Device and method for acquiring three-dimensional model of abdominal organ

Abstract

The invention provides a device and a method for acquiring a three-dimensional model of an abdominal organ. The device comprises a coil supporting assembly, a nuclear magnetic resonance spectrometer and image processing equipment, the coil supporting assembly is clamped and fixed to the side portion of the examination bed and used for supporting the abdomen coil to reduce compression on the abdomen of the user to be examined. The nuclear magnetic resonance spectrometer is arranged above the abdomen of a to-be-tested user, the axial line of the examination positioning lamp is aligned with the axial midpoint of the coil supporting assembly, and a magnetic resonance scanning sequence is obtained through scanning of the nuclear magnetic resonance spectrometer; and the image processing equipment is in communication connection with the nuclear magnetic resonance spectrometer, obtains a magnetic resonance scanning sequence and obtains an intra-abdominal organ three-dimensional model according to the magnetic resonance scanning sequence. According to the method, full-level accurate extraction of the abdominal organs is realized through a multi-region seed point strategy, adaptive threshold segmentation and other technologies, the low efficiency and subjective limitation of traditional manual segmentation are broken through, the reliability of the result is ensured through an automatic algorithm, and support is provided for image analysis of the abdominal organs.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to a device and method for acquiring a three-dimensional model of an abdominal organ. Background Art

[0002] In clinical practice, obtaining three-dimensional models of abdominal organs mainly relies on two technical means: CT scanning and MR scanning. Both require data acquisition and then three-dimensional image reconstruction, but each has limitations in application.

[0003] Among them, CT scanning produces ionizing radiation, which poses potential risks to radiation-sensitive groups such as pregnant women and adolescents. MR scanning does not produce ionizing radiation and is safer, but it faces the following key difficulties in actual operation: 1. Compression risk: The abdominal radiofrequency coil used in MR scanning has a certain weight and will directly compress the patient's abdominal cavity during the scan. For special populations such as pregnant women, this compression may cause fetal distress in the abdomen, posing a safety hazard; 2. Motion artifact interference: The acquisition time of MR sequences is usually long and is sensitive to motion artifacts. Abdominal organs are significantly affected by physiological movements such as respiration, gastrointestinal motility, and vascular pulsation. Although respiratory gating technology has been applied, this technology can only reduce the artifacts caused by respiratory movement to a certain extent, and cannot solve the interference caused by other physiological movements, affecting image quality; 3. Insufficient tissue contrast: The MR scanning sequence parameters are complex, and different scanning methods will affect the echo signal strength and image contrast between different tissues. Summary of the Invention

[0004] Based on this, it is necessary to provide a device and method for obtaining a three-dimensional model of abdominal organs in order to address the above technical problems.

[0005] A device for acquiring a three-dimensional model of an abdominal organ, comprising: a coil support assembly, a nuclear magnetic resonance apparatus, and an image processing device; The coil support assembly is clamped and fixed to the side of the examination bed, and is used to support the abdominal coil to reduce pressure on the abdomen of the user to be tested; the nuclear magnetic resonance apparatus is placed above the abdomen of the user to be tested, and the axial line of the inspection positioning light is aligned with the axial midpoint of the coil support assembly, and the magnetic resonance scanning sequence is obtained by scanning with the nuclear magnetic resonance apparatus; the image processing device is communicatively connected to the nuclear magnetic resonance apparatus, obtains the magnetic resonance scanning sequence, and obtains a three-dimensional model of the abdominal organs according to the magnetic resonance scanning sequence; Wherein, acquiring a three-dimensional model of an abdominal organ according to the magnetic resonance imaging sequence includes: Obtaining an abdominal visceral organ tissue segmentation result according to the magnetic resonance imaging sequence and the abdominal visceral organ automatic image segmentation technology based on grayscale features; generating an initial abdominal organ tissue model according to the abdominal organ tissue segmentation result, and adjusting the triangular facets by using a surface simplification coefficient to optimize the initial abdominal organ tissue model to obtain an optimized three-dimensional mesh model; The optimized three-dimensional mesh model is multi-dimensionally verified and standardized and exported to obtain a three-dimensional model of the abdominal viscera.

[0006] In one embodiment, the coil support assembly includes a side fixing bracket group, a fixing clamping mechanism is provided at the lower portion of the side fixing bracket group, and the fixing clamping mechanism is used to clamp the side fixing bracket group to the side of the examination bed; a height adjustment mechanism is provided at the upper portion of the side fixing bracket group, and a rotating shaft structure is provided at the top output end of the height adjustment mechanism, and the rotating shaft structure is connected to an arc-shaped support plate, and the arc-shaped support plate is rotatably installed at the top output end of the height adjustment mechanism through the rotating shaft structure; at least two abdominal coil binding through-holes are evenly provided on the arc-shaped support plate, and the abdominal coil binding through-holes are used to pass the binding belt of the abdominal coil.

[0007] In one embodiment, a nuclear magnetic resonance apparatus is placed above the abdomen of a user to be tested, an axial line of a check positioning light is aligned with an axial midpoint of a coil support assembly, and obtaining a magnetic resonance scan sequence by scanning with the nuclear magnetic resonance apparatus includes: Step S1, using a fast positioning scan sequence to acquire sagittal, coronal, and transverse positioning images, and determining the scanning range of the main sequence based on the positioning images; Step S2: Define the main magnetic field direction as the Z axis, parallel to the long axis of the human body and pointing to the foot side, the X axis pointing to the right side of the human body, and the Y axis pointing to the front side of the human body; set the gradient field spatial encoding as the Z axis as the slice direction, and set the direction with the smaller value of the abdominal anteroposterior diameter thickness and the left and right diameter width as the phase encoding direction, and the larger value as the frequency encoding direction; Step S3, based on the gradient field spatial encoding, continuously exciting the scanning range with a small-angle radio frequency pulse with a flip angle of 30°-65°, and acquiring echo signals by rapidly switching the readout gradient field to obtain raw MR signal data; Step S4: Processing the raw MR signal data, using half-echo acquisition based on Fourier space conjugate symmetry, and reconstructing a magnetic resonance scanning sequence.

[0008] In one embodiment, step S3 further includes: A preparation pulse combination of 90°, 180°, and -90° is added before the main sequence.

[0009] In one embodiment, step S3 further includes: Preset currents are applied to the gradient coils in the X-axis, Y-axis, and Z-axis directions.

[0010] In one embodiment, obtaining an abdominal visceral organ tissue segmentation result based on the magnetic resonance scanning sequence and the abdominal visceral organ automatic image segmentation technology based on grayscale features includes: Setting seed points in the high-signal area, low-signal area, and medium-signal area of ​​the target abdominal organ, and generating an initial mask covering the entire range of the abdominal organ by using a seed region growing algorithm; performing a boundary shrinkage operation on the initial mask to remove the blood vessel wall and retain the pure parenchyma area to obtain the organ parenchyma; For abdominal organs containing fluid, the fluid area is extracted using an adaptive threshold segmentation algorithm; For tiny low-signal abdominal organs, a signal threshold is obtained, and tiny low-signal structures are extracted based on a volume screening condition and the signal threshold; The intra-abdominal organ tissue segmentation result includes the organ substance, the liquid area, and the tiny low-signal structure.

[0011] In one embodiment, an initial abdominal organ tissue model is generated based on the abdominal organ tissue segmentation result, and the triangular facets are adjusted by using a surface simplification coefficient to optimize the initial abdominal organ tissue model. The optimized three-dimensional mesh model includes: Adjust the triangle patch according to the following formula: ; in, Indicates the number of original mesh faces, Indicates the number of target faces after simplification, represents the simplification factor.

[0012] In one embodiment, performing multi-dimensional verification and standardized export on the optimized three-dimensional mesh model to obtain a three-dimensional model of abdominal organs includes: Performing two-dimensional layer verification on the optimized three-dimensional mesh model, and obtaining a first verification model in response to a fit error between the segmented contour and the original image of ≤1 pixel; Superimposing the original image and the first verification model by volume rendering, and obtaining a three-dimensional model of the abdominal viscera in response to the absence or loss of perforators; The three-dimensional model export formats of the abdominal organs include STL format, VTK format and DICOM-SEG format.

[0013] A method for obtaining a three-dimensional model of an abdominal organ, used in the above-mentioned device for obtaining a three-dimensional model of an abdominal organ, comprising: Clamp the coil support assembly to the side of the examination bed; The user to be tested lies supine on the examination bed, and the abdominal coil is fixed above the abdomen by the coil support assembly, with the abdomen located at the center of the abdominal coil; The nuclear magnetic resonance apparatus is placed above the abdomen of the user to be tested, and the axis line of the positioning light is checked to be aligned with the axial midpoint of the coil support assembly, and the magnetic resonance scanning sequence is obtained by scanning with the nuclear magnetic resonance apparatus; An image processing device is in communication with the magnetic resonance imaging device to obtain the magnetic resonance scanning sequence and obtain a three-dimensional model of the abdominal organs according to the magnetic resonance scanning sequence; Wherein, acquiring a three-dimensional model of an abdominal organ according to the magnetic resonance imaging sequence includes: Obtaining an abdominal visceral organ tissue segmentation result according to the magnetic resonance imaging sequence and the abdominal visceral organ automatic image segmentation technology based on grayscale features; generating an initial abdominal organ tissue model according to the abdominal organ tissue segmentation result, and adjusting the triangular facets by using a surface simplification coefficient to optimize the initial abdominal organ tissue model to obtain an optimized three-dimensional mesh model; The optimized three-dimensional mesh model is multi-dimensionally verified and standardized and exported to obtain a three-dimensional model of the abdominal viscera.

[0014] Compared to existing technologies, the advantages and benefits of the present invention lie in: The coil support assembly, designed in this invention, utilizes adjustable side brackets, a height adjustment mechanism, and a curved support plate to stably support the abdominal coil, preventing abdominal compression in patients (especially those with special needs, such as pregnant women). Furthermore, the coil fit is ensured by the binding perforations, improving signal acquisition stability. The invention utilizes techniques such as small-angle RF pulses and extremely short repetition times combined with half-echo acquisition to significantly shorten data acquisition time. Combined with a preparatory pulse, it enhances soft tissue and fluid contrast, effectively suppressing motion artifacts and susceptibility artifacts caused by respiratory and gastrointestinal motility, thereby balancing imaging efficiency and quality. Automated segmentation technology based on grayscale features, using strategies such as multi-region seed point growth and adaptive threshold screening, enables precise extraction of organs and microstructures such as the liver, gallbladder, and stones. Surface simplification coefficients are combined to optimize 3D modeling, resulting in models that can be exported in multiple formats and are suitable for clinical scenarios such as 3D printing and surgical planning. This invention offers synergistic advantages in safety, imaging quality, automation, and clinical applicability, providing a highly efficient solution for accurately acquiring 3D models of abdominal organs. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of the structure of a device for acquiring a three-dimensional model of an abdominal organ according to a method in an embodiment; Figure 2 A schematic diagram of the coil support assembly structure in one embodiment; Figure 3A schematic diagram of shortening the scanning time based on phase encoding direction optimization in one embodiment; Figure 4 A schematic diagram of a sequence diagram of collected echo signals in one embodiment; Figure 5 Schematic diagram of the flip angle of a radio frequency pulse in one embodiment; Figure 6 is a schematic diagram of the longitudinal magnetization vector in one embodiment; Figure 7 Schematic diagram of a pulse combination of 90°, 180°, and -90° in one embodiment; Figure 8 A schematic diagram of introducing a magnetic resonance scanning sequence in one embodiment; Figure 9 A schematic diagram of trimming three-dimensional data in one embodiment; Figure 10 A schematic diagram of initial mask three-dimensional data in one embodiment; Figure 11 A schematic diagram of three-dimensional data of a liver in one embodiment; Figure 12 A schematic diagram of gallstones in one embodiment; Figure 13 A schematic diagram of adjusting a mesh triangle patch using a surface simplification coefficient in one embodiment; Figure 14 A schematic diagram of smoothing processing in one embodiment; Figure 15 A schematic diagram of a three-dimensional model of abdominal viscera in one embodiment; Figure 16 A schematic diagram of a model export type in one embodiment.

[0016] In the figure, 1-coil support assembly, 2-nuclear magnetic resonance apparatus, 3-image processing equipment, 10-side fixing bracket group, 11-fixed clamping mechanism, 12-height adjustment mechanism, 13-rotating axis structure, 14-arc support plate, 15-abdominal coil binding perforation. DETAILED DESCRIPTION

[0017] Before describing the specific embodiments of the present invention, the overall concept of the present invention is described as follows: This invention is mainly developed to address the problems of insufficient MR scanning safety, interference with imaging quality, and limited 3D modeling accuracy during the acquisition of 3D models of abdominal organs. Currently, CT scanning, which is relied upon in clinical practice, carries the risk of ionizing radiation, while traditional MR scanning faces the risk of abdominal coil compression (especially for pregnant women and other groups), significant motion artifact interference, and insufficient tissue contrast. These problems result in unstable quality of acquired image data, making it difficult for the subsequent 3D model segmentation and reconstruction accuracy to meet precise clinical needs, and may even affect examination efficiency due to scan interruptions.

[0018] After analysis, the inventors found that the main reason for the above problems is that the existing coils lack an adaptive support structure, which leads to direct compression of the abdominal cavity; the traditional scanning sequence acquisition time is too long and is sensitive to physiological movement, and lacks a targeted artifact suppression mechanism; the sequence parameter design does not fully highlight the signal difference between soft tissue and liquid, and the subsequent image segmentation mostly relies on manual operation and has a low degree of standardization. As long as innovative designs are made from multiple dimensions such as scanning device structure, sequence parameter optimization, motion artifact suppression and automated segmentation modeling, the aforementioned problems can be avoided. Therefore, the present invention proposes an MR scanning device and method for obtaining a three-dimensional model of abdominal viscera, which eliminates the risk of compression through an adjustable coil support component, improves imaging quality by using a fast acquisition sequence and multi-dimensional artifact suppression technology, and combines the automated segmentation and precise modeling process based on grayscale features to achieve safe, efficient and high-precision acquisition of three-dimensional models of abdominal viscera, effectively meeting the needs of clinical diagnosis and treatment planning.

[0019] After introducing the overall concept of the present invention, in order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below through specific implementation methods in conjunction with the accompanying drawings.

[0020] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of this specification should have the usual meanings understood by people with ordinary skills in the field to which the invention belongs. The words "first", "second" and similar terms used in one or more implementations of this specification do not indicate any order, quantity or importance, but are only used to distinguish different components. Words such as "include" or "comprise" mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Words such as "connect" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0021] For ease of understanding, the terms involved in the embodiments of the present invention are explained below: MR stands for Magnetic Resonance. In medicine, "MR" is often combined with "imaging," as in magnetic resonance imaging (MRI).

[0022] CT stands for Computed Tomography. It is an imaging technique that uses X-rays to scan specific parts of the human body and then reconstructs the tomographic images through computer processing.

[0023] TR time: Repetition Time, which means repetition time in Chinese. It is a key parameter in MRI scanning sequence and refers to the time interval between two adjacent radiofrequency pulse excitations.

[0024] T1 value: Longitudinal Relaxation Time. It describes the time constant required for tissue nuclei (primarily hydrogen protons) to recover from an excited state to an equilibrium state (longitudinal magnetization vector recovery) after the RF pulse stops. It reflects the speed of longitudinal magnetization recovery.

[0025] K-space (Fourier space) is a core concept in MRI image reconstruction. It is essentially a frequency-domain data space used to store the frequency and phase information of magnetic resonance signals, rather than direct anatomical spatial images.

[0026] TE (Echo Time) is a key parameter in magnetic resonance imaging (MRI), referring to the time interval between the excitation of the radio frequency pulse and the acquisition of the echo signal.

[0027] T2 value: Transverse Relaxation Time. It is the time constant that describes the decay of the transverse magnetization vector of hydrogen protons in tissue after the radiofrequency pulse stops, reflecting the speed at which the transverse magnetization disappears.

[0028] SAR value: Specific Absorption Rate, which means specific absorption rate in Chinese. It is a physical quantity that measures the rate at which the human body absorbs electromagnetic energy in an electromagnetic field.

[0029] DICOM (Digital Imaging and Communications in Medicine) is an internationally standardized data format for medical imaging. It is used to store, transmit, and share medical images and related information. It is the core data carrier for hospital PACS (Picture Archiving and Communication Systems) and RIS (Radiology Information Systems).

[0030] STL: Standard Tessellation Language. A file format used to represent the surface geometry of three-dimensional models.

[0031] VTK: Visualization Toolkit. An open source, cross-platform 3D visualization and graphics processing library.

[0032] DICOM-SEG: DICOM Segmentation Image IOD (Information Object Definition) is a file format in the DICOM standard specifically used to store and transmit medical image segmentation results.

[0033] RAI coordinate system: In the field of medical imaging, it is a form of expression of the anatomical coordinate system. R (Right) represents the negative left direction, that is, from right to left, A (Anterior) represents the negative posterior direction, that is, from front to back, and I (Inferior) represents the negative upward direction.

[0034] ROI: Region of Interest. It refers to the specific area in an image or 3D data that researchers or clinicians focus on.

[0035] SSF: Surface Simplification Factor, which means surface simplification coefficient in Chinese.

[0036] : Delta B-zero, describes the main magnetic field inhomogeneity, and represents the difference between the actual main magnetic field and the ideal uniform main magnetic field, that is, the main magnetic field inhomogeneity value.

[0037] In one embodiment, Figure 1 As shown, a device for acquiring a three-dimensional model of an abdominal organ is provided, comprising: a coil support assembly 1, a nuclear magnetic resonance apparatus 2, and an image processing device 3; The coil support assembly 1 is clamped and fixed to the side of the examination bed, and is used to support the abdominal coil to reduce the pressure on the abdomen of the user to be tested; the nuclear magnetic resonance apparatus 2 is set above the abdomen of the user to be tested, and the axial line of the inspection positioning light is aligned with the axial midpoint of the coil support assembly 1, and the magnetic resonance scanning sequence is obtained by scanning with the nuclear magnetic resonance apparatus 2; the image processing device 3 is communicated with the nuclear magnetic resonance apparatus 2 to obtain the magnetic resonance scanning sequence and obtain the three-dimensional model of the abdominal organs according to the magnetic resonance scanning sequence.

[0038] like Figure 2 As shown, the coil support assembly 1 includes a side fixing bracket group 10, and a fixing clamping mechanism 11 is provided at the lower part of the side fixing bracket group 10, and the fixing clamping mechanism 11 is used to clamp the side fixing bracket group 10 to the side of the examination bed; a height adjustment mechanism 12 is provided at the upper part of the side fixing bracket group 10, and a rotating shaft structure 13 is provided at the top output end of the height adjustment mechanism 12, and the rotating shaft structure 13 is connected to an arc-shaped support plate 14, and the arc-shaped support plate 14 is rotatably installed at the top output end of the height adjustment mechanism 12 through the rotating shaft structure 13; at least two abdominal coil binding through-holes 15 are evenly opened on the arc-shaped support plate 14, and the abdominal coil binding through-holes 15 are used to pass the binding belt of the abdominal coil.

[0039] Place the rear abdomen coil on the examination bed, and then fix the side fixing bracket group 10 on the left and right sides of the examination bed respectively through the fixing clamping mechanism 11, flip up the arc-shaped support plate 14, and the abdomen of the user to be tested can easily lie on the examination bed and the rear abdomen coil, then flip the arc-shaped support plate 14 so that it is buckled on the limiting support platform on the side fixing bracket group 10, and then place the front abdomen coil on the upper surface of the arc-shaped support plate 14, and adjust the extended height of the side fixing bracket 10 through the height adjustment mechanism 12 so that the lower surface of the arc-shaped support plate 14 basically fits the upper abdomen of the user to be tested without forming any oppression, and then use the binding belt to pass through the abdominal coil binding perforation 15 to achieve tightening and binding between the two abdominal coils, so that the front and back abdominal coils fit the human body better and the examination effect is better.

[0040] In this embodiment, the coil support group 1 adopts a combination design of a side fixing bracket group 10 and a fixed clamping mechanism 11, which can be stably clamped on the side of the examination bed, is easy to install and is adaptable to examination beds of different specifications; the height adjustment mechanism 12 accurately and synchronously adjusts the support height on both sides to ensure that the arc-shaped support plate 14 fits stably with the abdomen of the user to be tested, effectively disperses the weight of the abdominal coil, and avoids the risk of compression to the abdominal cavity (especially special groups such as pregnant women); the arc-shaped support plate 14 is flexibly flipped through the rotating shaft structure 13, which is convenient for the user to be tested to get on and off the bed, and forms a stable support through the arch-bridge design, and is matched with the abdominal coil to ensure its fit with the human body, thereby improving the stability of signal acquisition; the overall structure takes into account both safety and comfort, and adapts to subjects of different body shapes through multi-dimensional adjustment, providing a reliable coil support solution for high-quality MR scanning.

[0041] In one embodiment, the nuclear magnetic resonance apparatus 2 is placed above the abdomen of the user to be tested, and the axis line of the inspection positioning light is aligned with the axial midpoint of the coil support assembly 1. The magnetic resonance scanning sequence obtained by scanning with the nuclear magnetic resonance apparatus 2 includes: Step S1: Use a fast positioning scan sequence to acquire sagittal, coronal, and transverse positioning images, and determine the scanning range of the main sequence based on the positioning images.

[0042] Step S2: Define the main magnetic field direction as the Z axis, parallel to the long axis of the human body and pointing to the foot side, the X axis pointing to the right side of the human body, and the Y axis pointing to the front side of the human body; set the gradient field spatial encoding as the Z axis as the slice direction, and set the direction with the smaller value of the abdominal anteroposterior diameter thickness and the left and right diameter width as the phase encoding direction, and the larger value as the frequency encoding direction; Step S3, based on the gradient field spatial encoding, continuously exciting the scanning range with a small-angle radio frequency pulse with a flip angle of 30°-65°, and acquiring echo signals by rapidly switching the readout gradient field to obtain raw MR signal data; Step S4: Processing the raw MR signal data, using half-echo acquisition based on Fourier space conjugate symmetry, and reconstructing a magnetic resonance scanning sequence.

[0043] Specifically, a rapid positioning scan sequence is used to acquire sagittal, coronal, and transverse positioning images. The final scanning range of the main sequence is determined based on the obtained positioning image, which must include the entire abdominal viscera. The main magnetic field direction is defined as the Z-axis direction of the nuclear magnetic resonance apparatus 2, that is, when the user to be tested is in a supine position with the foot first, the Z-axis is parallel to the long axis of the human body and points to the foot side; the X-axis and Y-axis are perpendicular to the Z-axis, the Y-axis points to the front side of the human anatomical position, and the X-axis points to the right side of the human anatomical position. Gradient fields are applied to the gradient coils in the X, Y, and Z directions respectively for spatial encoding, and the Z-axis direction is set as the layer direction; according to the individual differences in the anteroposterior diameter thickness and left-right diameter width of the user to be tested, the direction with a smaller value is set as the phase encoding direction, and the direction with a larger value is set as the frequency encoding direction.

[0044] like Figure 3 As mentioned above, you can increase or decrease the number of scan slices based on the size of the scan area. This will increase or decrease the scan time accordingly, typically fluctuating by a few seconds. Scan time is directly related to the number of phase encoding steps. Setting the direction of shorter anatomical diameters as the phase encoding direction can reduce scan time by setting a rectangular field of view. In the figure, Kx represents the number of steps in the frequency encoding direction, and Ky represents the number of steps in the phase encoding direction.

[0045] Accurately measure the dimensions of the scanned area in both the long and short axis directions. If the scan area is large, such as a comprehensive scan of the entire abdomen to observe multiple organs, the anatomical diameters of this area are long. To fully cover this area, the number of scan layers should be appropriately increased. Conversely, if only a specific, smaller intra-abdominal organ (such as a localized gallbladder) is being scanned, the number of scan layers can be reduced accordingly due to its short anatomical diameter. Using a rapid positioning scan sequence to acquire sagittal, coronal, and transverse positioning images, the specific diameters of the scanned area can be accurately measured, providing a reliable basis for the preliminary determination of the number of scan layers. Assuming that the positioning image measures the coronal diameter of a patient's liver to be 15 cm, based on the equipment parameters, if a 3 mm slice thickness is used, to fully cover the liver, the preliminary estimate of the number of scan layers is approximately 150 mm ÷ 3 mm = 50, but this needs to be further adjusted based on other factors.

[0046] The selected slices are excited continuously by using a small-angle radio frequency pulse with a flip angle of 30°-65°, and the echo signals are collected by fast switching of the readout gradient field. Sequence diagram such as Figure 4 shown.

[0047] This sequence uses a very short repetition time (TR < 5ms) to continuously excite the tissue and generate a radiofrequency magnetic field (B1 field). The flip angle of each radiofrequency pulse is the same, but the polarity between each two radiofrequency pulses is different (for example, Figure 5 As shown, the flip angle of each RF pulse is 45°, but the first one is in the same direction as the main magnetic field, the second one is in the opposite direction, and so on. After the first 4-6 RF pulses, echo signals are not collected. Instead, echo signals are collected by rapidly switching the readout gradient field after the macroscopic longitudinal and macroscopic transverse magnetization vectors of the tissue have reached a relatively stable state.

[0048] Taking the macroscopic longitudinal magnetization vector as an example, before each RF pulse excitation, its size is the superposition of the longitudinal residual component and the longitudinal magnetization vector recovered in the previous repetition time (TR). Since the tissue is continuously excited with an extremely short repetition time shorter than the tissue T1 value, each tissue cannot complete sufficient longitudinal relaxation in each TR cycle. Assuming that the flip angle of the excitation pulse is α and the initial longitudinal magnetization vector size is 1, then after the first excitation, the size of the longitudinal residual component is 1×cosα. Before the second RF pulse excitation, the longitudinal magnetization vector becomes the sum of cosα and M1 (where M1 is the size of the magnetization vector restored to the longitudinal direction in the first TR); after the second excitation, the size of the longitudinal residual component is (cosα+M1)×cosα, and M2 represents the size of the magnetization vector restored to the longitudinal direction in the second TR. For example Figure 6As shown in the figure, after multiple excitations, the tissue's longitudinal magnetization vector gradually reaches a relatively stable state. Gradient fields with the same magnitude and opposite direction as the corresponding encoding gradients are applied in the slice selection, phase encoding, and frequency encoding directions, correcting the proton group dephasing caused by these three encoding gradient fields.

[0049] Because the Fourier space (K-space), the raw data space of the acquired MR signals, is mirror-symmetric (conjugate symmetric) in the frequency encoding direction, only half an echo is acquired. The other half can be simulated and filled in according to the principle of symmetry to obtain the magnetic resonance scan sequence. Using half-echo acquisition will further shorten the scan time of the main sequence.

[0050] The conjugate symmetry of K-space data originates from the real number characteristics of MRI signals. Assuming that the acquired time domain signal is a real number, then its Fourier transform is satisfy: ; in represents the complex conjugate. This means that in K space k and The position data are mutually conjugate complex numbers, with symmetric real parts and antisymmetric imaginary parts. ,but The data can be obtained through generate.

[0051] In this embodiment, the use of a rapid localization scan sequence to acquire sagittal, coronal, and transverse localization images and determine the main sequence scan range based on these images has multiple significant advantages. First, the rapid localization sequence can complete multi-plane image acquisition in a very short time, reducing the waiting and breath-holding burden on the examinee, making it particularly suitable for special populations (such as pregnant women and elderly patients) who cannot maintain a fixed posture for long periods of time. Second, the localization images in three orthogonal planes can fully and intuitively display the anatomical position, morphology, and adjacent relationships of the abdominal organs, providing a three-dimensional reference for accurately framing the main sequence scan range, avoiding unnecessary time waste due to an excessively large scan range, and preventing the omission of abdominal organs or lesions due to an undersized scan range. Furthermore, this approach can flexibly adjust the scan boundaries based on individual anatomical differences (such as the size and positional variation of abdominal organs), ensuring that the main sequence acquisition data fully covers the target area (e.g., the entire liver and gallbladder, the entire bladder), laying a complete and accurate data foundation for subsequent image reconstruction and three-dimensional modeling, thereby improving the efficiency and reliability of the overall examination.

[0052] Small-angle RF pulses can reduce excessive consumption of the longitudinal magnetization vector of the tissue and achieve continuous excitation at an extremely short repetition time (TR < 5ms). This can not only shorten the overall data acquisition time and support the user to be tested to complete the scan by holding their breath, effectively avoiding artifact interference caused by physiological movements such as breathing and gastrointestinal motility, but also maintain stable signal output and ensure the image signal-to-noise ratio. At the same time, the design of opposite polarity of every two RF pulses, combined with the strategy of waiting for the magnetization vector to stabilize after 4-6 excitations before acquiring the signal, can reduce the cumulative error of the residual magnetization vector and improve signal consistency. The rapid switching of the readout gradient field can accelerate the acquisition of echo signals and further compress the sequence time. By applying reverse gradient fields in the three encoding directions to correct the phase loss of the proton group, the artifacts can be significantly reduced. Ultimately, while ensuring high contrast between soft tissue and liquid, it provides high-quality and highly stable raw data for subsequent three-dimensional model reconstruction, taking into account both scanning efficiency and imaging accuracy.

[0053] The conjugate symmetry of K-space data (i.e., the k and -k position data are conjugate complex numbers of each other) provides a theoretical basis for half-echo acquisition. By acquiring only half of the signal after the echo peak (corresponding to the k ≥ 0 region), the other half of the echo data can be simulated and filled in using the principle of symmetry. This can significantly shorten the main sequence scanning time while ensuring image quality and reduce the breath-holding burden on the user under test. It is particularly suitable for areas that are susceptible to physiological movement, such as the abdomen. At the same time, half-echo acquisition avoids the longer gradient switching time required for full echo acquisition, reducing the risk of artifacts caused by gradient field changes. In addition, during the reconstruction process, mathematical symmetry is used to complete the data, ensuring that image resolution and signal integrity are not compromised. Ultimately, while improving scanning efficiency, it provides a high-quality imaging foundation for the precise segmentation and reconstruction of three-dimensional models of abdominal organs, achieving an effective balance between efficiency and accuracy.

[0054] In one embodiment, step S3 further includes: A preparation pulse combination of 90°, 180°, and -90° is added before the main sequence.

[0055] Specifically, such as Figure 7 As shown, a 90°, 180°, and -90° pulse combination is added before the main sequence as a preparation pulse combination. The first 90° pulse in this preparation pulse converts the tissue's macroscopic longitudinal magnetization vector into a macroscopic transverse magnetization vector. After the 90° pulse is turned off, a 180° RF pulse is applied at the appropriate time (at 1 / 2 the TE of the preparation pulse). The macroscopic transverse magnetization vector then refocuses at the TE of the preparation pulse. Due to the varying T2 relaxation rates of different tissues, there is T2 contrast, resulting in differences in the refocused macroscopic transverse magnetization vector. A -90° pulse then converts the macroscopic transverse magnetization vector back to the longitudinal direction, returning it to the macroscopic longitudinal magnetization vector.

[0056] In this embodiment, a 90° pulse first converts the longitudinal magnetization vector to the transverse direction. A 180° pulse, at a specific moment, causes the transverse magnetization vector to refocus, highlighting T2 relaxation differences between different tissues. Finally, a -90° pulse converts the refocused transverse magnetization vector back to the longitudinal direction. This process not only enhances the T2 contrast between soft tissues and fluids (such as bile, amniotic fluid, and blood), making the boundaries of intra-abdominal organs (such as the liver and gallbladder, and the fetus and amniotic fluid) more clearly visible in the image, but also reduces the interference of magnetic susceptibility artifacts on imaging quality through magnetic field adjustment. Furthermore, this preparation pulse combination is completed before the main sequence is started, without increasing the breath-holding time of the patient being tested. This is particularly suitable for special populations such as pregnant women who require shortened scan times. Ultimately, it provides high-quality image data with better contrast and fewer artifacts for the subsequent precise segmentation and reconstruction of 3D models of intra-abdominal organs, significantly improving the accuracy of identifying subtle structures (such as the common bile duct and stones) in clinical applications.

[0057] In one embodiment, because MRI acquires K-space data line by line through spatial encoding (phase encoding, frequency encoding, and inter-slice encoding) using gradient magnetic fields, the scan time is proportional to the number of encoding steps. Reducing the number of encoding steps (i.e., "undersampling") can shorten the scan time, but undersampling results in incomplete K-space data, and direct reconstruction can produce aliasing artifacts. A very short calibration prescan can be used to obtain spatial sensitivity information of the ankle coil in advance. The following algorithm can then be used to "de-alias" the undersampled data, reconstructing an artifact-free image.

[0058] ; in, Represents the original signal of the coil, Indicates the pixels, Indicates the total number of pixels within the scan range, Represents pixels The actual signal strength, Indicates the The coil unit is The sensitivity of the pixels, Indicates noise.

[0059] Specifically, before the main sequence scan, a very short (usually within a few seconds) calibration pre-scan is performed to allow the M units of the abdominal coil (such as an 8-channel or 16-channel array coil) to synchronously collect a small amount of fully sampled reference data (covering the key layers of the range to be scanned, which does not need to be completely covered, but must include the signal characteristics of all pixels). At this time, multiple units of the abdominal coil (such as different channels of the array coil) will synchronously collect a small amount of fully sampled reference data (covering the key layers of the range to be scanned). By analyzing these reference data, the sensitivity coefficient of each coil unit to the N pixels within the scanning range can be calculated (that is, the signal reception sensitivity of the coil unit to the pixel, expressed in matrix form as c, where represents the sensitivity of the m-th coil unit to the n-th pixel). This sensitivity information will be stored for subsequent anti-aliasing.

[0060] During the main sequence scan, the number of encoding steps in the phase encoding direction and / or the inter-slice encoding direction is actively reduced (i.e., undersampling). For example, only 50%-70% of the normal encoding steps are acquired. Since MRI scan time is directly related to the number of encoding steps, this can significantly shorten data acquisition time (e.g., from 10 seconds to 5-7 seconds) while also reducing RF energy deposition (SAR value). However, undersampling can result in incomplete K-space data. When directly reconstructing the image through inverse Fourier transform, signals from adjacent pixels will alias, resulting in streaky or overlapping artifacts.

[0061] The coil spatial sensitivity matrix c obtained in the first step is used to process the undersampled K-space data to eliminate aliasing artifacts. First, the undersampled K-space data is inversely Fourier transformed to obtain the original image containing aliasing artifacts and record the original coil signal. :The M units of the abdominal coil synchronously collect under-sampled data, and the original signal output by each unit is the formula , which contains the real signal of the target pixel, aliasing interference and noise (such as electronic noise, thermal noise).

[0062] A mathematical model is constructed based on the coil sensitivity matrix c, and the inverse problem is solved (such as using compressed sensing or parallel imaging algorithms). Separate the true signal of each pixel , so that the true signal of each pixel can be extracted separately.

[0063] After processing with the dealisimetry algorithm, the true signal intensity of each pixel is obtained. Conventional image reconstruction processes (such as Fourier transform and grayscale correction) then produce aliasing artifact-free MRI images. This image retains the time advantage of undersampling (scan time is shortened by 30%-50%) while compensating for data loss through coil sensitivity information. The image resolution and detail are consistent with full sampling results.

[0064] In this embodiment, the spatial sensitivity differences of the abdominal coils are exploited to reconstruct a complete image while minimizing the amount of acquired data, further shortening scan time and reducing SAR values. By sacrificing a small amount of calibration time in exchange for a significant reduction in main sequence time, while also utilizing the natural spatial sensitivity differences of the coils to address undersampling artifacts, this approach is particularly suitable for abdominal scans—it reduces the breath-holding burden on the patient (shortening scan time) while also avoiding image quality degradation caused by undersampling, providing clear original images for subsequent abdominal organ segmentation and 3D modeling. This method can be used in both phase encoding and interslice encoding directions, significantly reducing sequence acquisition time and SAR values.

[0065] In one embodiment, step S3 further includes: Preset currents are applied to the gradient coils in the X-axis, Y-axis, and Z-axis directions.

[0066] Specifically, for the scanning range, especially when the scanning range is close to the diaphragm or there is interference from intra-abdominal gas, appropriate currents can be applied to the gradient coils in the X, Y, and Z directions and superimposed on the original main magnetic field, thereby compensating for the unevenness of the original main magnetic field and further reducing the generation of magnetic susceptibility artifacts.

[0067] The specific compensation logic is as follows: The inhomogeneity of the main magnetic field will lead to differences in the proton resonance frequency, causing magnetic susceptibility artifacts. Applying current through the gradient coil can generate a linearly changing gradient field with a direction similar to that of the gradient coil. On the contrary, the inhomogeneity is offset. For example, if there is a vertical magnetic field drift in the diaphragm top area ( Along the Z axis), current is applied to the Z axis gradient coil to generate a Z axis gradient field, causing the magnetic field to change linearly in the Z axis direction, offsetting .

[0068] The gradient field strength is proportional to the coil current. in, is the gradient field strength, is the sensitivity coefficient of the gradient coil, which is determined by the coil design (for example, the Z-axis coil of a certain device has K=2mT / m / A, and the X / Y-axis coil has K=1.5mT / m / A). is the coil current.

[0069] The specific superposition method and applied current calculation example scenario are as follows: The problem of compensating for the Z-axis magnetic field unevenness in the diaphragm top area is manifested as follows: respiratory movement causes the main magnetic field in the diaphragm top area to be uneven in the Z-axis direction (up and down direction), resulting in signal loss or deformation.

[0070] First, measure the main magnetic field inhomogeneity value, and obtain the main magnetic field inhomogeneity value of the diaphragm top area through field map scanning (such as multi-echo phase difference method) Through the gyromagnetic ratio of hydrogen protons, Converted into the corresponding resonant frequency shift, ,in, Indicates the hydrogen proton resonance frequency shift caused by the inhomogeneity of the main magnetic field, represents the gyromagnetic ratio of hydrogen protons, Indicates the main magnetic field inhomogeneity value.

[0071] Then calculate the required gradient field, the gradient field needs to be compensated The spatial variation of the gradient field is calculated according to the following formula: ,in, represents the required gradient field, represents the main magnetic field inhomogeneity value, represents the gyromagnetic ratio of hydrogen protons, Indicates the length of the scanning range along the Z axis.

[0072] Finally, the applied current is calculated, ,in, represents the applied current, Indicates the Z-axis coil sensitivity, represents the required gradient field.

[0073] After the current is applied, the Z-axis gradient field ,offset , restore the magnetic field uniformity in the diaphragmatic region and reduce respiratory-related artifacts.

[0074] Superimposed current is different from passive field shimming (such as attaching field shimming sheets and adjusting the patient's position) and is more suitable for complex clinical scanning needs. The current is applied through the built-in field shimming coil of the MRI device (surrounding the scanning hole) without the need for any physical equipment to contact the patient's body.

[0075] The current parameters can be adjusted in real time according to individual patient differences (such as body shape, weight, anatomical variations) or characteristics of the scanning site (such as the diaphragm top, abdominal gas area, and near metal implants), and the compensation speed is very fast, making the magnetic field in the scanning area more uniform.

[0076] The magnitude, direction, and area of ​​action of the current can all be quantitatively adjusted via the MRI console. The strength of the main magnetic field is maintained by the fixed current of the superconducting magnet. The superimposed current only slightly modifies the local magnetic field and does not affect the overall strength of the main magnetic field.

[0077] In one embodiment, the image processing device 3 is in communication with the nuclear magnetic resonance apparatus 2, and obtaining the magnetic resonance scan sequence and obtaining the three-dimensional model of the abdominal viscera according to the magnetic resonance scan sequence includes: Obtaining an abdominal visceral organ tissue segmentation result according to the magnetic resonance imaging sequence and the abdominal visceral organ automatic image segmentation technology based on grayscale features; generating an initial abdominal organ tissue model according to the abdominal organ tissue segmentation result, and adjusting the triangular facets by using a surface simplification coefficient to optimize the initial abdominal organ tissue model to obtain an optimized three-dimensional mesh model; The optimized three-dimensional mesh model is multi-dimensionally verified and standardized and exported to obtain a three-dimensional model of the abdominal viscera.

[0078] The abdominal viscera tissue segmentation results obtained according to the magnetic resonance imaging sequence and the abdominal viscera automatic image segmentation technology based on grayscale features include: Setting seed points in the high signal area, low signal area and medium signal area of ​​the abdominal viscera respectively, and generating an initial mask covering the entire range of the abdominal viscera by using a seed region growing algorithm; performing a boundary shrinkage operation on the initial mask to remove the blood vessel wall and retain the pure parenchyma area to obtain the organ parenchyma; For abdominal organs containing fluid, the fluid area is extracted using an adaptive threshold segmentation algorithm; For tiny low-signal abdominal organs, a signal threshold is obtained, and tiny low-signal structures are extracted based on a volume screening condition and the signal threshold; The abdominal visceral tissue segmentation result includes the organ substance, the liquid area, and the tiny low-signal structure. An initial abdominal visceral tissue model is generated based on the abdominal visceral tissue segmentation result, and the triangular facets are adjusted by the surface simplification coefficient to optimize the initial abdominal visceral tissue model. The optimized three-dimensional mesh model includes: Adjust the triangle patch according to the following formula: ; in, Indicates the number of original mesh faces, Indicates the number of target faces after simplification, represents the simplification factor.

[0079] The optimized three-dimensional mesh model is multi-dimensionally verified and standardized and exported to obtain a three-dimensional model of abdominal organs including: Performing two-dimensional layer verification on the optimized three-dimensional mesh model, and obtaining a first verification model in response to a fit error between the segmented contour and the original image of ≤1 pixel; Superimposing the original image and the first verification model by volume rendering, and obtaining a three-dimensional model of the abdominal viscera in response to the presence or absence of perforators; The three-dimensional model export formats of the abdominal organs include STL format, VTK format and DICOM-SEG format.

[0080] This embodiment uses the medical image processing software 3D Slicer as an example to explain the above embodiments. The operation of the present invention includes but is not limited to this medical image processing software, and any medical image processing software that can achieve this function is available.

[0081] Specifically, first, Figure 8 As shown, an MRI scan sequence (DICOM data) is imported into 3D Slicer. This sequence has a "high fluid contrast" feature (sensitive to fluid signals such as bile and blood), laying the foundation for subsequent grayscale feature-based segmentation. By selecting a 3D gradient echo high fluid contrast sequence for the target organ, ensuring that the imported MRI scan sequence is the target high fluid contrast sequence, the MRI scan sequence is loaded as volume data to generate 3D data.

[0082] like Figure 9 As shown, to trim the 3D data, the data orientation is uniformly marked in the international standard RAI (right-anterior-inferior) coordinate system to ensure spatial positioning consistency. Voxel spacing is strictly verified, requiring ≤1.2mm×1.2mm×1.2mm. If the original slice thickness exceeds 1.5mm, the imported 3D data is spatially trimmed to retain the region of interest and remove irrelevant portions. Resampling is performed using 0.8mm isotropic voxels. This improves spatial resolution, eliminates the impact of slice thickness differences on subsequent 3D modeling accuracy, and ensures that details of the edges of abdominal organs and microstructures (such as the common bile duct and stones) are preserved.

[0083] After data preprocessing, the signal differences of tissues in the high-fluid-contrast sequence were used to set positive and negative seed points in the high-signal area, medium-signal parenchymal area, and surrounding low-signal background area of ​​the abdominal organs, respectively. The seed region growing algorithm was used to combine tissue signal characteristics and spatial continuity to generate an initial coarse mask covering the entire range of the abdominal organs.

[0084] A boundary shrinkage operation is performed on the obtained initial rough mask to remove non-substantial structures such as blood vessel walls that may be contained at the edge of the mask, and retain the pure substantial area of ​​the abdominal organs.

[0085] If the intra-abdominal organs contain fluid cavities (such as the cavity structure of a fluid-containing organ), the fluid region is extracted using an adaptive threshold segmentation algorithm. The threshold range is [SI_mean - 0.5SD, SI_mean + 2SD], where SI_mean is the signal mean of the fluid region and SD is the signal standard deviation. A smoothing algorithm is applied to the extracted tubular or irregular structure region to optimize the edges and eliminate stair-step artifacts to obtain the organ parenchyma.

[0086] For tiny low-signal structures that may exist in abdominal organs, a specific threshold screening method is used based on the signal contrast between them and surrounding tissues, and the threshold is set to be less than the signal mean of normal tissue around the tiny structure -2.5SD; volume screening operation is performed on candidate voxels that pass the threshold screening, and clinically significant tiny low-signal structures are retained. Their quantitative parameters such as volume and maximum diameter are automatically calculated, and finally a complete segmentation result including the abdominal organ substance, internal fluid and tiny abnormal structures is obtained.

[0087] Specifically, such as Figure 10 As shown, the above embodiments are described by taking the liver as an example.

[0088] Create a new segmentation node to integrate the segmentation results of liver-related structures.

[0089] A liver parenchyma segmentation sub-item was added. Using the seeded region growing method, the system exploits the high signal intensity of blood flow (portal vein and hepatic vein), intermediate signal intensity of liver parenchyma, and low signal intensity of background. A multi-region seeding strategy was implemented: 3-5 positive seed points were placed in the portal vein trunk (high signal area), 2-3 negative seed points were placed in the background outside the liver (low signal area), and 4-6 core seed points were randomly placed within the liver parenchyma (intermediate signal area). The signal gradient differences between the seed points were used to guide the segmentation range. An initial mask encompassing the liver parenchyma and vascular structures was generated, enabling automated preliminary localization of the overall liver contour.

[0090] Specifically, such as Figure 11 As shown, the above embodiments are described by taking liver parenchyma edge optimization and gallbladder-common bile duct separation as examples.

[0091] To address the issue of blood vessels being included in the initial mask, boundary adjustment operations are performed on the segmentation results (such as the contour masks of organs and lesions). By shrinking the boundaries, the blood vessel walls are accurately removed, the pure liver parenchyma area is retained, and the segmentation ambiguity caused by the overlap of blood vessel and parenchymal signals is resolved.

[0092] A new gallbladder bile segmentation sub-item was created. This algorithm leverages the characteristic high signal intensity of bile and low signal intensity of the gallbladder wall in the sequence. It uses an adaptive threshold segmentation algorithm: Automatic segmentation is performed based on image grayscale values. The mean (SI_bile_mean) and standard deviation (SD) of the bile signal are obtained by manually drawing a ROI within the gallbladder cavity. The threshold is dynamically set within the range [SI_bile_mean–0.5SD, SI_bile_mean+2SD]. This threshold range is adaptive to fluctuations in bile signal intensity across individuals, avoiding the inapplicability of fixed thresholds to signal variation.

[0093] Optimization operations are performed on the scattered small areas that may exist in the segmentation results (among the small areas that are not connected to the main area in the segmentation results, only the main area with the largest volume and the best connectivity is retained from all the segmented areas, and other small areas are automatically deleted). Small volume artifacts (such as interference from adjacent high-signal blood vessels) are removed through morphological screening to ensure the complete extraction of the gallbladder cavity.

[0094] A new common bile duct segmentation sub-item was created. Taking advantage of the signal consistency between the bile in the common bile duct and the gallbladder bile, 2-3 seed points were placed in the common bile duct region between the gallbladder neck and the pancreatic head. High-signal bile columns were quickly extracted using a semi-automatic segmentation tool based on a region growing algorithm. The segmentation results (such as the mask of the gallbladder bile region) were smoothed using a "median filter" with a filter kernel size of 1.0 mm to eliminate stair-step artifacts and achieve continuous segmentation of the gallbladder and common bile duct.

[0095] Specifically, such as Figure 12 As shown, the above embodiments are described by taking the automatic detection and marking of gallstones as an example.

[0096] The stone segmentation sub-item uses a specific threshold screening method based on the characteristic that stones show significant low signal against a high-signal bile background.

[0097] The threshold segmentation tool was used to set the low signal threshold to < bile signal mean – 2.5SD (standard deviation), and candidate voxels were extracted using the significant signal difference between stones and bile.

[0098] Optimize the scattered small areas that may exist in the execution of candidate voxels (from all the segmented areas, only the N main areas with the largest volume and the best connectivity are retained for the small areas that are not connected to the main area, and the other small areas are automatically deleted) (N value is set according to clinical needs, usually 5), remove tiny noise artifacts through volume screening, and accurately locate clinically significant stones.

[0099] The segmentation statistics tool is used to calculate and derive quantitative parameters such as volume and maximum diameter for each screened stone area, providing objective data support for clinical diagnosis.

[0100] Based on the results of abdominal organ segmentation, an initial 3D mesh model is generated. This model contains all the geometric details in the original image, but often increases the data volume and computational burden due to redundant patches (such as small fluctuations caused by noise). Figure 13 As shown in the figure, the surface simplification factor (SSF) is introduced to adjust the retention ratio of triangle faces in the mesh. It reduces the number of redundant faces while ensuring that key anatomical details (such as the texture of the liver capsule, the connection between the gallbladder neck and the common bile duct) are not lost, thereby balancing the model visualization accuracy and computational efficiency.

[0101] The calculation formula of the surface simplification coefficient is: ; in, is the number of original mesh faces, To simplify the number of target faces, The value range is 0.2-0.5. In practical applications, it can be dynamically adjusted based on organ complexity: for detail-rich areas such as the liver capsule and the gallbladder neck-common bile duct junction, the SSF is set to 0.4-0.5 to preserve more facets; for simple structural areas such as stones, the SSF is set to 0.2-0.3 to reduce computational complexity. Through parametric control, key clinical anatomical features can be accurately preserved while significantly reducing model complexity, taking into account the efficiency requirements of subsequent visualization analysis and surgical planning.

[0102] like Figure 14 As shown, the smoothing tool (10-20 iterations) smooths the surface of the organ's parenchyma (caused by noise). For liquid regions, multiple independent discrete regions in the segmentation result are merged into a continuous, integrated morphology to eliminate gaps and ensure continuity. For tiny, low-signal target structures, the simplification tool reduces the number of polygons to simplify the mesh while preserving core morphological features. The optimized model accurately reflects the spatial relationship between organs, providing high-fidelity digital models for clinical applications such as 3D printing and surgical navigation.

[0103] like Figure 15 As shown, the optimized three-dimensional grid model is verified in two dimensions, and the segmentation contour and the original image are compared layer by layer. The fitting error between the segmentation contour and the original image is ≤1 pixel, and a first verification model is obtained.

[0104] The original image is superimposed on the three-dimensional model through volume rendering to check whether the model has perforators (such as incorrect connection between the liver and diaphragm) or missing branches (such as omission of the edge of the liver lobe), ensure the accuracy of the three-dimensional spatial structure, and obtain a three-dimensional model of the abdominal organs.

[0105] like Figure 16 As shown in the figure, the three-dimensional model of abdominal organs supports exporting in STL format (suitable for 3D printing), VTK format (suitable for mechanical research such as finite element analysis), and DICOM-SEG format (suitable for clinical system archiving), enabling multi-scenario application adaptation.

[0106] In this embodiment, a multi-region seed point strategy accurately combines the features of high, medium, and low signal areas to quickly generate an initial mask covering the entire abdominal organ, balancing segmentation efficiency and regional integrity. By accurately removing the vascular wall through boundary shrinkage, the fuzzy boundary problem caused by the overlap of vascular and organ parenchymal signals is effectively resolved, ensuring the purity of the organ parenchymal area. Adaptive threshold segmentation is used for fluid-containing organs to dynamically match individual signal differences, avoiding the limitations of fixed thresholds on liquid area extraction. The tiny low-signal structure extraction strategy based on signal threshold and volume screening can accurately capture lesion features while eliminating noise interference. The final integrated segmentation result fully includes organ parenchyma, liquid areas, and tiny abnormal structures, achieving full-level coverage from macroscopic organs to microscopic lesions. It also significantly improves segmentation efficiency and objectivity through automated processes, providing an accurate and comprehensive imaging foundation for clinical diagnosis and quantitative analysis.

[0107] This method, which optimizes 3D mesh models by adjusting triangular facets using surface simplification coefficients, offers the dual advantages of accuracy and practicality. First, the initial model is generated based on the complete segmentation results (including organ parenchyma, fluid regions, and tiny low-signal structures), ensuring faithful restoration of 3D morphology and anatomical features, providing a reliable structural baseline for subsequent analysis. Second, by specifically adjusting the number of triangular facets using the surface simplification coefficient, redundant facets are reduced while preserving key morphological details (such as organ edges and tiny lesion outlines), significantly reducing the model data volume. This optimization not only ensures model clarity and spatial accuracy during visualization, but also improves its efficiency in interactive operations, data transmission, and subsequent applications (such as surgical simulation and 3D printing). It strikes a balance between accuracy and lightweightness, providing high-quality 3D structural support for scenarios such as clinical diagnosis and preoperative planning.

[0108] 2D layer verification ensures the accuracy of local model details at the tomographic level by strictly controlling the fit error between the segmented contour and the original image (≤1 pixel). 3D volume rendering overlay verification further identifies perforators or missing branches from the overall structure, ensuring the model fully restores the spatial form and anatomical continuity of the organ. This dual verification mechanism provides a high-quality 3D structural foundation for subsequent applications. Support for export in multiple formats, including STL, VTK, and DICOM-SEG, not only meets the needs of engineering applications such as 3D printing and surgical simulation, but also adapts to clinical and scientific research scenarios such as medical image archiving and cross-platform analysis, achieving the efficient transformation of high-precision models into practical value in multiple fields.

[0109] The present invention provides a device for acquiring three-dimensional models of abdominal organs. Through techniques such as a multi-region seed point strategy and adaptive threshold segmentation, it achieves full-level, precise extraction of organ parenchyma, fluid regions, and tiny, low-signal structures, balancing efficiency and integrity. Through steps such as boundary optimization and surface simplification, the model is lightweight while retaining key anatomical details, balancing accuracy and operational performance. A multi-dimensional verification mechanism (two-dimensional fit control and three-dimensional structural integrity inspection) ensures model authenticity at every level, while standardized multi-format export is suitable for diverse scenarios such as clinical diagnosis, surgical planning, and 3D printing. This overall process overcomes the inefficiency and subjective limitations of traditional manual segmentation while ensuring the reliability of the results through automated algorithms and rigorous verification. This provides a complete, efficient, accurate, and compatible solution for the image analysis and clinical application of abdominal organs.

[0110] It should be noted that the above description is limited to some embodiments of the present invention. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in an order different from that described in the above embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0111] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments, the present invention also provides a method for obtaining a three-dimensional model of an abdominal organ.

[0112] The method for obtaining a three-dimensional model of an abdominal organ comprises: Clamp the coil support assembly to the side of the examination bed; The user to be tested lies supine on the examination bed, and the abdominal coil is fixed above the abdomen by the coil support assembly, with the abdomen located at the center of the abdominal coil; The nuclear magnetic resonance apparatus is placed above the abdomen of the user to be tested, and the axis line of the positioning light is checked to be aligned with the axial midpoint of the coil support assembly, and the magnetic resonance scanning sequence is obtained by scanning with the nuclear magnetic resonance apparatus; An image processing device is in communication with the magnetic resonance imaging device to obtain the magnetic resonance scanning sequence and obtain a three-dimensional model of the abdominal organs according to the magnetic resonance scanning sequence; Wherein, acquiring a three-dimensional model of an abdominal organ according to the magnetic resonance imaging sequence includes: Obtaining an abdominal visceral organ tissue segmentation result according to the magnetic resonance imaging sequence and the abdominal visceral organ automatic image segmentation technology based on grayscale features; generating an initial abdominal organ tissue model according to the abdominal organ tissue segmentation result, and adjusting the triangular facets by using a surface simplification coefficient to optimize the initial abdominal organ tissue model to obtain an optimized three-dimensional mesh model; The optimized three-dimensional mesh model is multi-dimensionally verified and standardized and exported to obtain a three-dimensional model of the abdominal viscera.

[0113] Specifically, the user lies supine on the examination bed, with cotton balls or earplugs placed in both ears. The patient's feet are placed forward, with the body's long axis aligned with the bed's. The patient's hands are placed naturally at their sides, avoiding crossing their hands and feet to form a loop. An abdominal coil is used and secured above the abdomen using the coil support assembly. The abdomen is placed in the center of the coil. The internal spaces within the coil are filled with high-density sponge pads to minimize the occurrence of magnetic susceptibility artifacts.

[0114] The MRI scanner is placed above the abdomen of the user to be tested. The axial line of the check positioning light is aligned with the axial midpoint of the coil support assembly. The MRI scanner is used to scan and obtain a magnetic resonance scan sequence.

[0115] The image processing device is connected to the nuclear magnetic resonance apparatus for communication, and acquires a magnetic resonance scanning sequence and a three-dimensional model of the abdominal viscera according to the magnetic resonance scanning sequence.

[0116] The method of the above embodiment is used in a corresponding device for acquiring a three-dimensional model of an abdominal organ in any of the above embodiments, and has the beneficial effects of the corresponding device embodiment, which will not be described in detail here.

[0117] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present invention (including the claims) is limited to these examples. Within the scope of the present invention, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present invention as described above, which are not provided in detail for the sake of simplicity.

[0118] While specific details have been set forth to describe exemplary embodiments of the present invention, it will be apparent to those skilled in the art that embodiments of the present invention may be practiced without or with variations in these specific details. Therefore, these descriptions should be considered illustrative rather than restrictive. Although the present invention has been described in conjunction with specific embodiments thereof, many alternatives, modifications, and variations of these embodiments will be apparent to those skilled in the art in light of the foregoing description.

[0119] The embodiments of the present invention are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present invention should be included in the scope of protection of the present invention.

Claims

1. A device for acquiring a three-dimensional model of an abdominal organ, characterized in that: include: Coil support components, MRI machines, and image processing equipment; The coil support assembly is clamped and fixed to the side of the examination bed, and is used to support the abdominal coil to reduce pressure on the abdomen of the user to be tested; the nuclear magnetic resonance apparatus is placed above the abdomen of the user to be tested, and the axial line of the inspection positioning light is aligned with the axial midpoint of the coil support assembly, and the magnetic resonance scanning sequence is obtained by scanning with the nuclear magnetic resonance apparatus; the image processing device is communicatively connected to the nuclear magnetic resonance apparatus, obtains the magnetic resonance scanning sequence, and obtains a three-dimensional model of the abdominal organs according to the magnetic resonance scanning sequence; Wherein, acquiring a three-dimensional model of an abdominal organ according to the magnetic resonance imaging sequence includes: Obtaining an abdominal visceral organ tissue segmentation result according to the magnetic resonance imaging sequence and the abdominal visceral organ automatic image segmentation technology based on grayscale features; generating an initial abdominal organ tissue model according to the abdominal organ tissue segmentation result, and adjusting the triangular facets by using a surface simplification coefficient to optimize the initial abdominal organ tissue model to obtain an optimized three-dimensional mesh model; The optimized three-dimensional mesh model is multi-dimensionally verified and standardized and exported to obtain a three-dimensional model of the abdominal viscera.

2. The device for obtaining a three-dimensional model of an abdominal organ according to claim 1, characterized in that: The coil support assembly includes a side fixing bracket group, a fixing clamping mechanism is provided at the lower part of the side fixing bracket group, and the fixing clamping mechanism is used to clamp the side fixing bracket group to the side of the examination bed; a height adjustment mechanism is provided at the upper part of the side fixing bracket group, and a rotating shaft structure is provided at the top output end of the height adjustment mechanism, and the rotating shaft structure is connected to an arc-shaped support plate, and the arc-shaped support plate is rotatably installed at the top output end of the height adjustment mechanism through the rotating shaft structure; at least two abdominal coil binding through-holes are evenly provided on the arc-shaped support plate, and the abdominal coil binding through-holes are used to pass the binding belt of the abdominal coil.

3. The device for obtaining a three-dimensional model of an abdominal organ according to claim 1, characterized in that: The nuclear magnetic resonance apparatus is placed above the abdomen of the user to be tested, and the axis line of the inspection positioning light is aligned with the axis midpoint of the coil support assembly. The magnetic resonance scanning sequence obtained by scanning with the nuclear magnetic resonance apparatus includes: Step S1, using a fast positioning scan sequence to acquire sagittal, coronal, and transverse positioning images, and determining the scanning range of the main sequence based on the positioning images; Step S2: Define the main magnetic field direction as the Z axis, parallel to the long axis of the human body and pointing to the foot side, the X axis pointing to the right side of the human body, and the Y axis pointing to the front side of the human body; set the gradient field spatial encoding as the Z axis as the slice direction, and set the direction with the smaller value of the abdominal anteroposterior diameter thickness and the left and right diameter width as the phase encoding direction, and the larger value as the frequency encoding direction; Step S3, based on the gradient field spatial encoding, continuously exciting the scanning range with a small-angle radio frequency pulse with a flip angle of 30°-65°, and acquiring echo signals by rapidly switching the readout gradient field to obtain raw MR signal data; Step S4: Processing the raw MR signal data, using half-echo acquisition based on Fourier space conjugate symmetry, and reconstructing a magnetic resonance scanning sequence.

4. The device for acquiring a three-dimensional model of an abdominal organ according to claim 3, characterized in that: The step S3 further includes: A preparation pulse combination of 90°, 180°, and -90° is added before the main sequence.

5. The device for obtaining a three-dimensional model of an abdominal organ according to claim 3, characterized in that: The step S3 further includes: Preset currents are applied to the gradient coils in the X-axis, Y-axis, and Z-axis directions.

6. The device for acquiring a three-dimensional model of an abdominal organ according to claim 1, characterized in that: The abdominal viscera tissue segmentation result obtained according to the magnetic resonance scanning sequence and the abdominal viscera automatic image segmentation technology based on grayscale features includes: Setting seed points in the high signal area, low signal area and medium signal area of ​​the abdominal viscera respectively, and generating an initial mask covering the entire range of the abdominal viscera by using a seed region growing algorithm; performing a boundary shrinkage operation on the initial mask to remove the blood vessel wall and retain the pure parenchyma area to obtain the organ parenchyma; For abdominal organs containing fluid, the fluid area is extracted using an adaptive threshold segmentation algorithm; For intra-abdominal organs with tiny low signals, a signal threshold is obtained, and tiny low signal structures are extracted based on a volume screening condition and the signal threshold; The intra-abdominal organ tissue segmentation result includes the organ substance, the liquid area, and the tiny low-signal structure.

7. The device for obtaining a three-dimensional model of an abdominal organ according to claim 1, characterized in that: Generating an initial abdominal organ tissue model according to the abdominal organ tissue segmentation result, adjusting the triangular facets by using a surface simplification coefficient, and optimizing the initial abdominal organ tissue model to obtain an optimized three-dimensional mesh model includes: Adjust the triangle patch according to the following formula: ； in, Indicates the number of original mesh faces, Indicates the number of target faces after simplification, represents the simplification factor.

8. The device for obtaining a three-dimensional model of an abdominal organ according to claim 1, characterized in that: The multi-dimensional verification and standardized export of the optimized three-dimensional mesh model to obtain the three-dimensional model of the abdominal organs includes: Performing two-dimensional layer verification on the optimized three-dimensional mesh model, and obtaining a first verification model in response to a fit error between the segmented contour and the original image of ≤1 pixel; Superimposing the original image and the first verification model by volume rendering, and obtaining a three-dimensional model of the abdominal viscera in response to the absence or loss of perforators; The three-dimensional model export formats of the abdominal organs include STL format, VTK format and DICOM-SEG format.

9. A method for obtaining a three-dimensional model of an abdominal organ, characterized in that: A device for obtaining a three-dimensional model of an abdominal organ according to any one of claims 1 to 8, comprising: Clamp the coil support assembly to the side of the examination bed; The user to be tested lies supine on the examination bed, and the abdominal coil is fixed above the abdomen by the coil support assembly, with the abdomen located at the center of the abdominal coil; The nuclear magnetic resonance apparatus is placed above the abdomen of the user to be tested, and the axis line of the positioning light is checked to be aligned with the axial midpoint of the coil support assembly, and the magnetic resonance scanning sequence is obtained by scanning with the nuclear magnetic resonance apparatus; An image processing device is in communication with the magnetic resonance imaging device to obtain the magnetic resonance scanning sequence and obtain a three-dimensional model of the abdominal organs according to the magnetic resonance scanning sequence; Wherein, acquiring a three-dimensional model of an abdominal organ according to the magnetic resonance imaging sequence includes: Obtaining an abdominal visceral organ tissue segmentation result according to the magnetic resonance imaging sequence and the abdominal visceral organ automatic image segmentation technology based on grayscale features; generating an initial abdominal organ tissue model according to the abdominal organ tissue segmentation result, and adjusting the triangular facets by using a surface simplification coefficient to optimize the initial abdominal organ tissue model to obtain an optimized three-dimensional mesh model; The optimized three-dimensional mesh model is multi-dimensionally verified and standardized and exported to obtain a three-dimensional model of the abdominal viscera.

Images