine the spatiotemporal reconstruction plane by the sampling cut-off line and the virtual third coordinate direction.

[0030] Step S4: Based on the spatiotemporal reconstruction plane, perform multi-plane reconstruction or equivalent resampling on the derived body data to obtain a two-dimensional static spatiotemporal cross-section. One coordinate corresponds to the actual acquisition time through a mapping relationship, and the other coordinate corresponds to the anatomical spatial position on the sampling cut-off line.

[0031] Step S5: Determine the endometrial peristaltic waveform in the two-dimensional static spatiotemporal cross-section, and output at least one of the following quantitative parameters: number of peristalsis, peristaltic frequency, propagation direction, distance between adjacent waves, rhythmicity, and waveform height or peristaltic amplitude.

[0032] Specifically, in step S1, the dynamic magnetic resonance images are acquired continuously on the same two-dimensional imaging plane, meaning all time frames correspond to the same anatomical layer. This ensures that pixel changes at the same spatial location between different frames reflect changes in that anatomical location over time, rather than changes in the layer position. Multiple frames of dynamic magnetic resonance images can be acquired directly from the magnetic resonance imaging device, or read from a PACS system, image workstation, or storage medium. This invention does not limit the method of image acquisition.

[0033] The acquisition time information used to determine the time sequence of each frame can be the acquisition time tag (such as AcquisitionTime, AcquisitionDateTime) in the DICOM file, and the frame sequence number can be InstanceNumber or other fields that can reflect the acquisition order. When using the DICOM format, the acquisition time information recorded in each frame image can be used to verify the correctness of the image sorting; if the acquisition time information is missing, sorting can be done directly based on the frame sequence number. The number N of dynamic images acquired in this step is an integer greater than 1, typically on the order of tens of frames, such as 60 frames, but the present invention is not limited to this.

[0034] In step S2, the temporal information of multiple frames of magnetic resonance dynamic images is mapped to a virtual third coordinate, making the temporally continuous two-dimensional image sequence equivalent to a three-dimensional volume data in terms of data organization. Let the sorted i-th frame image be I_i(x,y), where x and y are the spatial coordinates of the original imaging plane. Then, a virtual three-dimensional image data V is constructed, with each frame image I_i(x,y) serving as a layer of V at the virtual third coordinate z_i. The virtual third coordinate z_i can be encoded using equal-interval coding.

[0035] It should be noted that the virtual third coordinate is not the actual anatomical slice thickness, nor is it equal to the repetition time (TR) of the magnetic resonance sequence, but is merely a time coding scale established for three-dimensional multiplanar reconstruction.

[0036] In step S3, the reference image can be any frame in the dynamic sequence (e.g., the first frame or an intermediate frame) or an average image generated from the dynamic sequence. The target anatomical site of the uterus to be observed is determined by the operator based on clinical needs, such as the uterine body, fundus, or other endometrial regions to be observed. The sampling cutoff line is a straight line or curve passing through the target anatomical site, and its position and angle are adjusted by the operator according to the local anatomical structure, and it is not required to be along a fixed fundus-cervix direction. For a single analysis, the sampling cutoff line remains consistent throughout all time frames, so that the spatial profile extracted from this cutoff line in each time frame corresponds to the same target anatomical location.

[0037] The spatiotemporal reconstruction plane is determined by the sampling intercept and the virtual third coordinate direction: the sampling intercept provides the spatial direction, and the virtual third coordinate direction provides the temporal direction; the plane spanned by the two is the spatiotemporal reconstruction plane P. Let L(u) denote the sampling intercept, and u be the spatial position parameter along the intercept; then the spatiotemporal reconstruction plane P is parameterized by the coordinate pair (u,z).

[0038] In step S4, multi-plane reconstruction (MPR) is performed on the derived data based on the spatiotemporal reconstruction plane P. This involves extracting a two-dimensional cross-section determined by plane P from the virtual 3D data V to obtain the cross-sectional image K(u,z). Then, based on the mapping relationship between z and real time t, K(u,z) is converted into K(u,t), i.e., a two-dimensional static spatiotemporal cross-sectional image. Equivalently, without relying on the general MPR function, the spatial profile G_i(u) = I_i(L(u)) on the sampling intercept line can be directly extracted from each frame image I_i in the program, and the G_i(u) images can be concatenated in parallel into a two-dimensional array according to chronological order. This implementation is mathematically equivalent to MPR.

[0039] In a two-dimensional static spatiotemporal cross-section, one coordinate (e.g., the horizontal axis) corresponds to the actual acquisition time through a mapping relationship, while the other coordinate (e.g., the vertical axis) corresponds to the anatomical spatial position on the sampling section. Therefore, each time position in the image represents a local anatomical section on the same sampling section at a certain moment, and the continuous change of the anatomical structure of the target site over time can be observed from left to right within the entire observation window. When the endometrium or its adjacent structures undergo periodic displacement or propagation over time, continuous or discontinuous waveforms, stripes, edges, or tilted trajectories will form in the static spatiotemporal cross-section, thus transforming the temporal changes in a dynamic film into geometric structures in a single static image.

[0040] In step S5, the endometrial peristaltic waveform is determined on a two-dimensional static spatiotemporal cross-sectional image. This can be achieved by a physician manually marking waveform feature points (such as peaks, troughs, and leading edges), or by a computer algorithm automatically or semi-automatically identifying the waveform. Based on the determined peristaltic waveform, at least one quantitative parameter is output: (1) Number of peristalsis: The number of mutually separated peristaltic waveforms identified within the preset observation time Tobs.

[0041] (2) Peristalsis frequency: It is calculated by converting the number of peristalsis and the observation time, for example, f=60×Nw / Tobs (times / minute).

[0042] (3) Direction of propagation: It can be determined by comparing the characteristic time of the same peristaltic event at multiple target anatomical sites, or by observing the tilt direction of the waveform in a single spatiotemporal graph.

[0043] (4) Interval between adjacent waves: The distance between the corresponding feature points of adjacent peristaltic waveforms in the direction of the time axis, which can be converted into the time interval between adjacent waves.

[0044] (5) Rhythm: Characterized by the consistency of multiple adjacent wave intervals (such as mean, standard deviation, coefficient of variation).

[0045] (6) Waveform height or creep amplitude: The displacement height of the crest of the creep waveform relative to the baseline or trough in the spatial axis direction, which can be converted into length based on the pixel spacing.

[0046] The above-mentioned quantitative parameters can be output individually or in combination; they can be displayed in numerical form or superimposed on a static spatiotemporal cross-section to form a traceable annotation record.

[0047] In one feasible implementation, the virtual third coordinate satisfies z_i = z_0 + (i-1)·Δz, and stores the correspondence between the virtual interlayer spacing Δz and the actual frame interval Δt; or, based on the actual acquisition time of each frame, a one-to-one correspondence table between the virtual third coordinate z_i and the actual acquisition time t_i is established, thereby converting the time axis of the two-dimensional static spatiotemporal cross-section from virtual coordinates to real time. Here, z_0 represents the virtual coordinate reference value corresponding to the first frame of the magnetic resonance dynamic image, i represents the i-th frame of the magnetic resonance dynamic image, and Δz represents the virtual interlayer spacing.

[0048] Specifically, this embodiment provides two methods for establishing the mapping relationship between virtual third coordinates and real time. The first method is a linear conversion method: the virtual third coordinates are encoded at equal intervals Δz, and the correspondence between Δz and the actual frame interval Δt (e.g., 3 seconds / frame) is recorded. Then, the virtual coordinate difference Δz' between any two points in the static image can be converted into the real time difference by (Δz' / Δz)×Δt. This method is simple to implement, has low storage overhead, and is suitable for conventional dynamic sequences acquired at equal intervals.

[0049] The second method is a one-to-one mapping table: This method records the actual acquisition time t_i corresponding to each virtual layer z_i in the form of a lookup table, where t_i is directly derived from the DICOM acquisition time field. This method is suitable for situations where the frame interval is not completely uniform, such as acquisition time jitter caused by reconstruction or heartbeat gating, and can provide more accurate time references. Either method can be used, or they can be used in combination; for example, using a linear relationship as the default and the DICOM time field as the verification.

[0050] Furthermore, if the time axis is interpolated during the multi-plane reconfiguration display process (e.g., resampling to improve display smoothness), the conversion is still based on the mapping between the virtual time coordinates and the original real time t_i, to avoid mistaking the interpolated pixel spacing as the new actual time resolution.

[0051] In one feasible implementation, the magnetic resonance dynamic images are obtained by acquiring a single-excitation fast spin echo (HASTE) dynamic sequence using sagittal half-Fourier acquisition; the real time corresponding to each frame is determined based on the DICOM acquisition time information and / or frame sequence information, and a one-to-one correspondence is established with the virtual third coordinate.

[0052] Specifically, the HASTE dynamic sequence is a single-shot fast spin echo sequence that, combined with half-Fourier acquisition, can acquire one frame of image within a single shot. It offers fast acquisition speed and is relatively insensitive to motion, making it suitable for displaying the boundary between the endometrium and myometrium, thus making it a preferred sequence for acquiring dynamic uterine images. The sagittal imaging plane can simultaneously display the fundus, body, and cervix along the long axis of the uterus, facilitating subsequent setting of target anatomical sites and sampling cutoff lines.

[0053] In a specific scanning protocol, TR=1000 ms, TE=68 ms, flip angle=120°, FOV=200×200 mm², matrix=192×128, slice thickness=5 mm, interslice gap=0 mm, NEX=1, and 60 frames of dynamic images are actually continuously acquired, with a nominal temporal resolution of 3 s / frame and a total acquisition time of 180 s. It should be noted that this method does not directly use TR as the frame interval, but rather establishes a true time axis based on the actual dynamic acquisition timing of 3 s / frame and reading the acquisition time and / or frame sequence information from the DICOM. That is, the MR sequence TR, the effective frame interval of the dynamic image, and the virtual third coordinate interslice gap are recorded separately to avoid confusion. This invention can also be applied to other magnetic field strengths (such as 1.5 T), other coil configurations, and other fast sequences capable of forming continuous dynamic uterine MR images on the same two-dimensional imaging plane, and is not limited to the specific parameters mentioned above.

[0054] In one feasible implementation, the derived data is a derived DICOM sequence; a corresponding derived DICOM instance is generated for each time frame, the SliceLocation and / or ImagePositionPatient are updated according to the virtual third coordinate, and the ImageOrientationPatient is set to adapt to the virtual 3D image data, so that software with 3D multi-plane reconstruction function can read each time frame according to the virtual third coordinate direction; the derived DICOM sequence has a new SeriesInstanceUID, and each derived DICOM instance has a different SOPInstanceUID from each other.

[0055] The purpose of this embodiment is to enable the present invention to directly utilize the existing 3D MPR function of the hospital imaging workstation without developing dedicated reconstruction software. Specifically, a derived DICOM instance is generated for each time frame, and the time sequence is encoded into the geometric position label of the DICOM. The ImagePositionPatient, SliceLocation, and ImageOrientationPatient of the reference frame are read, and the SliceLocation and / or the third component of ImagePositionPatient of each instance are updated frame by frame according to the virtual layer spacing Δz, so that the 60 time frames are identified as 60 "layers" arranged along a third direction in the DICOM geometry; simultaneously, a consistent ImageOrientationPatient is set for each layer to ensure that the normals of each layer are consistent and an ordered virtual 3D body can be formed.

[0056] To ensure standardized data management, derived sequences use a new SeriesInstanceUID to distinguish them from the original dynamic sequences, and each derived instance uses a different SOPInstanceUID to avoid conflicts with the original instance or other derived instances. Through the above processing, any workstation that conforms to the DICOM standard and has 3D MPR functionality (such as a PACS post-processing terminal) can read time frame sequences as three-dimensional volume data and perform arbitrary planar reconstruction, thereby achieving the spatiotemporal reconstruction of this invention at minimal cost.

[0057] In one feasible implementation, the derived data is MHD, NIfTI, NRRD, or a three-dimensional voxel array capable of recording; multiple frames of magnetic resonance dynamic images are written in chronological order into a three-dimensional array with dimensions [N,H,W] or their equivalent arrangement, and the voxel spacing in the virtual third coordinate direction is saved together with the real time calibration relationship.

[0058] This embodiment provides a 3D volumetric data implementation path independent of the DICOM system, suitable for self-developed post-processing software or scientific research analysis platforms. Taking the MHD (MetaImage) format as an example, 2D images sorted by time are written layer by layer into a 3D array of dimensions [N,H,W] (where N is the number of frames, H is the number of rows, and W is the number of columns; an equivalent arrangement such as [H,W,N] is also acceptable). The spacing_x and spacing_y of the original 2D plane retain their real spatial meaning, while the spacing_z of the virtual third coordinate direction is set to 1 or other values ​​convenient for software to read. Simultaneously, the mapping between each virtual layer and the actual acquisition time is recorded in the accompanying header file, database, or program variables. Formats such as NIfTI and NRRD have equivalent 3D voxel storage capabilities and can all be used as carriers of derived volumetric data. This path and the DICOM path of the previous embodiment can coexist or be used selectively, depending on the available software in the deployment environment.

[0059] In one feasible implementation, the sampling cutoff line is manually determined by the operator based on the target anatomical site, which may include the uterine body, fundus, or other endometrial regions to be observed; the position and / or angle of the sampling cutoff line can be adjusted and fixedly applied to all time frames in a single spatiotemporal reconstruction.

[0060] This embodiment differs from the fixed analysis method along the long axis of the uterine fundus-cervix. Physicians can freely select the target anatomical site based on clinical focus: when evaluating peristalsis in the uterine body, the cutoff line is set in the uterine body; when evaluating peristalsis in the uterine fundus, the cutoff line is moved to the uterine fundus. The angle of the cutoff line can also be adjusted according to the local endometrial course to ensure that the direction of the cutoff line is as consistent as possible with the direction of the peristaltic displacement to be observed, thus obtaining the most significant waveform representation. The interactive interface provides controls for dragging, rotating, and adjusting the length of the cutoff line; after the cutoff line is adjusted, the system immediately regenerates the spatiotemporal cross-section. In a single analysis, the determined cutoff line is fixedly applied to all time frames, meaning that spatial profiles are extracted from the same cutoff line in each time frame, which is a prerequisite for waveform comparability; if image registration is used, the cutoff line remains consistent in the registered unified reference coordinates.

[0061] In one feasible implementation, at least two different target anatomical sites are set along the anatomical sequence from the fundus to the cervix on the same set of magnetic resonance dynamic images, and sampling cut-off lines are set at each target anatomical site to generate corresponding two-dimensional static spatiotemporal cross-sections; in each two-dimensional static spatiotemporal cross-section, a consistent waveform feature point definition is used to determine the waveform feature time and / or phase belonging to the same endometrial peristaltic event.

[0062] This embodiment provides a data foundation for determining the direction of peristaltic propagation. On the same set of dynamic data, for example, three target anatomical sites are set: the fundus site F, the uterine body site B, and the cervical site C (e.g., ...). Figure 6 As shown in the previous embodiment, sampling cutoff lines are set and corresponding static spatiotemporal cross-sections K_F, K_B, and K_C are generated. Since the three spatiotemporal cross-sections originate from the same acquisition and their time axes are consistent, waveforms at different points can be directly compared over time.

[0063] The key lies in the correspondence between "consistent waveform feature point definitions" and "the same peristaltic event": for peristaltic events to be compared, the same feature point definitions are used in each spatiotemporal diagram, such as peaks, troughs, waveform leading edges, maximum gradient positions, or other repeatable feature points, and their corresponding real times t_F, t_B, and t_C are recorded. When the peristaltic wave propagates along the fundus-cervix direction, the waveforms formed by the same event at the three sites have a systematic temporal sequence, which can be used to determine the propagation direction. In automated implementation, cross-correlation analysis can also be performed on the spatiotemporal diagrams of different sites or the time signals extracted from them to estimate the time delay or phase difference between sites.

[0064] In one feasible implementation, the two-dimensional static spatiotemporal cross-section is displayed with the horizontal axis representing time and the vertical axis representing the spatial coordinates on the sampling cut-off line; the spatial direction of the vertical axis is determined by the position, angle and image orientation information of the sampling cut-off line in the original two-dimensional imaging plane.

[0065] This embodiment defines the display coordinate system for static spatiotemporal cross-sectional images. The horizontal axis is the time axis, and its physical scale is determined by the mapping relationship between virtual coordinates and real time. The vertical axis is the spatial coordinate on the sampling cutoff line, and its direction and physical scale are determined by the position, angle, and image orientation information (such as DICOM's PixelSpacing and ImageOrientationPatient) of the cutoff line in the original two-dimensional imaging plane. Therefore, the anatomical meaning of the vertical axis depends on the cutoff line position and angle set by the operator: when the cutoff line is set along the fundus-cervix direction, the vertical axis approximately corresponds to the fundus-cervix direction; when the cutoff line is set in the direction of the endometrial lining, the vertical axis corresponds to the local endometrial direction. This invention does not preset the vertical axis to a fixed fundus-cervix or cervix-fundus direction. This flexibility allows the invention to observe the peristaltic behavior of any local area. The displacement on the spatial axis can be converted to millimeters based on the pixel pitch, providing a physical calibration for the quantitative determination of peristaltic amplitude.

[0066] In one feasible implementation, the number of peristaltic movements Nw is the number of mutually separated peristaltic waveforms identified in the two-dimensional static spatiotemporal cross-section within the preset observation time Tobs; the peristaltic frequency f is calculated according to f=60×Nw / Tobs, where Tobs is in seconds and f is in times / minute.

[0067] This embodiment defines the most basic creep quantization parameters. The preset observation time Tobs is usually equal to the total acquisition time of the dynamic sequence, for example, 180 s; in a static spatiotemporal cross-section (such as... Figure 5 As shown in the figure, the peristaltic waveform appears as a wavy structure that repeats along the time axis. The number of mutually separated waveforms, counted manually or by algorithm, is Nw. The peristaltic frequency is converted to times per minute using f = 60 × Nw / Tobs. For example, when Tobs = 180 s, f = Nw / 3 times per minute. To improve the repeatability of the count, the counting area, the minimum continuous length of the waveform, and the passing rules for waveform feature points or the same spatial reference position can be predetermined. These counting rules can be set manually, fixed by software, or have two observers count independently and obtain consistent results to further reduce subjectivity.

[0068] In one feasible implementation, the adjacent wave spacing is obtained by measuring the distance between corresponding feature points of adjacent creeping waveforms in the time axis direction in a two-dimensional static spatiotemporal cross-section diagram, and is converted into adjacent wave time intervals based on the mapping between the virtual third coordinate and real time; the rhythmicity is characterized by the consistency of multiple adjacent wave spacings or adjacent wave time intervals, and the consistency includes at least one of the mean, standard deviation, coefficient of variation or range.

[0069] In this embodiment, defined at the same spatial reference position or by the same feature points, the spacing d_j between two adjacent peristaltic waveforms on the time axis is measured and converted into a time interval ΔT_j. For example, when 60 frames are calibrated at 3 s / frame, if the adjacent waves differ by m original frame intervals, ΔT_j=3m / second. Rhythmicity is characterized by the consistency of a plurality of ΔT_j: the average value reflects the average peristaltic cycle, and the standard deviation, range and coefficient of variation reflect the dispersion degree of the intervals. The more consistent the spacing between adjacent waves (the smaller the coefficient of variation), the more regular the peristaltic rhythm; cases with rhythm disorders show higher interval dispersion. This parameter provides an objective indicator for clinical evaluation of peristaltic regularity.

[0070] In a feasible embodiment, the waveform height or peristaltic amplitude is obtained by measuring the displacement height of the peak of the peristaltic waveform in the spatial axis direction of the sampling intercept relative to the baseline, adjacent trough or preset reference position, and converted into a length according to the pixel spacing of the sampling intercept spatial axis or the spatial calibration relationship after multi-planar reconstruction.

[0071] In this embodiment, on the static spatiotemporal切面 map with the spatial axis as the longitudinal direction, the vertical height h_j of the peak of a single peristaltic waveform relative to the local baseline, adjacent trough or preset reference position is measured. h_j can be directly recorded in units of displayed pixels, or can be converted into millimeters using the spatial calibration of MPR images or the original PixelSpacing combined with the actual geometry of the sampling intercept direction to form the peristaltic displacement amplitude A_j. It should be noted that when the sampling intercept is not parallel to the grid direction of the original imaging plane, the actual pixel spacing in the intercept direction needs to be corrected according to the intercept angle. When comparing different cases or different sites, it is preferable to adopt the same waveform height definition and the same spatial calibration method to ensure measurement consistency.

[0072] In a feasible embodiment, the anatomical propagation direction is determined according to the sequence and / or phase difference of the waveform characteristic times corresponding to the same endometrial peristalsis event at at least two different target anatomical sites; when the waveform characteristic time of the target anatomical site close to the uterine fundus side is earlier than that of the target anatomical site close to the cervical side, the propagation is determined to be from the uterine fundus side to the cervical side; when the waveform characteristic time of the target anatomical site close to the cervical side is earlier than that of the target anatomical site close to the uterine fundus side, the propagation is determined to be from the cervical side to the uterine fundus side; when the time difference between the target anatomical sites is lower than the preset time resolution threshold or the time sequence does not satisfy the unidirectional propagation relationship, at least one of the marks of undetermined direction, synchronous propagation or complex propagation is output.

[0073] Specifically, taking three sites as an example, if the same peristalsis event satisfies t_F<t_B<t_C (such as Figure 6as shown), which indicates that the event first appears on the fundus uteri side, then reaches the uterine body and subsequently the cervical side, and is determined to be propagation from the fundus uteri side to the cervical side; if t_C<t_B<t_F is satisfied, it is determined to be propagation from the cervical side to the fundus uteri side; when only two sites are provided, the judgment is made according to the sequence of the two. The determination is based on the chronological relationship between different sites, rather than interpreting the vertical axis of a single local intercept line as the fundus uteri-cervix direction, and this difference ensures the anatomical reliability of direction determination.

[0074] Considering the limitation of temporal resolution, a preset temporal resolution threshold is introduced: in the embodiment of 3 s / frame, manual interpretation may preferably take at least 1 original frame interval (3 s) as the resolvable threshold; if |Δt| between any two sites is less than the threshold, it is marked as undetermined direction or nearly synchronous; if the chronological order does not satisfy a monotonic unidirectional relationship (for example, presenting a mixed sequence), it is marked as complex propagation. In addition, the arrival time difference Δt_AB=t_B-t_A can be defined and expressed in seconds, and if the anatomical distance between sites can be obtained from a reference image, the apparent propagation velocity can also be calculated by the ratio of the distance to the time delay.

[0075] In a feasible embodiment, before constructing virtual three-dimensional image data, at least one of rigid registration, affine registration, non-rigid registration, region cropping, intensity normalization or motion artifact correction targeting the uterine region is performed on multiple frames of magnetic resonance dynamic images; after processing, the same sampling intercept line corresponds to the same target anatomical position in each temporal frame.

[0076] If there is respiration, overall uterine displacement or slight patient movement during dynamic scanning, the anatomical correspondence between frames will be affected. In this embodiment, preprocessing is performed on the image sequence before stacking: rigid registration with the uterine region as the region of interest can eliminate overall translation and rotation, affine registration can further correct scale and shear changes, and non-rigid registration can correct local deformation; region cropping can reduce data volume and highlight the target region, intensity normalization can eliminate inter-frame brightness inconsistency caused by receiving gain fluctuation, and motion artifact correction can improve single-frame image quality. After any one or a combination of the above processing, the same sampling intercept line corresponds to the same target anatomical position in each frame, thereby improving the spatial consistency of the waveform in the spatiotemporal切面 map. This preprocessing is an optional step provided for enhancing robustness, and the present invention can also be implemented without performing the preprocessing.

[0077] In a feasible embodiment, the step of determining the endometrial peristalsis waveform in the two-dimensional static spatiotemporal切面 map is completed by manual marking, or by automatic or semi-automatic recognition through edge detection, ridge tracing, local gradient, dynamic programming, cross-correlation, phase analysis or a learning-based model.

[0078] The simplest method, consistent with current clinical practice, is manual identification: physicians manually count and measure waveform features on a static spatiotemporal cross-section and manually mark the characteristic times of the same peristaltic event in multi-target site analysis. Automatic or semi-automatic identification methods include: extracting waveform boundaries using edge detection; extracting ridges along the waveform direction using ridge tracing; locating the leading edge or maximum change position of the waveform using local gradients; finding the optimal waveform path in the time-space two-dimensional graph using dynamic programming; analyzing the time signal extracted from the spatiotemporal graph using cross-correlation or phase analysis to determine the delay between waveform events and sites; and directly outputting waveform positions and feature points using a trained deep learning model (such as a segmentation network or trajectory regression network). Automatic identification results can be overlaid for physician confirmation and modification, forming a semi-automatic workflow. All of the above identification methods are optional enhancements. Regardless of the method used, the core data processing chain of this invention—"stacked time frames as a virtual third coordinate + target site sampling cutoff line + spatiotemporal reconstruction to form a static graph"—remains unchanged.

[0079] like Figure 2 As shown, the present invention provides an endometrial peristalsis magnetic resonance quantification system for implementing the above-mentioned endometrial peristalsis magnetic resonance quantification method, comprising: The dynamic image acquisition module 1 is used to acquire multiple frames of magnetic resonance dynamic images of the uterine region continuously acquired on the same two-dimensional imaging plane, as well as acquisition time information or frame number used to determine the time sequence of each frame.

[0080] The time-dimensional stacking module 2 is used to sort multiple frames of magnetic resonance dynamic images according to the acquisition time information or frame number, map the time information of each frame to a virtual third coordinate, and set the two-dimensional pixel array of each frame as the data layer of virtual three-dimensional image data according to the virtual third coordinate.

[0081] The derived body data generation module 3 is used to establish the mapping relationship between the virtual third coordinate and the corresponding real acquisition time, and to generate derived body data that can be reconstructed into multiple planes.

[0082] The target site and sampling cutoff line determination module 4 is used to determine the sampling cutoff line passing through the target anatomical site on the reference image based on the target anatomical site of the uterus to be observed, and the spatiotemporal reconstruction plane is determined by the sampling cutoff line and the virtual third coordinate direction.

[0083] The spatiotemporal reconstruction module 5 is used to perform multi-plane reconstruction or equivalent resampling on the derived body data based on the spatiotemporal reconstruction plane to obtain a two-dimensional static spatiotemporal cross-section. One coordinate corresponds to the actual acquisition time through a mapping relationship, and the other coordinate corresponds to the anatomical spatial position on the sampling cut-off line.

[0084] Quantization module 6 is used to determine the endometrial peristaltic waveform in a two-dimensional static spatiotemporal cross-sectional image and output at least one of the following quantization parameters: number of peristalsis, peristaltic frequency, propagation direction, spacing between adjacent waves, rhythmicity, and waveform height or peristaltic amplitude.

[0085] In one feasible implementation, the system is deployed on a magnetic resonance post-processing workstation, a PACS post-processing server, or a standalone medical image processing platform. The system's interface simultaneously displays the original dynamic magnetic resonance reference frame, the target anatomical site and sampling cut-off positioning window, a two-dimensional static spatiotemporal section, and quantization results. After the sampling cut-off is moved or rotated, the corresponding two-dimensional static spatiotemporal section is regenerated and the quantization results are updated. The system saves the sampling cut-off coordinates, orientation, real time mapping of each time frame, and manually or automatically marked points to form a traceable record.

[0086] This embodiment defines the deployment form and interactive interface of the system. In terms of deployment, the system can be used as a software plug-in to the post-processing workstation of an MR device, as a post-processing component of a PACS server providing services via network, or as independent research analysis software running on a regular computer. In terms of interface design, multi-view linkage is key: the positioning window displays a reference frame and interactive sampling cut-off lines. After the physician drags or rotates the cut-off lines, the system instantly reconstructs the spatiotemporal data and updates the spatiotemporal cross-sectional image and quantitative results, achieving a "what you see is what you get" interactive analysis. Regarding traceability, the system saves the cut-off line coordinates and direction, the time mapping table, and all manually or automatically marked points along with the analysis results, ensuring that any quantitative result can be reviewed, audited, and reproduced afterward, meeting clinical quality control requirements.

[0087] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor performs the above-described method for quantifying endometrial peristalsis by magnetic resonance imaging.

[0088] Specifically, the aforementioned computer-readable storage medium can be a non-volatile storage medium that can be read by a processor, including but not limited to: disks, optical discs (CD-ROM, DVD), USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, etc. The computer program stored on the storage medium contains instruction code that implements each step of the above method. When the processor executes the instructions, it completes the entire processing flow from dynamic image acquisition, time-dimensional virtual coordinate stacking, derived volume data generation, sampling cutoff location, spatiotemporal reconstruction to waveform quantization output. The processor can be a general-purpose CPU, GPU, processing unit of a medical imaging workstation, or an embedded processor; this invention is not limited to any particular type.

[0089] The present invention also provides a computer program product, including a computer program that, when run on a processor, causes the processor to execute the above-described method for quantifying endometrial peristalsis by magnetic resonance imaging.

[0090] Specifically, the computer program in the aforementioned computer program product can be distributed in software form, such as a software package downloaded and installed via the network, a program module called by a processing service deployed in the cloud, or a software component released together with the medical image post-processing system. When the program runs on the processor, it executes the above method to convert the input uterine magnetic resonance dynamic sequence into a two-dimensional static spatiotemporal cross-sectional image and output peristaltic quantification parameters. The computer program product may also include auxiliary programs for guiding installation, configuring imaging parameter interfaces, or interfacing with a PACS system; these are all equivalent embodiments within the scope of this invention.

[0091] Example 1 This embodiment recruits one female subject of childbearing age to undergo examination using a 16-channel phased array pelvic coil on a 3.0 T clinical MR system (such as MAGNETOM Prisma, Siemens Healthineers). The subject fasted for 4 hours and emptied their bladder before the scan to reduce motion artifacts. First, sagittal T2WI (TR=3910 ms, TE=80 ms, FOV=240×240 mm², slice thickness=5 mm) and axial T2WI (TR=5360 ms, TE=91 ms, FOV=300×300 mm², slice thickness=5 mm) were acquired to locate and visualize the endometrium, conjunctival zone, and myometrium. Subsequently, 60 frames of images were continuously acquired in the same sagittal plane using the sagittal HASTE dynamic sequence (TR=1000 ms, TE=68 ms, flip angle=120°, FOV=200×200 mm², matrix=192×128, slice thickness=5 mm, interslice=0 mm, NEX=1), with a nominal temporal resolution of 3 s / frame and a total acquisition time of 180 s.

[0092] During processing, the 60 frames of images are sorted according to the DICOM acquisition time and InstanceNumber. A virtual three-dimensional array with dimensions [60,H,W] is constructed with Δz=1 virtual unit and written into the MHD format derived data. Simultaneously, a time calibration table of "3 seconds per virtual layer" is saved. The operator sets a sampling cutoff line for the uterine body on the reference frame, adjusting the cutoff line direction along the local endometrial line. An example of the sagittal MR reference image and MPR positioning line for the uterus in this embodiment is shown below. Figure 3 As shown, the static spatiotemporal cross-section diagram formed by spatiotemporal stacking and recombination of the dynamic sequence in this embodiment is as follows: Figure 4 As shown.

[0093] A spatiotemporal recombination plane is determined by the intercept line and the virtual third coordinate, and multi-plane resampling is performed to obtain a static spatiotemporal sectional view with the horizontal axis representing time (0-180 s) and the vertical axis representing the spatial coordinate of the intercept line. Clear reciprocating peristaltic ripples can be seen on the diagram. Manual counting gives Nw=9 waveforms, so the peristaltic frequency f=60×9 / 180=3 times per minute; the time intervals between adjacent peaks are measured as 18, 21, 21, 18, 21, 24, 18, 21 s respectively, with an average value of 20.3 s, a standard deviation of 2.0 s, and a coefficient of variation of 9.8%, indicating a relatively regular rhythm; the spatial displacement of the wave peak relative to the adjacent wave trough is converted to approximately 2.5 mm through the pixel pitch, which is the peristaltic amplitude.

[0094] Further, three sampling intercept lines, namely the fundus uteri point F, the uterine body point B, and the cervical point C, are arranged from the fundus uteri side to the cervical side, and three spatiotemporal diagrams are generated. For the same peristaltic event, t_F=36 s, t_B=39 s, and t_C=42 s are recorded according to the characteristic points of the wave peaks, which satisfy t_F<t_B<t_C and the adjacent differences all reach the resolution threshold of one frame interval (3 s), so it is determined that the peristalsis propagates from the fundus uteri side to the cervical side.

[0095] Comparative Example 1 For the same set of 60 frames of dynamic data of the same subject in Example 1, without performing spatiotemporal stacking processing, two radiologists with experience in pelvic MR diagnosis respectively interpreted the data by repeatedly playing the dynamic movie in a conventional manner, and each estimated the number, frequency, rhythm and direction of peristalsis without using any static spatiotemporal characterization tool.

[0096] A comparison between Example 1 and Comparative Example 1 shows that: (1) In terms of objectivity and repeatability, the number, frequency, wave interval and amplitude of peristalsis in Example 1 are all directly marked and measured on a single static diagram, and the measurement results have clear geometric coordinates and time back-marking, which can be reviewed by a third party according to the same marking rules; while in the comparative example, the two physicians estimated the number of peristalsis to be 8 times and 10 times respectively (corresponding to frequencies of 2.7 and 3.3 times per minute), and the interpretation results of the same physician on the next day also fluctuated, indicating that there are obvious subjective differences in naked-eye dynamic interpretation.

[0097] (2) In terms of information completeness, after 180 s of dynamic information is compressed into a single static diagram, the number, spacing, inclination direction and height of the waveforms are clear at a glance, which avoids the problem that the information is fleeting during dynamic playback and relies on memory for estimation.

[0098] (3) In terms of quantification capability, it is difficult for naked-eye interpretation to obtain the coefficient of variation of adjacent wave intervals (rhythmicity) and the peristaltic amplitude in millimeters, while the present invention can directly provide these objective parameters based on time mapping and spatial calibration.

[0099] (4) Regarding the determination of the direction of propagation, the direction of visual judgment relies on the doctor's memory of the movie screen at different times, which is unreliable. This invention directly compares the characteristic time of the same event on the static map of multiple target sites, and gives a clear direction conclusion within the time resolution range. It also outputs the "direction undetermined / synchronous / complex propagation" mark for cases that do not meet the judgment conditions, thus avoiding forcibly giving unreliable conclusions.

[0100] Example 2 In another embodiment, a derived DICOM path is employed: Derived DICOM instances are generated frame-by-frame for the 60 frames of images from Example 1. The ImagePositionPatient, SliceLocation, and ImageOrientationPatient of the reference frame are read, and the SliceLocation is updated frame-by-frame at Δz=1. New SeriesInstanceUIDs are assigned to the derived sequences, and different SOPInstanceUIDs are assigned to each instance before sending the data to the PACS. The derived sequence is directly read using the 3D MPR function on a general-purpose 3D post-processing workstation in the hospital. The reconstruction plane is adjusted on the MPR positioning interface to be parallel to the virtual time axis and pass through the target cutoff line, thus obtaining a static spatiotemporal section equivalent to that of Example 1. Counting and measurement are then performed on the workstation's measurement tools. This embodiment demonstrates that the present invention can be implemented in existing clinical workflows without the need for dedicated software.

[0101] Compare with Example 2 This comparative example simply stacks 60 frames of images and interprets the third dimension according to the spatial layer spacing, without establishing a mapping between the virtual third coordinate and the actual acquisition time (i.e., without time scaling). Therefore, the horizontal axis of the reconstructed cross-section only has layer order significance. Although the operator can count the number of waveforms, they cannot accurately convert the creeping frequency to the number of seconds between adjacent waveforms. Especially when the MPR display interpolates the time axis, the correspondence between pixel spacing and actual time is further disrupted, leading to a systematic deviation in frequency calculation. This comparative example illustrates that "establishing a mapping relationship between the virtual third coordinate and the actual acquisition time" is a necessary technical feature for achieving accurate time quantization in this invention.

[0102] Compare with Example 3 This comparative example does not employ a target anatomical site sampling cutoff mechanism. Instead, it uses a fixed cutoff line along the long axis of the uterine fundus to the cervix for observation. When peristaltic events mainly occur locally in the uterine body and have small amplitudes, their waveforms may be submerged or superimposed by changes in other structures along the long axis. Furthermore, a single cutoff line cannot distinguish the temporal sequence of events at different sites, making it impossible to determine the direction of propagation. This invention overcomes this deficiency through a sampling cutoff mechanism that can be flexibly set according to target anatomical sites and allows for multiple cutoff lines to run in parallel.

[0103] The specific parameters of the 3.0 T device, 16-channel coil, and T2WI / HASTE described above are only used to illustrate a repeatable acquisition scheme. This invention can also be applied to other magnetic field strengths, coil configurations, and fast sequences capable of forming continuous dynamic uterine MR images on the same two-dimensional imaging plane; its core lies in the scalable virtual spatial encoding of the time dimension and the generation of static spatiotemporal cross-sections based on the sampling cut-off lines of the target sites.

[0104] Without altering the core concept of this invention, the numerical scale of the virtual third coordinate, the derived data format, the MPR interpolation method, the sampling intercept position or angle, the waveform recognition algorithm, and the quantization parameter combination can all be equivalently replaced. To avoid unnecessarily limiting the scope of protection to a particular software, the MPR in this specification also includes an equivalent implementation of directly resampling a static spatiotemporal cross-section from the time series based on the same spatiotemporal reconstruction plane.

[0105] The present invention outputs static spatiotemporal representations and quantitative intermediate parameters formed by medical image processing, which are used by clinicians to interpret in conjunction with other medical information; the method itself does not require the output of disease diagnosis conclusions or treatment plans as a necessary step.

[0106] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for quantifying endometrial peristalsis using magnetic resonance imaging, characterized in that, Includes the following steps: Acquire multiple frames of dynamic magnetic resonance images of the uterine region continuously acquired on the same two-dimensional imaging plane, as well as acquisition time information or frame number used to determine the temporal order of each frame. The multi-frame magnetic resonance dynamic images are sorted according to the acquisition time information or frame number, the time information of each frame is mapped to a virtual third coordinate, and the two-dimensional pixel array of each frame is set as the data layer of virtual three-dimensional image data according to the virtual third coordinate; at the same time, the mapping relationship between the virtual third coordinate and the corresponding real acquisition time is established to generate derivative data that can be reconstructed into multiple planes. Based on the target anatomical site of the uterus to be observed, a sampling cut-off line passing through the target anatomical site is determined on the reference image of the two-dimensional imaging plane, and the spatiotemporal reconstruction plane is determined by the sampling cut-off line and the virtual third coordinate direction. Based on the spatiotemporal reconstruction plane, the derived body data is reconstructed in multiple planes or equivalently resampled to obtain a two-dimensional static spatiotemporal cross-section. One coordinate corresponds to the actual acquisition time through the mapping relationship, and the other coordinate corresponds to the anatomical spatial position on the sampling section. The endometrial peristaltic waveform is determined in the two-dimensional static spatiotemporal cross-section, and at least one of the following quantitative parameters is output: number of peristalsis, peristaltic frequency, propagation direction, spacing between adjacent waves, rhythmicity, and waveform height or peristaltic amplitude.

2. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 1, characterized in that, The virtual third coordinate satisfies z_i=z_0+(i-1)·Δz, and stores the correspondence between the virtual interlayer spacing Δz and the actual frame interval Δt; Alternatively, a one-to-one correspondence table between the virtual third coordinate z_i and the real acquisition time t_i can be established based on the actual acquisition time of each frame, thereby converting the time axis of the two-dimensional static spatiotemporal cross-section from virtual coordinates to real time. Where z_0 represents the virtual coordinate reference value corresponding to the first frame of the magnetic resonance dynamic image, i represents the i-th frame of the magnetic resonance dynamic image, and Δz represents the virtual interlayer spacing.

3. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 1, characterized in that, The magnetic resonance dynamic images are obtained by acquiring a single-excitation fast spin echo dynamic sequence using sagittal half-Fourier acquisition; the real time corresponding to each frame is determined based on the DICOM acquisition time information and / or frame sequence information, and a one-to-one correspondence is established with the virtual third coordinate.

4. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 1, characterized in that, The derived data is a derived DICOM sequence; a corresponding derived DICOM instance is generated for each time frame, the SliceLocation and / or ImagePositionPatient are updated according to the virtual third coordinate, and an ImageOrientationPatient adapted to the virtual 3D image data is set so that software with 3D multi-plane reconstruction function can read each time frame according to the virtual third coordinate direction; the derived DICOM sequence has a new SeriesInstanceUID, and each derived DICOM instance has a different SOPInstanceUID.

5. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 1, characterized in that, The derived data is MHD, NIfTI, NRRD, or a three-dimensional voxel array capable of recording; the multi-frame magnetic resonance dynamic images are written in chronological order into a three-dimensional array with dimensions [N,H,W] or equivalent arrangements thereof, and the voxel spacing in the virtual third coordinate direction is saved together with the real time calibration relationship.

6. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 1, characterized in that, The sampling cutoff line is determined manually by the operator based on the target anatomical site, which includes the uterine body, fundus, or the endometrial region to be observed. The position and / or angle of the sampling cutoff line can be adjusted and fixedly applied to all time frames in a single spatiotemporal reconstruction.

7. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 6, characterized in that, On the same set of magnetic resonance dynamic images, at least two different target anatomical sites are set along the anatomical sequence from the fundus to the cervix, and sampling cut-off lines are set at each target anatomical site to generate corresponding two-dimensional static spatiotemporal cross-sections. In each two-dimensional static spatiotemporal cross-section, a consistent waveform feature point definition is used to determine the waveform feature time and / or phase belonging to the same endometrial peristaltic event.

8. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 1, characterized in that, The two-dimensional static spatiotemporal cross-section is displayed with the horizontal axis representing time and the vertical axis representing the spatial coordinates on the sampling cut-off line; the spatial direction of the vertical axis is determined by the position, angle and image orientation information of the sampling cut-off line in the original two-dimensional imaging plane.

9. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 1, characterized in that, The number of peristaltic movements Nw is the number of mutually separated peristaltic waveforms identified in the two-dimensional static spatiotemporal cross-section within the preset observation time Tobs; the peristaltic frequency f is calculated according to f=60×Nw / Tobs, where Tobs is in seconds and f is in times / minute.

10. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 1, characterized in that, The adjacent wave spacing is obtained by measuring the distance between corresponding feature points of adjacent peristaltic waveforms in the time axis direction in the two-dimensional static spatiotemporal section diagram, and is converted into adjacent wave time intervals based on the mapping between the virtual third coordinate and real time; the rhythmicity is characterized by the consistency of multiple adjacent wave spacings or adjacent wave time intervals, and the consistency includes at least one of the mean, standard deviation, coefficient of variation or range.

11. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 1, characterized in that, The waveform height or creep amplitude is obtained by measuring the displacement height of the crest of the creep waveform relative to the baseline, adjacent trough, or preset reference position in the direction of the sampling cut-off spatial axis, and is converted into length according to the pixel spacing of the sampling cut-off spatial axis or the spatial calibration relationship after multi-plane reconstruction.

12. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 7, characterized in that, The anatomical propagation direction is determined based on the chronological order and / or phase difference of the waveform characteristic time of the same endometrial peristaltic event corresponding to at least two different target anatomical sites; when the waveform characteristic time of the target anatomical site closer to the fundus is earlier than that of the target anatomical site closer to the cervix, it is determined to be propagation from the fundus to the cervix; when the waveform characteristic time of the target anatomical site closer to the cervix is ​​earlier than that of the target anatomical site closer to the fundus, it is determined to be propagation from the cervix to the fundus; when the time difference between each target anatomical site is lower than a preset time resolution threshold or the chronological order does not satisfy the unidirectional propagation relationship, at least one of the following markers is output: undetermined direction, synchronous propagation, or complex propagation.

13. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to claim 1, characterized in that, Before constructing the virtual three-dimensional image data, at least one of the following is performed on the multi-frame magnetic resonance dynamic images, targeting the uterine region: rigid registration, affine registration, non-rigid registration, region cropping, intensity normalization, or motion artifact correction; after processing, the same sampling cutoff line corresponds to the same target anatomical location in each time frame.

14. The method for quantifying endometrial peristalsis using magnetic resonance imaging according to any one of claims 1 to 13, characterized in that, The step of determining the endometrial peristaltic waveform in the two-dimensional static spatiotemporal cross-section is completed by manual marking or by automatic or semi-automatic identification using edge detection, ridge tracing, local gradient, dynamic programming, cross-correlation, phase analysis, or a learning model.

15. A magnetic resonance imaging (MRI) system for quantifying endometrial peristalsis, used to implement the MRI method for quantifying endometrial peristalsis as described in any one of claims 1 to 14, characterized in that, include: The dynamic image acquisition module is used to acquire multiple frames of continuous magnetic resonance dynamic images of the uterine region on the same two-dimensional imaging plane, as well as acquisition time information or frame number for determining the time sequence of each frame. The time-dimensional stacking module is used to sort the multi-frame magnetic resonance dynamic images according to the acquisition time information or frame number, map the time information of each frame to a virtual third coordinate, and set the two-dimensional pixel array of each frame as the data layer of virtual three-dimensional image data according to the virtual third coordinate. The derived body data generation module is used to establish the mapping relationship between the virtual third coordinate and the corresponding real acquisition time, and to generate derived body data that can be reconstructed into multiple planes. The target site and sampling cutoff line determination module is used to determine the sampling cutoff line passing through the target anatomical site on the reference image based on the target anatomical site of the uterus to be observed, and the spatiotemporal reconstruction plane is jointly determined by the sampling cutoff line and the virtual third coordinate direction. The spatiotemporal reconstruction module is used to perform multi-plane reconstruction or equivalent resampling on the derived body data based on the spatiotemporal reconstruction plane to obtain a two-dimensional static spatiotemporal cross-section, where one coordinate corresponds to the actual acquisition time through the mapping relationship, and the other coordinate corresponds to the anatomical spatial position on the sampling section line. The quantization module is used to determine the endometrial peristaltic waveform in the two-dimensional static spatiotemporal cross-section and output at least one of the following quantization parameters: number of peristalsis, peristaltic frequency, propagation direction, spacing between adjacent waves, rhythmicity, and waveform height or peristaltic amplitude.

16. The endometrial peristalsis magnetic resonance quantification system according to claim 15, characterized in that, The endometrial peristalsis magnetic resonance quantification system is deployed on a magnetic resonance post-processing workstation, a PACS post-processing server, or a standalone medical image processing platform. The interface of the endometrial peristalsis magnetic resonance quantification system simultaneously displays the original dynamic magnetic resonance reference frame, the target anatomical site and sampling cut-off positioning window, a two-dimensional static spatiotemporal section, and the quantification results. After the sampling cut-off is moved or rotated, the corresponding two-dimensional static spatiotemporal section is regenerated and the quantification results are updated. The endometrial peristalsis magnetic resonance quantification system saves the sampling cutoff coordinates, direction, real time mapping of each time frame, and manual or automatic marker points to form a traceable record.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the processor performs the endometrial peristalsis magnetic resonance quantification method according to any one of claims 1 to 14.

18. A computer program product, comprising a computer program, characterized in that, When the computer program is run on a processor, it causes the processor to execute the endometrial peristalsis magnetic resonance quantification method according to any one of claims 1 to 14.

Images