A MRI-based placental location detection system and method

Abstract

The present application relates to the field of medical information technology, and particularly relates to a placenta position detection system and method based on MRI, the method comprising: obtaining target MRI image data, performing image segmentation processing on the target MRI image data, and extracting three-dimensional models of the placenta, uterus and cervix; based on the three-dimensional models, identifying the positional relationship between the placenta and the cervix, judging the placenta type, and evaluating the placenta implantation risk to generate risk warning information; displaying a three-dimensional visual model of the placenta and the uterus, and labeling a high-risk area on the three-dimensional visual model based on the risk warning information. The present application improves the placenta position detection efficiency, precision and automation, can realize full-process automatic detection, and reduces human error.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, specifically to an MRI-based placental location detection system and method. Background Technology

[0002] Placenta previa is a serious complication during pregnancy that threatens the safety of both mother and fetus. It is characterized by abnormal attachment of the placenta to the lower segment of the uterus or even covering the internal cervical os, which may lead to critical situations such as antepartum hemorrhage and fetal distress. Accurately determining the location of the placenta (e.g., complete, partial, or marginal placenta previa) is crucial for developing a clinical plan.

[0003] In recent years, with the continuous advancement of medical imaging technology, magnetic resonance imaging (MRI) has been increasingly applied to assist in the identification of placenta previa location. Compared to two-dimensional planar images, MRI, as a non-invasive, high-resolution imaging technology, can provide clearer soft tissue contrast, thus more accurately displaying the positional relationship between the placenta and the cervix. However, existing MRI detection methods still have significant shortcomings. On the one hand, traditional methods heavily rely on the experience and expertise of physicians, making the detection process highly subjective, especially in resource-limited primary hospitals where the error rate is high due to insufficient experience of technicians. On the other hand, existing MRI detection procedures are inefficient, requiring physicians to spend a significant amount of time observing and analyzing complex MRI images frame by frame, which is insufficient to meet clinical needs. Furthermore, existing technologies have significant deficiencies in automated risk assessment, lacking tools capable of automatically identifying placental location, morphology, and risk level, thus failing to provide effective support for precise surgical incision design.

[0004] Therefore, existing technologies face multiple bottlenecks in the detection of placenta previa, including efficiency, accuracy, and automation. There is an urgent need for a detection technology that integrates MRI three-dimensional reconstruction and intelligent analysis to achieve fully automated detection and reduce human error. Summary of the Invention

[0005] In view of this, the present invention provides an MRI-based placental location detection system and method to solve the multiple bottlenecks in efficiency, accuracy and automation faced by existing technologies in the detection of placenta previa.

[0006] In a first aspect, the present invention provides a placental location detection method based on MRI, the method comprising:

[0007] Acquire target MRI image data;

[0008] The target MRI image data is processed by image segmentation, and three-dimensional models of the placenta, uterus and cervix are extracted.

[0009] Based on the three-dimensional model, the positional relationship between the placenta and the cervix is ​​identified, the relationship between the placenta and the myometrium is determined, thereby determining the placental type and assessing the risk of placenta accreta to generate risk warning information.

[0010] A three-dimensional visualization model of the placenta and uterus is displayed, and high-risk areas are marked on the three-dimensional visualization model based on the risk warning information.

[0011] In one optional implementation, the target MRI image data is multi-sequence pelvic MRI image data, and before performing image segmentation processing on the target MRI image data, the method includes:

[0012] The target MRI image data is preprocessed, including motion artifact correction, noise reduction, multimodal non-rigid registration, and contrast enhancement.

[0013] In one optional implementation, the step of performing image segmentation processing on the target MRI image data and extracting a three-dimensional model of the placenta, uterus, and cervix includes:

[0014] The target MRI image data is segmented using a U-Net 3D segmentation network incorporating an attention mechanism, and the 3D models of the placenta, uterus, and cervix, along with segmentation confidence scores, are output.

[0015] In an optional implementation, the method further includes:

[0016] If the segmentation confidence level is lower than a preset threshold, a manual review mechanism is triggered to manually review the target MRI image data and the image segmentation processing results.

[0017] The results of manual verification are input into the U-Net 3D segmentation network, and the model parameters of the U-Net 3D segmentation network are optimized through incremental learning.

[0018] In one optional implementation, the step of identifying the positional relationship between the placenta and the cervix based on the three-dimensional model, determining the relationship between the placenta and the myometrium, thereby determining the placental type and assessing the risk of placenta accreta, includes:

[0019] Based on the three-dimensional model, the positional relationship between the placenta and the cervix was identified, and the placental attachment location, placental thickness, cervical morphology, cervical canal length, bridging vessels, placental vessel distribution density, and myometrial invasion depth were extracted.

[0020] The placental type is determined based on the positional relationship between the placenta and the cervix; the placental type includes complete placenta previa, partial placenta previa, and marginal placenta previa.

[0021] The placental type, placental attachment location, placental thickness, cervical morphology, cervical canal length, bridging vessels, placental vessel distribution density, and myometrial invasion depth are input as features into a pre-trained risk assessment model, and the probability assessment result of placental implantation risk is output; the placental implantation risk includes adhesive placenta, implanted placenta, and penetrating placenta.

[0022] In one optional implementation, generating risk warning information includes:

[0023] A risk heatmap is generated based on the probability assessment results, and a risk warning message is issued when the probability assessment results exceed a preset threshold; the risk warning message includes warning information on the coordinates of the cut operation avoidance.

[0024] In one optional implementation, the step of displaying a three-dimensional visualization model of the placenta and uterus, and marking high-risk areas on the three-dimensional visualization model based on the risk warning information, includes:

[0025] Based on the risk heatmap and the three-dimensional models of the placenta, uterus and cervix, a three-dimensional visualization model of the placenta and uterus is generated and displayed through a user interaction module.

[0026] Based on the risk warning information, high-risk areas are marked on the three-dimensional visualization model, and a detection report is generated; the detection report includes placental location and type, risk level, image markings, and incision operation planning suggestions.

[0027] Secondly, the present invention provides an MRI-based placental location detection system, the system being used to perform an MRI-based placental location detection method as described above, the system comprising:

[0028] The image processing module is used to acquire target MRI image data, perform image segmentation processing on the target MRI image data, and extract three-dimensional models of the placenta, uterus, and cervix.

[0029] The intelligent analysis module is used to identify the positional relationship between the placenta and the cervix, determine the relationship between the placenta and the myometrium, and thus determine the placental type based on the three-dimensional model.

[0030] The risk warning module is used to assess the risk of placenta accreta and generate risk warning information.

[0031] The user interaction module is used to display a three-dimensional visualization model of the placenta and uterus, and to mark high-risk areas on the three-dimensional visualization model based on the risk warning information.

[0032] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to perform an MRI-based placental location detection method according to the first aspect or any corresponding embodiment described above.

[0033] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform an MRI-based placental location detection method according to the first aspect or any corresponding embodiment described above.

[0034] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute an MRI-based placental location detection method according to the first aspect or any corresponding embodiment described above.

[0035] The technical solution provided by this invention may include the following beneficial effects:

[0036] 1. This invention utilizes multimodal data fusion technology to achieve millimeter-level 3D modeling, accurately locating the distance between the lower edge of the placenta and the internal cervical os. Combined with an intelligent segmentation algorithm, it can automatically distinguish between complete, partial, and marginal placenta previa, significantly reducing the risk of misdiagnosis due to insufficient experience. This high-precision detection method is particularly suitable for primary care hospitals, effectively improving the objectivity and accuracy of diagnosis.

[0037] 2. This invention employs a fully automated design, supporting parallel processing optimization from image input to the generation of 3D models and detection reports. It can complete image segmentation and 3D reconstruction in real time. This efficient processing method significantly improves system response speed, meets clinical operational needs, and greatly reduces the workload of doctors.

[0038] 3. Based on the density of placental blood vessel distribution and the depth of uterine myometrial invasion, this invention can predict the risk of placenta accreta and automatically issue warnings and mark high-risk bleeding areas. Combining a three-dimensional placental model with uterine anatomy, it recommends cesarean section incision coordinates to avoid areas covered by the placenta.

[0039] 4. This invention supports VR / AR panoramic visualization. Users can manipulate the 3D model with gestures to dynamically observe the spatial relationship between the placenta and the uterus, and simulate the effects of different surgical approaches. This can help users better observe and analyze the placental condition and also improve the accuracy of detection.

[0040] 5. This invention supports cloud and edge computing, allowing primary hospitals to obtain diagnostic support at the level of tertiary hospitals simply by uploading image data. Simultaneously, the intelligent analysis module has continuous update capabilities, adapting to the diversity of ethnicities, gestational ages, and pathological characteristics across different regions. This not only reduces the misdiagnosis rate in primary hospitals but also promotes the widespread coverage of high-quality medical resources. Attached Figure Description

[0041] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0042] Figure 1 This is a schematic diagram of a placental location detection system based on MRI according to an embodiment of the present invention;

[0043] Figure 2 This is a flowchart of a placental location detection method based on MRI according to an embodiment of the present invention;

[0044] Figure 3 This is a flowchart of another MRI-based placental location detection method according to an embodiment of the present invention;

[0045] Figure 4 This is a flowchart of another MRI-based placental location detection method according to an embodiment of the present invention;

[0046] Figure 5 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] According to an embodiment of the present invention, an MRI-based placental location detection system is provided. Figure 1 This is a schematic diagram of a placental location detection system based on MRI according to an embodiment of the present invention. This system is used to perform tasks such as... Figure 2 In the MRI-based placental location detection method shown, such as Figure 1 As shown, the system includes:

[0049] Image processing module 01 is used to acquire target MRI image data, perform image segmentation processing on the target MRI image data, and extract three-dimensional models of the placenta, uterus and cervix.

[0050] Intelligent analysis module 02 is used to identify the positional relationship between the placenta and the cervix and to determine the relationship between the placenta and the myometrium based on the three-dimensional model, thereby determining the placenta type;

[0051] Risk warning module 03 is used to assess the risk of placenta accreta and generate risk warning information;

[0052] User interaction module 04 is used to display a three-dimensional visualization model of the placenta and uterus, and to mark high-risk areas on the three-dimensional visualization model based on the risk warning information.

[0053] Furthermore, the image processing module 01 is used to receive image data (i.e., target MRI image data) acquired by the MRI equipment, and extract three-dimensional models of key anatomical structures such as the placenta, uterus, and cervix through image segmentation algorithms. By receiving target MRI image data, applying segmentation algorithms, and reconstructing three-dimensional models, the image processing module provides strong technical support for the accurate detection of placenta previa, significantly reducing the time doctors spend manually analyzing images and improving detection efficiency.

[0054] The Intelligent Analysis Module 02 is the core of the system. Its function is to automatically identify the positional relationship between the placenta and cervix based on deep learning algorithms, determine the placental type (e.g., complete, partial, marginal placenta previa), and assess the risk of placenta accreta. By optimizing the deep learning algorithm, the Intelligent Analysis Module significantly reduces the risk of misjudgment due to human factors, especially in resource-constrained primary hospitals. The module employs advanced deep learning algorithms to automatically learn features and patterns in target MRI image data, thereby achieving accurate identification of placental location and morphology. Furthermore, the Intelligent Analysis Module introduces an attention mechanism to further enhance the identification ability of key areas (such as the placental-cervix junction). This mechanism automatically focuses on important features in the image, thereby improving detection accuracy.

[0055] The risk warning module 03 analyzes parameters such as the distance between the placenta and cervix, placental thickness, and placental vascular distribution to generate risk warning information, providing support for clinical decision-making. This module reduces the decision-making burden on physicians in complex cases, especially in primary care hospitals where limited resources and insufficient technical personnel experience lead to higher misjudgment rates. The application of the intelligent analysis module can significantly improve detection efficiency and accuracy.

[0056] User interaction module 04 is used to support panoramic observation of the placenta-uterus spatial relationship in VR / AR mode, display a three-dimensional visualization model of the placenta and uterus, mark key anatomical structures and risk areas, and provide test reports.

[0057] In summary, the technical solution provided in this embodiment can include the following beneficial effects:

[0058] 1. This embodiment utilizes multimodal data fusion technology to achieve millimeter-level 3D modeling, accurately locating the distance between the lower edge of the placenta and the internal cervical os. Combined with intelligent segmentation algorithms, it can automatically distinguish between complete, partial, and marginal placenta previa, significantly reducing the risk of misdiagnosis due to insufficient experience. This high-precision detection method is particularly suitable for primary care hospitals, effectively improving the objectivity and accuracy of diagnosis.

[0059] 2. This embodiment adopts a fully automated design, from image input to the generation of 3D models and detection reports. The entire process supports parallel processing optimization and can complete image segmentation and 3D reconstruction in real time. This efficient processing method significantly improves the system response speed, meets clinical operation needs, and greatly reduces the workload of doctors.

[0060] 3. This embodiment, based on placental vascular distribution density and uterine myometrial invasion depth, can predict the risk of placenta accreta and automatically issue warnings and mark high-risk bleeding areas. Combining a three-dimensional placental model with uterine anatomy, it recommends cesarean section incision coordinates to avoid areas covered by the placenta.

[0061] 4. This embodiment supports VR / AR panoramic visualization. Users can manipulate the 3D model with gestures to dynamically observe the spatial relationship between the placenta and the uterus, and simulate the effects of different surgical approaches. This can help users better observe and analyze the placental condition and also improve the accuracy of detection.

[0062] 5. This embodiment supports cloud and edge computing. Primary hospitals only need to upload image data to obtain diagnostic support at the level of tertiary hospitals. Simultaneously, the intelligent analysis module has continuous update capabilities, adapting to the diversity of ethnicities, gestational ages, and pathological characteristics across different regions. This not only reduces the misdiagnosis rate in primary hospitals but also promotes the widespread coverage of high-quality medical resources.

[0063] According to an embodiment of the present invention, an embodiment of a placental location detection method based on MRI is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0064] This embodiment provides an MRI-based method for placental location detection, which can be applied to... Figure 1 The illustrated system is an MRI-based placental location detection system. Figure 2 This is a flowchart of a placental location detection method based on MRI according to an embodiment of the present invention, as shown below. Figure 2 As shown, the process includes the following steps:

[0065] Step S201: Acquire target MRI image data.

[0066] Furthermore, this embodiment scans the pregnant woman's pelvic region to acquire multi-sequence MRI images (such as T1-weighted images, T2-weighted images, and fat-suppressed sequences), covering areas such as the placental attachment site, the internal cervical os, and the myometrium, thereby obtaining target MRI image data. Multi-sequence MRI images can provide different soft tissue contrasts (e.g., T2-weighted images highlight the boundary between the placenta and the uterus, and vascular enhancement sequences show the distribution of placental vessels), providing multi-dimensional information for subsequent 3D modeling and risk assessment.

[0067] Step S202: Perform image segmentation processing on the target MRI image data and extract the three-dimensional models of the placenta, uterus and cervix.

[0068] Furthermore, this embodiment can first preprocess the target MRI image data, such as performing motion artifact correction (eliminating image blurring caused by the pregnant woman's breathing and fetal movement) and noise filtering (improving tissue boundary clarity). Then, multimodal non-rigid registration is used to align different sequence images to the same coordinate system to compensate for differences in organ deformation (such as the elastic changes of the uterus with gestational age). After preprocessing, this embodiment can use a U-Net 3D segmentation network with an attention mechanism to segment the target tissue voxel by voxel through an encoder-decoder structure, generating 3D mesh models of the placenta, uterus, and cervix. This transforms 2D slice images into quantifiable 3D structures, providing geometric and spatial data support for subsequent positional relationship analysis and risk assessment.

[0069] Step S203: Based on the three-dimensional model, identify the positional relationship between the placenta and the cervix, determine the relationship between the placenta and the myometrium, thereby determining the placental type and assessing the risk of placental implantation, so as to generate risk warning information.

[0070] Furthermore, this embodiment extracts placental thickness, vascular distribution density (quantified by vascular enhancement sequences of the number of vessels per unit volume), and myometrial invasion depth based on a three-dimensional model. These parameters are then input into a pre-trained risk assessment model (such as a deep learning classifier), which outputs the probabilities of adhesive, implanted, and penetrating placentas, thus obtaining risk warning information. This embodiment replaces manual experience-based judgment with automated algorithms, achieving standardized classification of placental types and quantitative assessment of implantation risk, reducing subjective errors.

[0071] Step S204: Display a three-dimensional visualization model of the placenta and uterus, and mark high-risk areas on the three-dimensional visualization model based on the risk warning information.

[0072] Furthermore, this embodiment utilizes VR / AR technology or specialized medical software to reconstruct a three-dimensional model of the placenta-uterus, supporting gesture-based controls (such as rotation and sectional viewing) for dynamic observation of spatial relationships (e.g., how the placenta covers the cervix and its location within the myometrium). A risk heatmap is integrated with the 3D model, and warning labels are marked in areas where the placenta covers the internal cervical os, areas with dense blood vessels, or sites of myometrial infiltration. This embodiment enhances the physician's understanding of complex anatomical structures through immersive interaction and optimizes surgical plans by combining quantitative risk information.

[0073] In summary, the technical solution provided in this embodiment can include the following beneficial effects:

[0074] 1. This embodiment utilizes multimodal data fusion technology to achieve millimeter-level 3D modeling, accurately locating the distance between the lower edge of the placenta and the internal cervical os. Combined with intelligent segmentation algorithms, it can automatically distinguish between complete, partial, and marginal placenta previa, significantly reducing the risk of misdiagnosis due to insufficient experience. This high-precision detection method is particularly suitable for primary care hospitals, effectively improving the objectivity and accuracy of diagnosis.

[0075] 2. This embodiment adopts a fully automated design, from image input to the generation of 3D models and detection reports. The entire process supports parallel processing optimization and can complete image segmentation and 3D reconstruction in real time. This efficient processing method significantly improves the system response speed, meets clinical operation needs, and greatly reduces the workload of doctors.

[0076] 3. This embodiment, based on placental vascular distribution density and uterine myometrial invasion depth, can predict the risk of placenta accreta and automatically issue warnings and mark high-risk bleeding areas. Combining a three-dimensional placental model with uterine anatomy, it recommends cesarean section incision coordinates to avoid areas covered by the placenta.

[0077] 4. This embodiment supports VR / AR panoramic visualization. Users can manipulate the 3D model with gestures to dynamically observe the spatial relationship between the placenta and the uterus, and simulate the effects of different surgical approaches. This can help users better observe and analyze the placental condition and also improve the accuracy of detection.

[0078] 5. This embodiment supports cloud and edge computing. Primary hospitals only need to upload image data to obtain diagnostic support at the level of tertiary hospitals. Simultaneously, the intelligent analysis module has continuous update capabilities, adapting to the diversity of ethnicities, gestational ages, and pathological characteristics across different regions. This not only reduces the misdiagnosis rate in primary hospitals but also promotes the widespread coverage of high-quality medical resources.

[0079] This embodiment provides an MRI-based method for placental location detection, which can be applied to... Figure 1 The illustrated system is an MRI-based placental location detection system. Figure 3 This is a flowchart of another MRI-based placental location detection method according to an embodiment of the present invention, such as... Figure 3 As shown, the process includes the following steps:

[0080] Step S301: Acquire target MRI image data. This target MRI image data is multi-sequence pelvic MRI image data.

[0081] Furthermore, this embodiment acquires multi-sequence pelvic MRI image data (such as T1-weighted images, T2-weighted images, fat-suppressed sequences, and vascular enhancement sequences) of the pregnant woman's pelvic region, covering key anatomical structures such as the uterus, placenta, and cervix. Compared to single-sequence data, multi-sequence data can more comprehensively reflect placental morphology (such as thickness and edge clarity), blood supply characteristics, and spatial relationship with surrounding tissues, which is the basis for accurate detection.

[0082] Step S302: Preprocess the target MRI image data, including motion artifact correction, noise reduction, multimodal non-rigid registration, and contrast enhancement.

[0083] Furthermore, motion artifact correction aims to eliminate image blurring caused by the pregnant woman's breathing movements and fetal activity, ensuring clear boundaries of anatomical structures. Noise reduction aims to suppress random noise during MRI acquisition and improve tissue contrast. Multimodal non-rigid registration aims to align MRI images from different sequences to the same spatial coordinate system, compensating for nonlinear organ deformation. Contrast enhancement aims to highlight the boundaries of the placenta, uterus, and cervix, facilitating segmentation algorithm recognition. These preprocessing steps in this embodiment provide high-quality input to the subsequent segmentation network, reducing segmentation errors caused by image noise or registration errors, and improving the accuracy of 3D model reconstruction.

[0084] For further details, please see Figure 4 The flowchart of another MRI-based placental location detection method is shown. First, multi-sequence pelvic MRI image data, including T1-weighted and T2-weighted multimodal sequences, is acquired in the image processing module 01. Then, motion artifact correction is performed to remove artifacts caused by the movement of the pregnant woman or fetus during the scanning process, ensuring image quality. Next, image noise filtering is used to remove random noise generated during MRI acquisition, highlighting important anatomical structures. Finally, multimodal Attention U-Net segmentation processing is applied, using a U-Net 3D segmentation network incorporating attention mechanisms to segment the preprocessed MRI image data, extracting 3D models of the placenta, uterus, and cervix. The 3D models of the placenta, uterus, and cervix are output, providing a foundation for subsequent analysis.

[0085] Step S303: The target MRI image data is segmented using a U-Net three-dimensional segmentation network that includes an attention mechanism, and the three-dimensional models of the placenta, uterus and cervix and the segmentation confidence are output.

[0086] Furthermore, this embodiment uses a U-Net 3D segmentation network (Attention U-Net / 3D) incorporating an attention mechanism. Its network structure is as follows: the encoder extracts multi-scale features (such as the overall morphology of the placenta and local details of the cervix), and the decoder dynamically focuses on key regions (such as the blurred boundary at the junction of the placenta and cervix) through the attention mechanism, suppressing interference from irrelevant tissues such as fat and bladder. Finally, a 3D mesh model of the placenta, uterus, and cervix is ​​generated, and segmentation confidence (such as a Softmax probability value, reflecting the probability that each voxel belongs to the target tissue) is output.

[0087] In one optional implementation, if the segmentation confidence level is lower than a preset threshold, a manual review mechanism is triggered to manually review the target MRI image data and the image segmentation processing results.

[0088] The results of manual verification are input into the U-Net 3D segmentation network, and the model parameters of the U-Net 3D segmentation network are optimized through incremental learning.

[0089] Furthermore, in this embodiment, if the segmentation confidence level is less than a preset threshold (e.g., 90%), uncertain regions (e.g., blurred placental edges) can be automatically marked, triggering a manual review process where doctors manually correct the segmentation results (e.g., outlining blurred boundaries). This embodiment uses the manually reviewed labeled data as training samples, updating the model parameters of the U-Net 3D segmentation network through transfer learning.

[0090] Furthermore, such as Figure 4 As shown, in the intelligent analysis module 02, this embodiment aligns multimodal MRI images to the same coordinate system through multimodal non-rigid registration processing, allowing local and nonlinear deformations to compensate for differences between different imaging modalities; and enhances the placental edges through algorithms to make them clearer and facilitate subsequent analysis; and marks areas where there may be uncertainty during segmentation to remind doctors to pay close attention. The segmentation confidence level is determined as follows: if the segmentation confidence level is below 90%, a manual review mechanism is triggered, and the doctor checks the segmentation results. If the segmentation confidence level is above or equal to 90%, the placental type (complete, partial, or marginal placenta previa) is determined based on the segmentation results.

[0091] Step S304: Based on the three-dimensional model, identify the positional relationship between the placenta and the cervix, determine the relationship between the placenta and the myometrium, thereby determining the placental type and assessing the risk of placental implantation, so as to generate risk warning information.

[0092] In one optional implementation, step S304 includes:

[0093] Based on this 3D model, the positional relationship between the placenta and the cervix was identified, and the placental attachment location, placental thickness, cervical morphology, cervical canal length, bridging vessels, placental vessel distribution density, and myometrial invasion depth were extracted.

[0094] The placental type is determined based on the positional relationship between the placenta and the cervix; the placental type includes complete placenta previa, partial placenta previa, and marginal placenta previa.

[0095] The placental type, placental attachment location, placental thickness, cervical morphology, cervical canal length, bridging vessels, placental vascular distribution density, and myometrial invasion depth are input features into a pre-trained risk assessment model, which outputs a probability assessment result of placental implantation risk. The placental implantation risk includes adhesion-type placenta, implanted placenta, and penetrating placenta.

[0096] A risk heatmap is generated based on the probability assessment result, and a risk warning is issued when the probability assessment result exceeds a preset threshold; the risk warning information includes warning information on the coordinates of the cut operation avoidance.

[0097] Furthermore, this embodiment calculates the minimum distance between the lower edge of the placenta and the internal cervical os, and classifies placental types based on the proportion of three-dimensional coverage area. Placental thickness can be obtained by measuring the maximum thickness on the cross-section of the three-dimensional model. Vascular density can be obtained by calculating the number of vascular branches per unit volume based on MRI dynamic contrast-enhanced (DCE) sequences using a vascular segmentation algorithm. The depth of myometrial invasion can be obtained by assessing the depth of placental signal intrusion into the myometrium in T2-weighted images.

[0098] The risk assessment model can be a pre-trained XGBoost ensemble model or a 3D ResNet network. Its input is the aforementioned feature vector, and its output is the probability distribution of adhesive, implanted, and penetrating placenta. In this embodiment, based on the probability distribution, a color gradient heatmap, i.e., a risk heatmap, is generated through Gaussian kernel density estimation (KDE). For example, in this risk heatmap, red represents high risk; orange represents medium risk; yellow represents low risk; and blue represents a safe zone. Furthermore, adaptive color display can be implemented according to actual needs.

[0099] Furthermore, such as Figure 4As shown, in the risk warning module 03, this embodiment extracts a feature map of the placenta-uterus interface, including placental thickness, placental vascular distribution density, and myometrial invasion depth. This feature map is input into a pre-trained risk assessment model to calculate the probability of placenta accreta. Based on the probability assessment results, a risk heatmap is generated, visually displaying the risk levels in different areas. Finally, the risk heatmap is combined with a three-dimensional model of the placenta and uterus to generate a three-dimensional visualization model, and a risk level is generated based on the risk assessment results, providing decision support for doctors.

[0100] Step S305: Based on the risk heat map and the three-dimensional models of the placenta, uterus and cervix, generate a three-dimensional visualization model of the placenta and uterus, and display the three-dimensional visualization model of the placenta and uterus through the user interaction module.

[0101] Furthermore, this embodiment integrates risk heatmaps with three-dimensional models of the placenta, uterus, and cervix, and generates an interactive 3D scene, namely a three-dimensional visualization model of the placenta and uterus, through the VTK or ITK visualization library. The three-dimensional visualization model of the placenta and uterus supports VR / AR interaction (such as gesture rotation, zoom, and sectioning), and doctors can observe the spatial relationship between the placenta and cervix from any angle (such as whether the placenta wraps around the cervix from behind).

[0102] Step S306: Based on the risk warning information, mark high-risk areas on the three-dimensional visualization model and generate a detection report; the detection report includes placental location and type, risk level, image markings, and incision operation planning suggestions.

[0103] Furthermore, this embodiment overlays warning labels onto the 3D visualization model to mark high-risk areas. The generated test report includes placental type (e.g., "partial placenta previa"), risk level (e.g., "high-risk accreta"), key image markers (e.g., screenshot of the internal cervical os location), and surgical planning recommendations (e.g., recommending the coordinates of 3 cesarean section incisions that avoid the placental coverage area, along with the probability of bleeding risk).

[0104] Furthermore, such as Figure 4 As shown, in the user interaction module 04, doctors can view the 3D visualization model and correct the results. If the result differs from the doctor's judgment by more than 15%, a consultation process is triggered to ensure the accuracy of the diagnosis.

[0105] In summary, the technical solution provided in this embodiment can include the following beneficial effects:

[0106] 1. This embodiment utilizes multimodal data fusion technology to achieve millimeter-level 3D modeling, accurately locating the distance between the lower edge of the placenta and the internal cervical os. Combined with intelligent segmentation algorithms, it can automatically distinguish between complete, partial, and marginal placenta previa, significantly reducing the risk of misdiagnosis due to insufficient experience. This high-precision detection method is particularly suitable for primary care hospitals, effectively improving the objectivity and accuracy of diagnosis.

[0107] 2. This embodiment adopts a fully automated design, from image input to the generation of 3D models and detection reports. The entire process supports parallel processing optimization and can complete image segmentation and 3D reconstruction in real time. This efficient processing method significantly improves the system response speed, meets clinical operation needs, and greatly reduces the workload of doctors.

[0108] 3. This embodiment, based on placental vascular distribution density and uterine myometrial invasion depth, can predict the risk of placenta accreta and automatically issue warnings and mark high-risk bleeding areas. Combining a three-dimensional placental model with uterine anatomy, it recommends cesarean section incision coordinates to avoid areas covered by the placenta.

[0109] 4. This embodiment supports VR / AR panoramic visualization. Users can manipulate the 3D model with gestures to dynamically observe the spatial relationship between the placenta and the uterus, and simulate the effects of different surgical approaches. This can help users better observe and analyze the placental condition and also improve the accuracy of detection.

[0110] 5. This embodiment supports cloud and edge computing. Primary hospitals only need to upload image data to obtain diagnostic support at the level of tertiary hospitals. Simultaneously, the intelligent analysis module has continuous update capabilities, adapting to the diversity of ethnicities, gestational ages, and pathological characteristics across different regions. This not only reduces the misdiagnosis rate in primary hospitals but also promotes the widespread coverage of high-quality medical resources.

[0111] This invention also provides a computer device; please refer to [link / reference]. Figure 5 , Figure 5 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 5As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 5 Take a processor 10 as an example.

[0112] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0113] The memory 20 stores instructions executable by at least one processor 10 to cause at least one processor 10 to perform the method shown in the above embodiments.

[0114] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0115] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0116] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0117] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0118] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0119] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and all such modifications and variations fall within the defined scope.

Claims

1. A method for placental location detection based on MRI, characterized in that, The method includes: Acquire target MRI image data; The target MRI image data is processed by image segmentation, and three-dimensional models of the placenta, uterus and cervix are extracted. Based on the three-dimensional model, the positional relationship between the placenta and the cervix is ​​identified, the relationship between the placenta and the myometrium is determined, thereby determining the placental type and assessing the risk of placenta accreta to generate risk warning information. Display a three-dimensional visualization model of the placenta and uterus, and mark high-risk areas on the three-dimensional visualization model based on the risk warning information; Based on the aforementioned three-dimensional model, the positional relationship between the placenta and cervix is ​​identified, the relationship between the placenta and the myometrium is determined, thereby identifying the placental type and assessing the risk of placenta accreta, including: Based on the three-dimensional model, the positional relationship between the placenta and the cervix was identified, and the placental attachment location, placental thickness, cervical morphology, cervical canal length, bridging vessels, placental vessel distribution density, and myometrial invasion depth were extracted. The placental type is determined based on the positional relationship between the placenta and the cervix; the placental type includes complete placenta previa, partial placenta previa, and marginal placenta previa. The placental type, placental attachment location, placental thickness, cervical morphology, cervical canal length, bridging vessels, placental vessel distribution density, and myometrial invasion depth are input as features into a pre-trained risk assessment model, which outputs a probability assessment result for placental implantation risk. The placental implantation risk includes adhesion-type placenta, implanted placenta, and penetrating placenta. The generation of risk warning information includes: Based on the probability assessment results, a color gradient heatmap is generated by Gaussian kernel density estimation, and a risk warning message is issued when the probability assessment results exceed a preset threshold; the risk warning message includes a warning message for the coordinates of the cut operation avoidance.

2. The method according to claim 1, characterized in that, The target MRI image data is multi-sequence pelvic MRI image data. Before performing image segmentation processing on the target MRI image data, the method includes: The target MRI image data is preprocessed, including motion artifact correction, noise reduction, multimodal non-rigid registration, and contrast enhancement.

3. The method according to claim 1, characterized in that, The step of performing image segmentation processing on the target MRI image data and extracting three-dimensional models of the placenta, uterus, and cervix includes: The target MRI image data is segmented using a U-Net 3D segmentation network incorporating an attention mechanism, and the 3D models of the placenta, uterus, and cervix, along with segmentation confidence scores, are output.

4. The method according to claim 3, characterized in that, The method further includes: If the segmentation confidence level is lower than a preset threshold, a manual review mechanism is triggered to manually review the target MRI image data and the image segmentation processing results. The results of manual verification are input into the U-Net 3D segmentation network, and the model parameters of the U-Net 3D segmentation network are optimized through incremental learning.

5. The method according to claim 1, characterized in that, The three-dimensional visualization model displaying the placenta and uterus, and marking high-risk areas on the three-dimensional visualization model based on the risk warning information, includes: Based on the color gradient heatmap and the three-dimensional models of the placenta, uterus and cervix, a three-dimensional visualization model of the placenta and uterus is generated and displayed through a user interaction module. Based on the risk warning information, high-risk areas are marked on the three-dimensional visualization model, and a detection report is generated; the detection report includes placental location and type, risk level, image markings, and incision operation planning suggestions.

6. A placental location detection system based on MRI, characterized in that, The system is used to perform the placental location detection method based on MRI according to any one of claims 1 to 5, the system comprising: The image processing module is used to acquire target MRI image data, perform image segmentation processing on the target MRI image data, and extract three-dimensional models of the placenta, uterus, and cervix. The intelligent analysis module is used to identify the positional relationship between the placenta and the cervix and to determine the relationship between the placenta and the myometrium based on the three-dimensional model, thereby determining the placental type. The risk warning module is used to assess the risk of placenta accreta and generate risk warning information. The user interaction module is used to display a three-dimensional visualization model of the placenta and uterus, and to mark high-risk areas on the three-dimensional visualization model based on the risk warning information. The intelligent analysis module is also used for: Based on the three-dimensional model, the positional relationship between the placenta and the cervix is ​​identified, and the placental attachment location, placental thickness, cervical morphology, cervical canal length, bridging vessels, placental vessel distribution density, and myometrial invasion depth are extracted. Based on the positional relationship between the placenta and the cervix, the placental type is determined. The placental type includes complete placenta previa, partial placenta previa, and marginal placenta previa. The risk warning module is also used for: The placental type, placental attachment location, placental thickness, cervical morphology, cervical canal length, bridging vessels, placental vessel distribution density, and myometrial invasion depth are input as features into a pre-trained risk assessment model, which outputs a probability assessment result for placental implantation risk. The placental implantation risk includes adhesion-type placenta, implanted placenta, and penetrating placenta. Based on the probability assessment results, a color gradient heatmap is generated by Gaussian kernel density estimation, and a risk warning message is issued when the probability assessment results exceed a preset threshold; the risk warning message includes a warning message for the coordinates of the cut operation avoidance.

7. A computer device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the MRI-based placental location detection method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform any one of claims 1 to 5, a method for placental location detection based on MRI.

Images