Ostomy bag base information generation method and system based on ostomy state acquisition result

Abstract

This invention belongs to the field of medical and health data information processing, and particularly relates to a method and system for generating ostomy bag base information based on ostomy status acquisition results. It includes a medical image acquisition unit, a medical data processing unit, and a design unit. The medical image acquisition unit is used to acquire multimodal medical image data of the patient's ostomy area, including at least color and depth images. The medical image acquisition unit includes a ring-shaped light-averaging module that provides shadowless illumination and a user guidance interface for guiding the user to complete standard pose acquisition. By adopting the ring-shaped light-averaging module and the image data acquisition unit, the problem of poor image quality caused by uneven ambient lighting is solved. Simultaneously, the ring-shaped light-averaging ensures the realism and shadowlessness of the two-dimensional color image's color and texture, while the depth sensor directly acquires three-dimensional geometric information, providing reliable input for high-precision segmentation and reconstruction, avoiding the shortcomings of insufficient information from a single sensor.

Description

Technical Field

[0001] This invention belongs to the field of medical and health data information processing, and in particular relates to a method and system for generating ostomy bag base information based on ostomy status acquisition results. Background Technology

[0002] Traditional ostomy bag bases (or base plates) are mostly prefabricated standard shapes and sizes, requiring manual cutting to fit the different stoma shapes and abdominal contours of each patient. This method has the following inherent drawbacks: manual cutting relies on the experience of nursing staff, making it difficult to accurately match irregular stoma boundaries and complex abdominal curvatures, which can easily lead to leakage of excrement and cause skin inflammation; rigid or semi-rigid standard base plates cannot adapt to individualized changes in abdominal curvature, easily causing local pressure, leading to pain and pressure sores; in addition, the shape of the patient's stoma and surrounding tissues will change during the postoperative recovery period, and traditional methods cannot achieve rapid and low-cost adaptation.

[0003] In recent years, with the development of healthcare informatics, some systems attempting to use digital means to assist stoma care have gradually emerged. For example, two-dimensional image acquisition and simple measurements can assist nurses in cutting and planning. However, these systems are mostly limited to two-dimensional information processing, lacking accurate reconstruction and mechanical analysis of the three-dimensional shape of the patient's abdomen, and thus unable to achieve truly personalized, conformal fit design. In addition, existing systems still have significant shortcomings in the acquisition, processing, and modeling of medical data: the acquisition process is greatly affected by ambient lighting, resulting in unstable image quality; the segmentation algorithm has limited accuracy in recognizing stoma boundaries, especially prone to errors in areas with complex textures and low contrast; the three-dimensional reconstruction model is coarse and cannot accurately reflect key medical and health data such as stoma bulges, depressions, or the curvature of the surrounding abdominal wall; in the design stage, there is a lack of a systematic method that combines patient-specific anatomical features with mechanical performance requirements, still relying on manual experience for adjustments, resulting in low design efficiency and unreliable results. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for generating ostomy bag base information based on ostomy status acquisition results, so as to solve the above-mentioned technical problems.

[0005] To achieve the above objectives, the specific technical solution of the method and system for generating ostomy bag base information based on ostomy status acquisition results of the present invention is as follows:

[0006] An ostomy bag base information generation system based on ostomy status acquisition results includes a medical image acquisition unit, a medical data processing unit, and a design unit. The medical image acquisition unit is used to acquire multimodal medical image data of the patient's ostomy area, including at least color and depth images. The medical image acquisition unit includes a ring-shaped light uniformity module that provides shadowless illumination and a user guidance interface for guiding the user to complete standard pose acquisition.

[0007] The medical data processing unit is communicatively connected to the medical image acquisition unit and is used to process the multimodal medical image data to generate a set of biometric parameters including stoma morphological features and surrounding abdominal wall curvature geometric features; the medical data processing unit includes a medical information image segmentation module, a medical data three-dimensional reconstruction module, and a feature quantization module;

[0008] The medical information image segmentation module is used to segment the stoma region from the multimodal medical image data and output its pixel-level boundary mask. It adopts a model based on an encoder-decoder architecture, and the training loss function of the model includes a boundary consistency loss to improve the boundary segmentation accuracy.

[0009] The medical data 3D reconstruction module is used to reconstruct the abdominal wall triangular mesh model containing the stoma structure based on the multimodal medical image data and the pixel-level boundary mask, using the Poisson surface reconstruction algorithm. The reconstruction accuracy is controlled by setting the octree depth during the reconstruction process.

[0010] The feature quantization module is used to extract the set of biological feature parameters from the pixel-level boundary mask and the abdominal wall triangular mesh model;

[0011] The design unit is communicatively connected to the medical data processing unit and is used to automatically generate a three-dimensional digital model of the ostomy bag base that conforms to the three-dimensional morphology of the patient's abdominal wall and has a spatial gradient mechanical property distribution based on the biometric parameter set; the design unit includes a parametric modeling engine and a simulation optimization engine.

[0012] The parametric modeling engine is used to generate an initial base model with a spatial gradient thickness field based on the set of biometric parameters. The thickness field is determined by the shortest distance from the grid point to the stoma contour boundary and the average curvature of the local surface at that point, so that the base thickness gradually transitions from the central region to the edge region, and the thickness is compensated and adjusted according to the average curvature of the local surface.

[0013] The simulation optimization engine is used to perform finite element analysis on the initial base model. Under simulated abdominal pressure load and material volume constraints, it optimizes the thickness field to minimize structural flexibility and obtains a three-dimensional digital model with optimized mechanical properties.

[0014] Furthermore, the medical information image segmentation module adopts an encoder-decoder architecture-based model, whose training loss function is a weighted sum of region loss, classification loss, and boundary consistency loss. The boundary consistency loss is constructed by calculating the gradient difference between the predicted probability map and the true label map at the boundary pixels.

[0015] Furthermore, the Poisson surface reconstruction algorithm used in the medical data 3D reconstruction module has an octree depth set to any value of 9, 10, or 11 in its solution process.

[0016] Furthermore, the set of biometric parameters extracted by the feature quantization module includes:

[0017] The two-dimensional morphological parameters of the ostomy obtained from the pixel-level boundary mask include at least its projected area and perimeter.

[0018] The three-dimensional geometric parameters of the stoma region obtained from the abdominal wall triangular mesh model include at least the average curvature of the abdominal wall within a preset annular region around the stoma.

[0019] Furthermore, the pre-defined annular region around the stoma refers to an annular region with a radius of 15mm to 25mm centered on the stoma boundary, and the average curvature of the abdominal wall is the arithmetic mean of the average curvature of all grid vertices within this region.

[0020] Furthermore, in the spatial gradient thickness field defined by the parametric modeling engine, the thickness gradient from the center region to the edge region is achieved through a Sigmoid-type transition function, the transition steepness coefficient and the transition center point of which are adjustable.

[0021] Furthermore, the optimization problem performed by the simulation optimization engine includes minimizing the structural flexibility under the constraints of satisfying the structural equilibrium equation, ensuring that the material volume does not exceed the target value, and ensuring that the thickness is within the upper and lower limits.

[0022] Furthermore, the simulated abdominal pressure load is 3 kPa to 7 kPa, and the target volume is 85% to 95% of the initial model volume.

[0023] Furthermore, it also includes an additive manufacturing unit, which is communicatively connected to the design unit and is used to manufacture a physical ostomy bag base according to the three-dimensional digital model. The additive manufacturing unit is a medical-grade 3D printer using fused deposition modeling or digital light processing technology, and the printing material used is a biocompatible flexible thermoplastic polyurethane or flexible photosensitive resin.

[0024] Furthermore, the method for generating ostomy bag base information based on ostomy status acquisition results includes the following steps:

[0025] S100: Acquire multimodal medical image data of the patient's stoma area, including at least color and depth images;

[0026] S200: Process the multimodal medical image data to generate a set of biometric parameters including stoma morphological features and surrounding abdominal wall curvature geometric features; specifically including:

[0027] S210: The multimodal medical image data is processed using a segmentation model to obtain a pixel-level boundary mask for the stoma region;

[0028] S220: Based on the multimodal medical image data and the pixel-level boundary mask, generate an abdominal wall triangular mesh model through multi-view stereo vision and surface reconstruction algorithms;

[0029] S230: Quantitatively extract the set of biological feature parameters from the pixel-level boundary mask and the abdominal wall triangular mesh model;

[0030] S300: Based on the aforementioned biometric parameter set, a three-dimensional digital model of the ostomy bag base is automatically generated through generative design, conforming to the three-dimensional morphology of the patient's abdominal wall and possessing a spatial gradient mechanical property distribution; specifically including:

[0031] S310: Based on the set of biological characteristic parameters, generate an initial base model with a defined spatial gradient thickness field using a parameterized formula. The thickness field is determined by the distance from the point to the stoma boundary and the local curvature.

[0032] S320: Perform finite element analysis on the initial base model, and adjust the thickness field through iterative optimization under material volume constraints to obtain a final three-dimensional digital model with optimized mechanical properties;

[0033] S400: Based on the three-dimensional digital model, a solid ostomy bag base is manufactured using additive manufacturing technology.

[0034] The advantages of this invention are:

[0035] 1. By adopting a ring-shaped light-averaging module and an image data acquisition unit, the problem of poor image quality caused by uneven ambient lighting is solved. At the same time, the ring-shaped light-averaging module ensures the authenticity and shadowlessness of the color and texture of the two-dimensional color image, while the depth sensor directly acquires three-dimensional geometric information, providing reliable input for high-precision segmentation and reconstruction, and avoiding the shortcomings of insufficient information from a single sensor.

[0036] 2. By employing a loss mechanism that includes boundary consistency The hybrid loss function and Poisson reconstruction with octree depths of 9, 10, or 11 solve the problems of blurred two-dimensional boundary segmentation of the stoma and coarse distortion of the three-dimensional model. At the same time, the boundary loss function forces the segmentation network to learn accurate edge features, while the Poisson reconstruction with limited depth range ensures that these fine two-dimensional boundaries can be "carved" onto the high-fidelity three-dimensional model, so that the extracted biological feature parameters truly reflect the physiological structure.

[0037] 3. By adopting the average curvature around the stoma Parameters and spatial gradient thickness field The formula solves the problem that traditional base plates cannot conform to complex abdominal curvatures. Furthermore, the average curvature quantifies the macroscopic shape of the bonding surface, while the thickness field formula addresses the local curvature. With distance By using these variables together and dynamically adjusting the thickness, the final base not only fits perfectly in shape but also achieves a gradient distribution of "rigid support at the center and smooth sealing at the edges" in terms of mechanical properties.

[0038] 4. Defined by using a parametric modeling engine The simulation optimization engine (performs topology optimization) solves the problems of personalized design relying on experience and difficulty in guaranteeing structural strength. At the same time, parametric modeling quickly generates personalized initial models, while simulation optimization is based on real abdominal pressure loads. and volume constraints We conducted mechanical verification and iterative fine-tuning to ensure that, while meeting clinical strength requirements, we could save materials to the greatest extent and improve wearing comfort. Attached Figure Description

[0039] Figure 1 This is a diagram showing the overall architecture of the ostomy bag base information generation system based on ostomy status acquisition results of the present invention.

[0040] Figure 2 This is a flowchart of the method for generating ostomy bag base information based on ostomy status acquisition results according to the present invention;

[0041] Figure 3 for Figure 2 Flowchart of step S200;

[0042] Figure 4 for Figure 2 Flowchart of step S300. Detailed Implementation

[0043] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0044] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0045] Example 1

[0046] like Figure 1 As shown, the ostomy bag base information generation system based on ostomy status acquisition results includes a medical image acquisition unit, a medical data processing unit, and a design unit. The medical image acquisition unit is used to acquire multimodal medical image data of the patient's ostomy area, including at least color and depth images. The medical data processing unit is communicatively connected to the medical image acquisition unit and is used to process the multimodal medical image data to generate a set of biometric parameters including ostomy morphological features and the geometric features of the surrounding abdominal wall surface. The design unit is communicatively connected to the medical data processing unit and is used to automatically generate a three-dimensional digital model of the ostomy bag base that conforms to the three-dimensional morphology of the patient's abdominal wall and has a spatial gradient mechanical property distribution based on the biometric parameter set.

[0047] The medical image acquisition unit includes an RGB color camera, a depth sensor, a ring-shaped light distribution module, and a user guidance interface. The RGB color camera is used to acquire the color image, the depth sensor is used to acquire the depth image, the ring-shaped light distribution module is used to provide shadowless illumination during acquisition, and the user guidance interface is used to guide the user to complete image acquisition in a standard pose through augmented reality contour overlay or voice prompts.

[0048] The medical image acquisition unit is an integrated device that combines a high-resolution RGB color camera, a depth sensor, and a ring-shaped light-diffusing module. The high-resolution RGB color camera uses 12 megapixels, and the ring-shaped light-diffusing module consists of multiple groups of LED beads arranged in a ring, providing uniform and shadowless illumination at the moment of acquisition. The color temperature is stable at around 5500K, ensuring color fidelity and texture clarity of the color image under different ambient lighting conditions. The device is equipped with a touch screen as a user guidance interface. During acquisition, the screen uses augmented reality (AR) technology to overlay a semi-transparent standard stoma outline in real time, accompanied by voice prompts (such as "Please align the device with the stoma and keep the green outline roughly aligned with the stoma edge"), guiding the user or medical staff to place the device directly above the abdominal wall surface, and automatically triggering image acquisition. The acquired data includes at least one set of registered and aligned color and depth images, providing reliable multimodal input for subsequent processing. This setup, by employing a ring-shaped light-averaging module and an image data acquisition unit, solves the problem of poor image quality caused by uneven ambient lighting. At the same time, the ring-shaped light-averaging ensures the realism and shadowlessness of the colors and textures in the two-dimensional color images, while the depth sensor directly acquires three-dimensional geometric information, providing reliable input for high-precision segmentation and reconstruction, and avoiding the shortcomings of insufficient information from a single sensor.

[0049] The depth sensor is a structured light camera, a time-of-flight camera, or a binocular stereo camera. The working distance of the depth sensor is 20-40 cm, within which the accuracy of the depth image can be better than ±0.5 mm.

[0050] The system also includes an additive manufacturing unit, which is communicatively connected to the design unit. This additive manufacturing unit is used to manufacture a physical ostomy bag base based on the three-dimensional digital model. The additive manufacturing unit is a medical-grade 3D printer using fused deposition modeling (FDM) or digital light processing (DLP) technology. The printing material used is a biocompatible flexible thermoplastic polyurethane or flexible photosensitive resin. The printing layer thickness of the additive manufacturing unit is set to 50 micrometers to ensure the surface quality and dimensional accuracy of the model. After printing, the base undergoes cleaning (to remove uncured resin), post-curing (to ensure complete material polymerization), and standard ethylene oxide or low-temperature plasma sterilization processes, ultimately becoming a sterile, personalized ostomy bag base suitable for clinical use.

[0051] Example 2

[0052] like Figure 1As shown, the medical data processing unit includes a medical information image segmentation module, a medical data 3D reconstruction module, and a feature quantization module. The medical information image segmentation module is used to segment the stoma region from the multimodal medical image data and output its pixel-level boundary mask. The medical data 3D reconstruction module is used to reconstruct an abdominal wall triangular mesh model containing the stoma structure based on the multimodal medical image data and the pixel-level boundary mask. The feature quantization module is used to extract the biological feature parameter set from the pixel-level boundary mask and the abdominal wall triangular mesh model.

[0053] The medical data processing unit is typically deployed on a local server or workstation.

[0054] The medical information image segmentation module employs an encoder-decoder architecture-based model, and the training loss function of this model is... for:

[0055]

[0056] in, These are the weighting coefficients, and ;

[0057] The The boundary consistency loss is calculated using the following formula:

[0058]

[0059] In the formula, For the set of true boundary pixels, To predict the probability map, This is a real label image. This is the gradient operator.

[0060] In this embodiment, the medical information image segmentation module uses an encoder-decoder architecture-based model that has been pre-trained on a dataset containing a large number of labeled stoma images. During training, in addition to using the conventional Dice loss and cross-entropy loss, a boundary consistency loss has also been specifically added. Its weighting coefficient The optimal value is 0.1. This allows the model to focus more on the accuracy of boundary pixels during segmentation, resulting in clear edges in the output pixel-level boundary mask and effectively reducing blurry areas.

[0061] In this embodiment, for The weighting coefficient of the Dice loss is used to adjust the model's focus on the "overall integrity and overlap of the stoma region." This factor measures the overlap area between the predicted and ground truth regions. Increasing this weight forces the model to prioritize ensuring that the predicted stoma region is as consistent as possible with the ground truth in shape and size, avoiding large omissions or oversegments, and ensuring that the segmented stoma region is accurate as a whole, providing a basis for subsequent calculations of two-dimensional parameters such as area and perimeter. The preferred value is 0.5;

[0062] In this embodiment, for The weighting coefficient of the cross-entropy loss is used to adjust how much the model focuses on the accuracy of classifying each pixel. This is a typical classification loss, penalizing each pixel for misclassification. Increasing this weight forces the model to focus on the classification confidence of all pixels, optimizing global pixel-level classification accuracy and ensuring the basic pixel-level accuracy of the segmentation result. This is the cornerstone of obtaining a reliable segmentation mask. The preferred value is 0.4;

[0063] In the encoder-decoder architecture of the medical information image segmentation module, the encoder can employ a ResNet-50 backbone network pre-trained on ImageNet to extract rich multi-level features. The decoder fuses shallow detail features and deep semantic features from the encoder through skip connections. During training, in addition to using a hybrid loss function including L_boundary, an online hard example mining strategy can be introduced, focusing on stoma region samples with low contrast to the background and blurred boundaries, thereby improving the robustness of the model in complex real-world scenarios. During inference, for predicted regions with low confidence, a lightweight post-processing based on graph cut can be used to further refine the segmentation boundaries.

[0064] The medical data 3D reconstruction module employs the Poisson surface reconstruction algorithm, which solves the Poisson equation. Convert point clouds into continuous triangular mesh models The octree depth in the solution process is set to 9, 10, or 11. In this implementation, the preferred octree depth for solving the Poisson equation is 10. This depth achieves a good balance between reconstruction accuracy and computational efficiency, resulting in a triangular mesh model of the ventral wall. The surface is smooth and faithfully reflects the three-dimensional bulge structure of the stoma and the complex curvature of the surrounding abdominal wall. This design, employing a boundary consistency loss mechanism, allows for further optimization. The hybrid loss function and Poisson reconstruction with octree depths of 9, 10, or 11 solve the problems of blurred two-dimensional boundary segmentation and coarse and distorted three-dimensional models of the stoma. At the same time, the boundary loss function forces the segmentation network to learn accurate edge features, while the Poisson reconstruction with limited depth range ensures that these fine two-dimensional boundaries can be "carved" onto the high-fidelity three-dimensional model, so that the extracted biological feature parameters truly reflect the physiological structure.

[0065] The medical data 3D reconstruction module first uses a statistical filter to remove outlier noise points when processing point clouds. During Poisson reconstruction, in addition to setting the octree depth, other parameters such as the reconstruction scale can be adjusted to control the density of the final mesh. For the stoma region, which is the core focus, the system can automatically identify this region and use a finer reconstruction scale there, thereby ensuring the 3D geometric fidelity of the stoma and its edges while keeping the overall model data volume under control.

[0066] The set of biometric parameters extracted by the feature quantization module includes:

[0067] The two-dimensional morphological parameters of the stoma obtained from the pixel-level boundary mask include at least its projected area. With perimeter ;

[0068] The three-dimensional geometric parameters of the stoma region obtained from the abdominal wall triangular mesh model include at least the average curvature of the abdominal wall within a preset annular region around the stoma. ;

[0069] The mean curvature The calculation formula is:

[0070]

[0071] in, Indicates the radius centered on the stoma boundary. The set of all mesh vertices within a ring-shaped region ranging from 15mm to 25mm. for The average curvature.

[0072] The feature quantization module automatically calculates, from the aforementioned mesh model and two-dimensional mask, features such as the stoma projection area and perimeter, as well as the average curvature within a pre-defined annular region (e.g., radius r is preferably 20 mm). Key biological characteristic parameters;

[0073] The set of biometric parameters extracted by the feature quantification module serves as direct input for subsequent personalized design. In addition to the basic geometric parameters mentioned, more descriptive parameters can be added in practice. For example:

[0074] Stoma protrusion height: Calculate the vertical distance of the highest point of the stoma relative to the surrounding abdominal wall reference plane.

[0075] Quantification of local undulations in the abdominal wall: Within a larger area centered on the stoma (e.g., a radius of 40 mm), the average curvature or Gaussian curvature of different sub-regions is calculated to form a curvature distribution map, which is used to assess the overall flatness or ruggedness of the abdominal wall.

[0076] Stoma major axis direction: The major axis direction is determined by calculating the minimum circumscribed ellipse of the two-dimensional projection of the stoma. This information can be used to guide the directional design of the base opening, making it more consistent with the anatomical position and movement trend.

[0077] These richer feature parameters provide more comprehensive patient-specific information for the design unit, making the generated base model not only more macroscopically fit but also more accurately in terms of microscopic mechanical adaptation.

[0078] The design unit includes a parametric modeling engine and a simulation optimization engine. The parametric modeling engine is used to generate a thickness field with spatial gradient based on the set of biometric parameters. The initial base model is characterized by its non-uniform thickness, which is determined by a spatial gradient thickness field. The simulation optimization engine is defined as being used to perform finite element analysis on the initial base model and optimize the thickness field under material volume constraints to minimize structural flexibility. This is achieved by using the average curvature around the stoma. Parameters and spatial gradient thickness field The formula solves the problem that traditional base plates cannot conform to complex abdominal curvatures. Furthermore, the average curvature quantifies the macroscopic shape of the bonding surface, while the thickness field formula addresses the local curvature. With distance By using these variables together and dynamically adjusting the thickness, the final base not only fits perfectly in shape but also achieves a gradient distribution of "rigid support at the center and smooth sealing at the edges" in terms of mechanical properties.

[0079] The spatial gradient thickness field defined by the parametric modeling engine. The following relationship must be satisfied:

[0080]

[0081] in,

[0082] This represents the shortest distance from a grid point to the stoma outline boundary;

[0083] The mean curvature of the local surface at the grid point is... for The normalized value;

[0084] The center reference thickness ranges from 2.5 mm to 4.0 mm.

[0085] The edge reference thickness ranges from 1.0 mm to 1.5 mm.

[0086] It is the Sigmoid transition function. For shape parameters;

[0087] This is the curvature compensation coefficient, with a value ranging from 0.05mm to 0.20mm.

[0088] Wherein, the Sigmoid transition function Specifically:

[0089]

[0090] in, This is the transition steepness coefficient, with a value ranging from 0.2 to 0.5; The value ranges from 10mm to 20mm, serving as the transition center point.

[0091] In this embodiment, a reference thickness is set for the stoma center region. Preferably 3.0 mm, the reference thickness of the edge region Preferably 1.2 mm, transition function Transition steepness coefficient The preferred value is 0.3, the transition center point. Preferably 15 mm, curvature compensation coefficient The preferred thickness is 0.08 mm, which means that in convex areas with high curvature, the model will be locally thickened to provide better support; in flat or concave areas, it will be thinned accordingly to improve flexibility.

[0092] The mathematical model of the optimization problem executed by the simulation optimization engine is as follows:

[0093]

[0094] in, For structural flexibility, For displacement vectors, For thickness The relevant stiffness matrix, To simulate the load vector of abdominal pressure, The total number of finite element elements. For unit volume, For the target volume, and These represent the lower and upper limits of thickness.

[0095] Wherein, the load vector The simulated abdominal pressure values ​​ranged from 3 kPa to 7 kPa, and the target volume... The volume is 85% to 95% of the initial model volume.

[0096] The simulation optimization engine is suitable for fields with spatial gradient thickness. Finite element analysis was performed on the initial base model. In this embodiment, a uniform 5 kPa ventral pressure (load vector F) was applied, and under the condition that the total material volume did not exceed 90% of the initial model volume, iterative optimization was performed with the goal of minimizing the structural flexibility C, ultimately obtaining a three-dimensional digital model with optimized mechanical properties. This setup was achieved by defining a parametric modeling engine. The simulation optimization engine (performs topology optimization) solves the problems of personalized design relying on experience and difficulty in guaranteeing structural strength. At the same time, parametric modeling quickly generates personalized initial models, while simulation optimization is based on real abdominal pressure loads. and volume constraints We conducted mechanical verification and iterative fine-tuning to ensure that, while meeting clinical strength requirements, we could save materials to the greatest extent and improve wearing comfort.

[0097] Example 3

[0098] like Figure 2 As shown, the method for generating ostomy bag base information based on ostomy status acquisition results includes the following steps:

[0099] S100: Acquire multimodal medical image data of the patient's stoma area, including at least color and depth images;

[0100] S200: Process the multimodal medical image data to generate a set of biometric parameters that includes stoma morphological features and surrounding abdominal wall curvature geometric features;

[0101] S300: Based on the set of biometric parameters, a three-dimensional digital model of the ostomy bag base is automatically generated through generative design, which conforms to the three-dimensional shape of the patient's abdominal wall and has a spatial gradient mechanical property distribution.

[0102] S400: Based on the three-dimensional digital model, a solid ostomy bag base is manufactured using additive manufacturing technology.

[0103] Step S200 includes:

[0104] S210: The multimodal medical image data is processed using a segmentation model to obtain a pixel-level boundary mask for the stoma region;

[0105] S220: Based on the multimodal medical image data and the pixel-level boundary mask, generate an abdominal wall triangular mesh model through multi-view stereo vision and surface reconstruction algorithms;

[0106] S230: Quantitatively extract the set of biological feature parameters from the pixel-level boundary mask and the abdominal wall triangular mesh model.

[0107] In step S210, the segmentation model minimizes the mixture loss function. Training is performed, which includes a boundary consistency loss term to improve boundary segmentation accuracy. .

[0108] In step S230, the extracted set of biometric parameters includes the average curvature of the annular region surrounding the stoma. It is obtained by calculating the arithmetic mean of the average curvature of all grid vertices in the region.

[0109] Step S300 includes:

[0110] S310: Based on the set of biometric parameters, generate a spatially gradient thickness field using a parameterized formula. The initial base model;

[0111] S320: Perform finite element analysis on the initial base model, and adjust the thickness field through iterative optimization under material volume constraints to obtain a final three-dimensional digital model with optimized mechanical properties.

[0112] In step S320, the goal of the iterative optimization is to minimize the flexibility of the base structure while controlling the material volume between 85% and 95% of the initial volume under the condition of applying simulated abdominal pressure loads of 3 kPa to 7 kPa.

[0113] In step S400, medical-grade flexible materials are used for 3D printing, and after printing, the materials are cleaned, cured, and disinfected.

[0114] Steps S300 and S400 can be implemented in variations according to different clinical needs and resource conditions.

[0115] Design optimization variant:

[0116] In step S300, the topology optimization performed by the simulation optimization engine can have different focuses.

[0117] Variant A (comfort as the primary objective): The optimization objective is set to minimize the average contact pressure between the base and the abdominal wall while meeting certain strength thresholds (e.g., maximum stress not exceeding 50% of the material's yield strength). This requires iterative optimization using contact algorithms in finite element analysis, aiming to achieve the design with the softest wearing comfort.

[0118] Variant B (for rapid response): This simplifies the optimization process for environments requiring urgent replacement or with limited resources. The simulation optimization engine can be replaced by a rule-based thickness adjuster. This adjuster incorporates a base of expert-based rules, such as: "If the average curvature..." >0.2mm -1 "If the stoma protrusion height is >10mm, then increase the global thickness base value by 10%"; "If the stoma protrusion height is >10mm, then increase the thickness of the central area". While its optimality may be slightly inferior to complete topology optimization, it can generate a customized design that is significantly better than the standard baseplate in a very short time.

[0119] Manufacturing method variants:

[0120] In step S400, the additive manufacturing technology can be selected based on the material properties and production scale.

[0121] Variant C (Multi-material Printing): This uses 3D printing technologies that support multi-material jetting (such as PolyJet). Different areas of the base can be designed to be made of materials with varying hardness: the sealing ring around the stoma uses the softest material (such as Shore A hardness 40) to ensure a seal and comfort; the load-bearing structural areas use medium-hardness materials (such as Shore A hardness 70) to provide support; and the edge fixing parts can use slightly harder materials to maintain shape. This multi-material integrated printing better realizes the concept of "spatial gradient mechanical property distribution."

[0122] Variant D (Batch Customization and Distributed Manufacturing): The 3D digital model generated by the design unit can be uploaded to a cloud data center. After being authorized, patients or medical institutions can download the digital file and send it to a local or nearby third-party certified 3D printing service center for manufacturing. This model connects professional manufacturing capabilities with a wide range of end-user needs, making it particularly suitable for community healthcare or home care scenarios, enabling rapid, decentralized, and personalized production.

[0123] This embodiment addresses the problems of traditional personalized design, such as reliance on human experience, long cycles, and limited manufacturing methods, by providing a design framework that combines parametric modeling and simulation optimization, along with diverse additive manufacturing variants. It achieves automated and intelligent rapid response in the design process and can flexibly select the most suitable design and manufacturing strategies based on different clinical priorities (such as optimal comfort or fastest delivery) and resource conditions, greatly improving the clinical adaptability and scalability of the technology.

[0124] Example 4

[0125] This embodiment illustrates how the system adapts to and handles special stoma conditions.

[0126] Scenario 1: Retracted Stoma: When the stoma is below the abdominal wall surface, traditional two-dimensional measurements fail. This system's depth imaging can directly capture the depth and extent of the retraction. The feature quantization module calculates the retraction depth and the slope of the retraction edges accordingly. When generating the model, the curvature H in the thickness field formula is made negative, allowing for the design of special guide and support structures at the retraction edges. This ensures the base conforms to the retraction contour while effectively collecting excrement and preventing it from seeping into the retraction.

[0127] Scenario 2: Parastomal Hernia: When there is abdominal wall bulging around the stoma, the abdominal wall surface is extremely complex and dynamically changing. During data acquisition, the system can guide the patient to acquire images in both relaxed and slightly tense abdominal states. The medical data processing unit reconstructs abdominal wall models in both states. During simulation optimization, the design unit can set the load conditions within a range (e.g., 3-7 kPa) to simulate changes in abdominal pressure, optimizing a robust design that maintains good sealing and low stress under both extreme conditions, or directly generating a series of bases suitable for different states.

[0128] This embodiment addresses the problem that traditional methods cannot effectively handle complex pathological anatomical structures such as depressed stomas and parastomal hernias by directly acquiring three-dimensional geometric information through depth images and by employing a design strategy based on multi-state acquisition and robust optimization. It expands the applicability of personalized systems, enabling them to provide safe and effective customized solutions for various special and complex stoma conditions, thereby improving the quality and safety of care.

[0129] Example 5: Example of ostomy bag base fabrication based on specific patient data

[0130] This example uses a patient, "Wang Moumou," as an example. Patient information: Wang Moumou, male, 6 weeks post-sigmoid colostomy. Clinical assessment indicates the stoma is an irregular oval shape, with localized bulging on the surrounding abdominal wall due to mild abdominal wall weakness. Execution steps and data examples:

[0131] S100: Acquire multimodal medical image data of the patient's stoma region, including at least color and depth images.

[0132] Operation Procedure: The nurse uses the integrated data acquisition device as described in Example 1. The patient lies supine with the abdomen fully exposed. After the device is activated, the ring-shaped light distribution module illuminates, and a green guide frame is displayed on the touchscreen via AR overlay. Following the voice prompt, "Please move the device slowly, ensuring the stoma is completely within the frame and remains stable," the nurse places the device approximately 30cm directly above the abdominal wall. The device automatically acquires and registers one high-resolution color image (3840×2160 pixels) and one set of corresponding depth images (depth map resolution 1024×768, accuracy ±0.4mm@30cm). To ensure data quality, the acquisition is repeated three times, and the system automatically selects the set of data with the highest image clarity and smallest registration error for subsequent processing.

[0133] Output: A set of registered color-depth image pairs.

[0134] S200: Process the multimodal medical image data to generate a set of biometric parameters that includes stoma morphological features and the geometric features of the surrounding abdominal wall surface.

[0135] S210: The multimodal medical image data is processed using a segmentation model to obtain a pixel-level boundary mask for the stoma region;

[0136] The medical data processing unit loads data based on the encoder-decoder architecture as described in Example 2, and includes a boundary loss term. A segmentation model with a similarity of 0.1 was used. The model infers from the input color image to obtain a predicted probability map of the stoma region, then refines the boundaries through CRF post-processing, and finally outputs a pixel-level boundary mask. Compared with the standard, the Dice similarity coefficient of this segmentation reached 0.96, and the boundary IoU was 0.93.

[0137] S220: Based on the multimodal medical image data and the pixel-level boundary mask, generate an abdominal wall triangular mesh model through multi-view stereo vision and surface reconstruction algorithms;

[0138] Using the aforementioned mask, a dense point cloud (approximately 550,000 points) focusing on the stoma and surrounding abdominal wall was extracted from the depth image. The medical data 3D reconstruction module employed the Poisson surface reconstruction algorithm with an octree depth set to 10, generating a triangular mesh model of the abdominal wall. This model contained approximately 280,000 vertices and 560,000 triangular faces, clearly reconstructing the three-dimensional morphology of the stoma and the surrounding raised abdominal wall surface.

[0139] S230: Quantitatively extract the set of biological feature parameters from the pixel-level boundary mask and the ventral wall triangular mesh model:

[0140] Calculation of the stoma projection area from a two-dimensional mask. =945mm²; Perimeter =125mm.

[0141] Calculated from a 3D mesh model:

[0142] Mean curvature around the stoma Taking the stoma boundary as the center, a ring-shaped region with a radius of r = 20 mm is selected. The arithmetic mean of the average curvature of all mesh vertices within this region is calculated to obtain... =0.12mm -1 (A positive value indicates that the area as a whole has a convex trend.)

[0143] Stoma protrusion height: The distance between the highest point of the stoma and the surrounding reference plane was calculated and measured to be 4.8 mm.

[0144] Abdominal wall curvature distribution characteristics: The system identified a region with relatively high local curvature (curvature up to 0.25 mm) at the 10 o'clock position of the stoma. -1 ).

[0145] Output: A structured set of biometric parameters, including all the geometric and curvature parameters mentioned above.

[0146] S300: Based on the set of biometric parameters, a three-dimensional digital model of the ostomy bag base is automatically generated through generative design, which is conformal to the three-dimensional shape of the patient's abdominal wall and has a spatial gradient mechanical property distribution.

[0147] S310: Based on the set of biometric parameters, generate a spatially gradient thickness field using a parameterized formula. Initial base model:

[0148] The parametric modeling engine of the design unit receives a set of feature parameters. Based on clinical rules and patient data, the engine sets the parameters: =3.2mm (considering moderate protrusion height), =1.2mm, Transition function In =0.35, =18mm. Specifically, for identified high-curvature regions, the system automatically adjusts the curvature compensation coefficient for that region. A thickness of 0.15 mm (higher than the default of 0.1 mm) is provided to offer additional support. The spatial gradient thickness field formula is applied. An initial three-dimensional digital model of the base was generated, with an estimated volume of 28.5 cm³.

[0149] S320: Perform finite element analysis on the initial base model, and adjust the thickness field through iterative optimization under material volume constraints to obtain a final three-dimensional digital model with optimized mechanical properties.

[0150] The simulation optimization engine performs finite element analysis on the initial model. The load condition is set to simulate a uniform intra-abdominal pressure of 5 kPa (load vector F). The optimization objective is to optimize the material volume constraint (…). Minimize structural flexibility (assuming 90% of the initial volume ≈ 25.65 cm³). After five iterations of optimization, the optimized model volume was reduced to 25.3 cm³, satisfying the constraints; the maximum von Mises stress decreased from the initial 0.78 MPa to 0.52 MPa, and the stress distribution became more uniform; the structural flexibility was reduced by approximately 18%. The optimization process focused on strengthening the internal structure of high curvature regions, while material reduction was implemented in non-load-bearing areas.

[0151] Output: A final 3D digital model (STL format) of the ostomy bag base with optimized mechanical properties.

[0152] S400: Based on a three-dimensional digital model, a solid ostomy bag base is manufactured using additive manufacturing technology.

[0153] Manufacturing process: The final STL model is sent to a medical-grade DLP 3D printer. A flexible, biocompatible photosensitive resin with a Shore A hardness of 55 is selected as the printing material. Printing parameter settings: layer thickness = 50μm, exposure time per layer = 8 seconds. The total printing time is 4 hours and 20 minutes.

[0154] Post-processing: After printing, the base is processed sequentially as follows:

[0155] Cleaning: Ultrasonic cleaning in isopropanol for 20 minutes to remove uncured resin.

[0156] Post-curing: Curing at 60°C and 405nm wavelength ultraviolet light for 45 minutes to ensure that the material properties fully meet the standards.

[0157] Sterilization and Packaging: Low-temperature ethylene oxide sterilization cycle is adopted. After analysis and inspection, aseptic packaging is carried out.

[0158] Final delivery: From image acquisition to obtaining a sterile, personalized base, the entire process was completed within 16 hours. Clinical trial feedback indicated that the base fit snugly against the patient's abdominal wall, with no leakage observed during daily activities such as coughing and bending over. Patients reported significantly greater comfort levels compared to the previously used standard cut baseboard.

[0159] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.

Claims

1. A system for generating ostomy bag base information based on ostomy status acquisition results, characterized in that, It includes a medical image acquisition unit, a medical data processing unit, and a design unit. The medical image acquisition unit is used to acquire multimodal medical image data of the patient's stoma region, including at least color and depth images. The medical image acquisition unit includes a ring-shaped light homogenizing module that provides shadowless illumination and a user guidance interface for guiding the user to complete standard pose acquisition. The medical data processing unit is communicatively connected to the medical image acquisition unit and is used to process the multimodal medical image data to generate a set of biometric parameters including stoma morphological features and surrounding abdominal wall curvature geometric features; the medical data processing unit includes a medical information image segmentation module, a medical data three-dimensional reconstruction module, and a feature quantization module; The medical information image segmentation module is used to segment the stoma region from the multimodal medical image data and output its pixel-level boundary mask. The medical information image segmentation module adopts a model based on an encoder-decoder architecture. The training loss function of this model includes a boundary consistency loss to improve the boundary segmentation accuracy. The medical data 3D reconstruction module is used to reconstruct the abdominal wall triangular mesh model containing the stoma structure based on the multimodal medical image data and the pixel-level boundary mask, and adopts the Poisson surface reconstruction algorithm. The reconstruction accuracy is controlled by setting the octree depth during the reconstruction process. The feature quantization module is used to extract the set of biological feature parameters from the pixel-level boundary mask and the abdominal wall triangular mesh model; The design unit is communicatively connected to the medical data processing unit and is used to automatically generate a three-dimensional digital model of the ostomy bag base that conforms to the three-dimensional morphology of the patient's abdominal wall and has a spatial gradient mechanical property distribution based on the biometric parameter set; the design unit includes a parametric modeling engine and a simulation optimization engine. The parametric modeling engine is used to generate an initial base model with a spatial gradient thickness field based on the set of biometric parameters. The thickness field is determined by the shortest distance from the grid point to the stoma contour boundary and the average curvature of the local surface at that point, so that the base thickness gradually transitions from the central region to the edge region, and the thickness is compensated and adjusted according to the average curvature of the local surface. The simulation optimization engine is used to perform finite element analysis on the initial base model. Under simulated abdominal pressure load and material volume constraints, it optimizes the thickness field to minimize structural flexibility and obtains a three-dimensional digital model with optimized mechanical properties.

2. The system according to claim 1, characterized in that, The medical information image segmentation module adopts an encoder-decoder architecture-based model, whose training loss function is a weighted sum of region loss, classification loss, and boundary consistency loss. The boundary consistency loss is constructed by calculating the gradient difference between the predicted probability map and the true label map at the boundary pixels.

3. The system according to claim 1, characterized in that, The medical data 3D reconstruction module uses the Poisson surface reconstruction algorithm, and the octree depth in its solution process is set to any value of 9, 10 or 11.

4. The system according to claim 1, characterized in that, The set of biological feature parameters extracted by the feature quantization module includes: The two-dimensional morphological parameters of the ostomy obtained from the pixel-level boundary mask include at least its projected area and perimeter. The three-dimensional geometric parameters of the stoma region obtained from the abdominal wall triangular mesh model include at least the average curvature of the abdominal wall within a preset annular region around the stoma.

5. The system according to claim 4, characterized in that, The pre-defined annular region around the stoma refers to an annular region with a radius of 15mm to 25mm centered on the stoma boundary, and the average curvature of the abdominal wall is the arithmetic mean of the average curvature of all grid vertices in this region.

6. The system according to claim 1, characterized in that, In the spatial gradient thickness field defined by the parametric modeling engine, the thickness gradient from the center region to the edge region is achieved through a Sigmoid-type transition function, where the transition steepness coefficient and the transition center point are parametrically defined controllable variables.

7. The system according to claim 1, characterized in that, The optimization problem performed by the simulation optimization engine includes minimizing structural flexibility under the constraints of satisfying the structural equilibrium equation, ensuring that the material volume does not exceed the target value, and ensuring that the thickness is within the upper and lower limits.

8. The system according to claim 7, characterized in that, The simulated abdominal pressure load is 3 kPa to 7 kPa, and the target volume is 85% to 95% of the initial model volume.

9. The system according to claim 1, characterized in that, It also includes an additive manufacturing unit, which is communicatively connected to the design unit and is used to manufacture a physical ostomy bag base according to the three-dimensional digital model. The additive manufacturing unit is a medical-grade 3D printer using fused deposition modeling or digital light processing technology, and the printing material used is a biocompatible flexible thermoplastic polyurethane or flexible photosensitive resin.

10. A method for generating ostomy bag base information based on ostomy status acquisition results, applicable to the system described in any one of claims 1-9, characterized in that, Includes the following steps: S100: Acquire multimodal medical image data of the patient's stoma area, including at least color and depth images; S200: Process the multimodal medical image data to generate a set of biometric parameters including stoma morphological features and surrounding abdominal wall curvature geometric features; specifically including: S210: The multimodal medical image data is processed using a segmentation model to obtain a pixel-level boundary mask for the stoma region; S220: Based on the multimodal medical image data and the pixel-level boundary mask, generate an abdominal wall triangular mesh model through multi-view stereo vision and surface reconstruction algorithms; S230: Quantitatively extract the set of biological feature parameters from the pixel-level boundary mask and the abdominal wall triangular mesh model; S300: Based on the aforementioned biometric parameter set, a three-dimensional digital model of the ostomy bag base is automatically generated through generative design, conforming to the three-dimensional morphology of the patient's abdominal wall and possessing a spatial gradient mechanical property distribution; specifically including: S310: Based on the set of biological characteristic parameters, generate an initial base model with a defined spatial gradient thickness field using a parameterized formula. The thickness field is determined by the distance from the point to the stoma boundary and the local curvature. S320: Perform finite element analysis on the initial base model, and adjust the thickness field through iterative optimization under material volume constraints to obtain a final three-dimensional digital model with optimized mechanical properties; S400: Based on the three-dimensional digital model, a solid ostomy bag base is manufactured using additive manufacturing technology.

Images