A method and system for generating dynamic visualization of a fetus based on 2D ultrasound data

Abstract

The application discloses a kind of fetal dynamic visualization generation method and system based on 2D ultrasonic data, first to the fetal two-dimensional ultrasonic gray data after desensitization processing is carried out structure feature extraction, construct adaptive structure constraint model suitable for different gestational age;Then establish articulated drivable dynamic model, and generate time-continuous fetal action sequence in combination with timing consistency constraint rule, after interframe smoothing optimization processing, output stable structure, visual coherent three-dimensional animation result.The animation excludes clinical diagnosis, health assessment, growth quantification and other medical purposes, only applicable to family memorial, popular science education and entertainment interaction class non-medical visualization scene.This method realizes stable dynamic visualization based on two-dimensional ultrasonic data without three-dimensional ultrasonic acquisition equipment, with adjustable parameters, dynamic controllable, playback repeatable and data processing full-process compliance and other characteristics.

Description

Technical Field

[0001] This invention relates to the fields of medical image processing and computer graphics technology, specifically to a method and system for generating dynamic visualizations of the fetus based on two-dimensional ultrasound grayscale image data. This technology is primarily geared towards non-medical applications, achieving the generation and display of three-dimensional fetal animation through structural modeling and dynamic driving processing of two-dimensional ultrasound data. Background Technology

[0002] In recent years, China has made progress in hardware integration, miniaturization, and wireless data transmission for portable handheld ultrasound devices (hereinafter referred to as handheld ultrasound devices). Their application scenarios are gradually expanding from traditional medical institutions to primary healthcare units, maternity care facilities, and some home environments. Against this backdrop, non-professional operators (such as expectant parents and related professionals) can also acquire two-dimensional grayscale ultrasound data of the fetus in utero under certain conditions, thus demonstrating a trend of ultrasound data acquisition expanding from being primarily driven by professional medical needs to encompassing multiple scenarios.

[0003] From the current industry development perspective, handheld ultrasound devices are still primarily geared towards medical diagnostic needs, with their image processing and display methods largely focused on medical interpretation. In non-medical scenarios, relevant data is usually presented as raw two-dimensional grayscale images, which have a relatively simple visualization format. For non-professional users, two-dimensional grayscale images have certain limitations in terms of spatial representation and structural understanding, making it difficult to intuitively reflect the three-dimensional morphology and dynamic changes of the fetus.

[0004] Meanwhile, as application scenarios expand, users' needs for displaying the fetal state in utero have gradually shifted from acquiring basic information to requiring more intuitive and interactive visualizations. For example, in home use scenarios, users prefer to understand the fetal state through clear shapes and dynamic continuity, while traditional two-dimensional images are somewhat inadequate in this type of application.

[0005] In existing technologies, clinically common three-dimensional and four-dimensional ultrasound imaging methods typically rely on dedicated three-dimensional ultrasound probes to acquire volumetric data, and then combine volume rendering or surface reconstruction algorithms to reconstruct the three-dimensional morphology of the fetus. While these methods offer advantages in structural accuracy, they require significant hardware and computing resources, usually necessitating the use of dedicated ultrasound hosts and image processing systems. Furthermore, the acquisition and processing processes are largely performed by professionals, thus limiting their applicability in resource-constrained mobile devices or home settings.

[0006] For non-medical applications, some existing technical solutions use pre-set 3D models or animation templates for visualization. While this lowers the barrier to entry to some extent, some publicly available technologies lack a stable structural mapping between the model and actual 2D ultrasound data, potentially leading to insufficient structural matching during model generation and display. Furthermore, in terms of dynamic generation, some solutions employ template-based looping animations or simple frame sequence splicing, lacking systematic constraints on cross-frame state changes, which may result in insufficient continuity of the animation over time.

[0007] Furthermore, at the data processing level, some technical solutions for non-medical applications do not pay enough attention to the privacy attributes of ultrasound data and have not formed a complete data desensitization, encrypted storage and access control mechanism, which may pose certain data security risks in practical applications.

[0008] In summary, existing technologies still have certain gaps in their compatibility between high-precision clinical imaging solutions and lightweight visualization solutions for civilian use. In particular, there is still room for optimization in achieving 3D visualization based on 2D ultrasound data that exhibits high structural consistency, continuous dynamic processes, and a certain degree of adjustability, without relying on 3D ultrasound acquisition equipment and high-performance computing platforms. Therefore, it is necessary to propose a new technical solution to improve the visualization capabilities of 2D ultrasound data in non-medical scenarios. Summary of the Invention

[0009] This invention aims to address the problems of insufficient visualization capabilities of 2D ultrasound data in non-medical scenarios, reliance on specialized equipment for 3D reconstruction, limited continuity of dynamic animation, and lack of data privacy protection in existing technologies, and provides a method and system for generating dynamic visualizations of fetuses based on 2D ultrasound data.

[0010] This invention enables the conversion of two-dimensional ultrasound grayscale data into structurally stable, temporally continuous, and dynamically controllable three-dimensional animation without the need for a three-dimensional ultrasound probe and a high-performance computing platform. This allows for the fulfillment of non-medical visualization needs such as family pregnancy commemoration and parent-child entertainment interaction, while ensuring the safety and compliance of data processing.

[0011] This invention is based on a structure constraint-driven dynamic visualization technology path using two-dimensional ultrasound data. It takes the desensitized ultrasound image as input and sequentially completes data preprocessing, key structural feature extraction and adaptive modeling, three-dimensional dynamic model construction, dual-constraint dynamic generation and smooth optimization output to form a structurally faithful, temporally coherent and controllable three-dimensional animation.

[0012] It is worth noting that the generation of two-dimensional structural models into three-dimensional space can be achieved using spatial algorithms, including but not limited to traditional geometric mapping, spatial interpolation algorithms (such as thin plate splines and radial basis functions), nonlinear spatial mapping, affine transformation, and deep learning spatial mapping networks. These spatial algorithms can be used independently or in combination with existing skeletal constraint models, physical constraints, and temporal constraints to ensure that the generated three-dimensional structure maintains reasonableness in terms of morphology, proportion, and dynamic continuity. By clarifying the feasibility of spatial algorithms, the scope of this invention covers any implementation method capable of mapping two-dimensional structures to three-dimensional space and generating structurally faithful, dynamically continuous, and controllable animations.

[0013] During the data preprocessing stage, the acquired two-dimensional ultrasound grayscale data undergoes privacy desensitization processing, including removing sensitive information such as the pregnant woman's identity and geographical location. Data security is ensured through encrypted storage and hierarchical access control mechanisms. Image optimization processing includes noise suppression, grayscale enhancement, and region-of-interest (ROI) cropping, providing high-quality input for subsequent modeling. The system can incorporate user safety guidelines for pregnant women, providing daily or weekly usage time suggestions and operation prompts for home users, intelligently suggesting acquisition frequency and optimal operation time periods to ensure safe use. The user guidelines described in this invention are entirely geared towards non-medical home scenarios and do not involve clinical diagnosis or medical decisions; they are intended only as a reasonable use reference.

[0014] In the structural modeling stage, a two-dimensional structural model is constructed through key point detection, contour segmentation, and topological relationships. Subsequently, an age-adaptive calibration engine is used to adjust the node spacing, proportions, and key morphologies to ensure that the model conforms to the actual anatomical features of fetuses at different gestational ages. The two-dimensional structural model is then used to generate a three-dimensional skeletal articulation model through spatial algorithms, assigning degrees of freedom and constraint thresholds to key nodes to ensure that the dynamic process conforms to the intrauterine movement patterns of the fetus.

[0015] The dynamic generation stage employs structural rigidity constraints and temporal consistency constraints to generate continuous motion sequences. Dynamic driving supports both standard template-driven and user-defined parameter-driven approaches, meeting diverse home scenarios and personalized needs. The generated 3D motion sequences undergo smoothing optimization and rendering enhancement, making them adaptable to mobile terminals, home display devices, or web platforms. They enable 360° multi-angle viewing, zooming, and looping playback, and are explicitly labeled "For non-medical use only."

[0016] The visualization system provided by this invention adopts a modular architecture, including a data acquisition and de-identification module, a structural feature extraction and modeling module, a dynamic driving and optimization module, and a rendering output and compliance verification module. Each module is seamlessly connected through standardized interfaces, achieving fully automated operation. The system can be deployed on mobile terminals, lightweight edge devices, home display devices, or web platforms, requiring no high-performance computing platform, allowing even non-professional users to complete animation generation and interactive operations.

[0017] Through the above technical solutions, this invention achieves structurally faithful, dynamically continuous, and controllable 3D animation generation based on 2D ultrasound data, and establishes a closed loop for full-process privacy protection and compliance. By introducing flexible implementation methods of spatial algorithms, the scope of protection of this invention covers all technical means that can achieve equivalent 2D to 3D mapping, skeletal constraints, and dynamic generation, taking into account innovation, feasibility, and wide applicability. It can be widely used in non-medical scenarios such as family pregnancy commemoration, pregnancy and childbirth science education, and parent-child entertainment interaction, and has the potential for large-scale industrialization. Attached Figure Description

[0018] To facilitate understanding of the technical solution of this invention, the description is provided in conjunction with the accompanying drawings, which are used to illustrate the processing flow and structural relationships of this invention and do not constitute a limitation on the scope of protection of this invention. The drawings may include overall process diagrams, structural model diagrams, and dynamically generated process diagrams.

[0019] Figure 1 shows a schematic diagram of the overall modular architecture of a family scene fetal visualization generation system in one embodiment of the present disclosure.

[0020] Figure 2 is a schematic diagram illustrating the process of preprocessing, structural extraction, and desensitization and encryption of the original ultrasound image in one embodiment of this disclosure.

[0021] Figure 3 illustrates a schematic diagram of constructing a three-dimensional fetal model based on structured security data and generating dynamic visualization content in one embodiment of this disclosure.

[0022] Figure 4 illustrates a schematic diagram of task scheduling and rendering output under a mid-cloud hybrid deployment architecture in one embodiment of this disclosure.

[0023] Figure 5 is a schematic diagram illustrating a multimodal user interaction and compliance behavior control process in one embodiment of this disclosure.

[0024] Figure 6 is a schematic diagram illustrating the closed-loop control of the entire process from data acquisition to compliant output in one embodiment of this disclosure. Specific implementation methods

[0025] This invention provides a method and system for generating dynamic visualizations of the fetus based on two-dimensional ultrasound data. Specific implementation methods include, but are not limited to, the following. Those skilled in the art can, based on the disclosure in this specification, employ any technical means, algorithm combinations, parameter adjustments, or module modifications capable of achieving the same or equivalent functions and effects, all of which fall within the protection scope of this invention. This implementation method is used to illustrate the technical solution and is not a limiting description.

[0026] During implementation, the fetal two-dimensional grayscale ultrasound image data is first collected using legally compliant, mass-produced handheld ultrasound devices or home portable terminals. These include devices capable of acquiring two-dimensional grayscale images, such as wireless probes, smartphones with integrated ultrasound modules, and embedded portable terminals. The data collection process strictly adheres to the informed consent and authorization procedures for the pregnant woman, employing written authorization, electronic authorization, or biometric verification methods. Information such as collection time, actual gestational age, device model, collection position, number of image frames, and collection environment parameters are recorded to establish a dedicated data archive. This provides a complete basis for subsequent gestational age adaptive modeling, parameter calibration, and result retrospective analysis, ensuring the legality and compliance of the data source and avoiding modeling biases caused by missing or non-compliant data.

[0027] Before data enters the core processing stage, the original images and associated metadata undergo full-process desensitization and security control. Desensitization methods include automated batch processing scripts, rule engines, and AI-assisted recognition, removing the pregnant woman's identity information and medical annotations, retaining only the fetal grayscale image and necessary non-privacy metadata (such as gestational age and frame sequence). Data transmission uses SSL / TLS, end-to-end encryption, or equivalent security mechanisms, while storage uses AES-256 or encryption algorithms compliant with national standards, supplemented by hierarchical access control, operation logs, and regular audits. This achieves closed-loop management of the entire lifecycle of data collection, transmission, storage, processing, use, and destruction, ensuring compliance with the Personal Information Protection Law and data ethics norms. Any equivalent desensitization, encryption, and control methods are within the scope of protection of this invention.

[0028] In the preprocessing stage, to address issues such as uneven grayscale, speckle noise, and blurred contours in 2D ultrasound images, a multi-level optimization method is employed. This includes arbitrary combinations of edge-preserving filtering, adaptive noise reduction, Gaussian filtering, median filtering, wavelet transform, BM3D, nonlocal means, and lightweight neural networks. This aims to suppress noise while fully preserving the fetal contours and key structural details. Adaptive grayscale enhancement (including CLAHE, local histogram equalization, Retinex, and gamma correction) optimizes local grayscale differences, enhancing the contrast between the fetal head, torso, limbs, and background. Preprocessing also includes intelligent fetal subject recognition and automatic ROI cropping. This can be based on template matching, contour detection, lightweight CNN, or Transformer segmentation methods, and adaptively adjusts the cropping boundaries and range based on gestational age parameters to adapt to different body shapes and postures of fetuses at different gestational weeks. The algorithms in this stage are lightweight and can run efficiently on mobile terminals, edge computing, or cloud-based collaborative modes without requiring high-performance workstations, while maintaining both accuracy and real-time performance.

[0029] In the structural modeling stage, a two-dimensional structural model is constructed through keypoint detection, contour segmentation, and topological relationships, including a set of keypoints, node topology, and contour curves. The two-dimensional structural model is then mapped to a three-dimensional skeletal joint articulation model using a spatial mapping algorithm. Spatial mapping can employ any combination of thin-plate spline interpolation, radial basis functions, affine / nonlinear transformations, physical constraint optimization, and depth mapping networks, and can be optimized using statistical fetal movement models or physiological data. Each key node is assigned appropriate degrees of freedom and rotation / translation constraints, and the node spacing, limb proportions, and trunk curvature are automatically adjusted by a gestational age adaptive calibration engine (lookup table method, regression model, rule engine, or lightweight neural network) to ensure the three-dimensional model conforms to the anatomical features of the actual gestational age. Different mapping strategies can be selected based on the terminal's computing power during the modeling process to achieve structural fidelity, naturalness, and stability. Any equivalent method for achieving two-dimensional to three-dimensional structural mapping and physiological constraints falls within the scope of this invention.

[0030] During the dynamic generation phase, a dual control approach is employed, combining structural rigidity constraints and temporal consistency. Structural constraints monitor the relative positions, topological relationships, and proportional constraints of key nodes in real time, ensuring the overall structure remains distortion-free or does not experience clipping during movement. Temporal consistency, combined with cross-frame differential analysis, attitude prediction, and AI-assisted large-scale model generation, achieves continuous motion optimization and global coherence, including but not limited to Kalman filtering, particle filtering, LSTM / GRU sequence prediction, optical flow constraints, and trajectory smoothing. Dynamic driving can utilize fetal intrauterine movement templates (resting, slight movements, turning / kicking, etc.) or user-defined parameters (amplitude, speed, direction, number of cycles, and specific action selection), and iteratively optimizes continuous motion sequences using an AI large-scale model, ensuring natural, coherent, and controllable movements. This approach is fully applicable to home settings and does not involve medical diagnosis or clinical decision-making.

[0031] The generated 3D motion sequence undergoes further processing including trajectory smoothing, low-pass filtering, and rendering enhancement. The trajectory can be achieved using a combination of cubic polynomial interpolation, B-splines, Bézier curves, and frequency domain low-pass filtering; the 3D mesh can be smoothed using Laplacian smoothing, Taubin smoothing, and bilateral filtering, combined with soft texture mapping, progressive lighting, and semi-transparent effects to improve visual realism and smoothness. The rendering method adapts to terminal performance, including lightweight mesh rendering, point cloud rendering, volume rendering, and semi-transparent texture overlay, and supports 360° viewing, zooming, panning, looping, slow motion, and screenshot / recording operations. The output animation is clearly labeled "For family commemoration, pregnancy science popularization, and parent-child entertainment only; not for medical diagnosis or clinical use," ensuring compliance throughout the entire process.

[0032] This invention's visualization system adopts a modular architecture, with each module seamlessly connected through standardized interfaces, achieving fully automated operation without manual intervention. The system can run on mobile terminals, home display devices, edge computing platforms, and web interfaces, allowing even non-professional users to independently complete the entire process of data uploading, parameter input, animation generation, viewing, saving, and sharing.

[0033] Through the above implementation methods, this invention, while ensuring data security and usage standards, achieves high-precision conversion of two-dimensional ultrasound data into structurally accurate, dynamically continuous, and adjustable three-dimensional animation through spatial mapping, AI-assisted dynamic generation, and lightweight design. It balances professionalism, practicality, and innovation, and is suitable for non-medical scenarios such as family pregnancy commemoration, science popularization, and parent-child entertainment, and has the potential for industrialization and promotion.

Claims

1. A method for generating dynamic three-dimensional visualization of the fetus based on two-dimensional ultrasound data, characterized in that, Includes the following steps: Step 1: Obtain the fetal two-dimensional grayscale ultrasound image sequence and the corresponding gestational age parameters; Step 2: Extract structural descriptive information representing fetal morphology from the image sequence; Step 3: Generate or select a corresponding three-dimensional structural prior model based on the gestational age parameters. The three-dimensional structural prior model is used to define the proportion, morphological range, and topological relationship of the fetal anatomical structure. Step 4: Through the structure-guided fusion mechanism, the structural description information is fused with the three-dimensional structural prior model to generate a three-dimensional structural representation that conforms to physiological constraints. Step 5: Based on the three-dimensional structural representation, a continuous fetal dynamic sequence is generated through a motion generation mechanism jointly controlled by structural constraints and temporal continuity constraints. Step 6: Smooth the dynamic sequence and output an interactive 3D visualization result.

2. The method according to claim 1, characterized in that, The three-dimensional structural prior model is generated or selected driven by the gestational age parameter, and at least limits the proportion of skeletal nodes, the range of limb length, the curvature of the trunk, and the positional relationship of major anatomical landmarks.

3. The method according to claim 1 or 2, characterized in that, The structure-guided fusion mechanism includes mapping the two-dimensional structure description information to the three-dimensional structure prior model, and performing morphological correction and consistency optimization of the three-dimensional structure through a constraint adjustment process to ensure the fit between the three-dimensional structure and the actual anatomical features of the fetus.

4. The method according to any one of claims 1 to 3, characterized in that, The structural constraints include limitations on the topological relationships, relative distance ratios, and degrees of freedom of motion between key nodes, in order to prevent distortion or clipping of the three-dimensional structure during dynamic changes.

5. The method according to any one of claims 1 to 4, characterized in that, The temporal continuity constraint is used to limit the range of pose changes between adjacent frames, smooth the motion trajectory, and maintain the overall coherence of the dynamic sequence.

6. The method according to any one of claims 1 to 5, characterized in that, The motion generation mechanism is driven by a preset library of fetal intrauterine motion patterns, user-adjustable parameters, or a combination of both.

7. The method according to any one of claims 1 to 6, characterized in that, Before extracting the structural description information, the image sequence is preprocessed, including one or more of noise suppression, contrast enhancement, and region of interest extraction.

8. The method according to any one of claims 1 to 7, characterized in that, It also includes steps for de-identifying the collected data and generated results, controlling secure transmission, and managing access permissions.

9. A fetal dynamic three-dimensional visualization generation system for implementing the method of any one of claims 1 to 8, characterized in that, include: The data acquisition unit is used to acquire two-dimensional ultrasound image sequences of the fetus and gestational age parameters. The structure extraction unit is used to extract fetal morphological structure description information from image sequences; Prior model unit, used to generate or select a three-dimensional structural prior model based on gestational age parameters; The fusion generation unit is used to perform structure-guided fusion to produce a three-dimensional structural representation that conforms to physiological constraints; Dynamic generation unit, used to generate continuous dynamic sequences under structural constraints and temporal continuity constraints; The output unit is used to smooth dynamic sequences and output visualization results.

10. A computer-readable storage medium or electronic device, characterized in that, It stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 8.

Images