A method and equipment for manufacturing an automated applicator
By using an automated dressing manufacturing method and image recognition and numerical control technology, the problems of insufficient precision and safety in the preparation of the P-32 dressing have been solved, enabling efficient and safe personalized treatment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-02-02
- Publication Date
- 2026-06-02
AI Technical Summary
The existing method for preparing the P-32 applicator relies on manual operation, which makes it difficult to accurately reproduce the complex boundaries of irregular scars. It also lacks precise quantitative control, resulting in problems such as long preparation cycle, low safety, and poor product consistency.
An automated dressing manufacturing method is adopted, which uses image recognition models (such as TransUNet network) for feature extraction and segmentation, combined with CNC cutting and quantitative solution application technology to generate dressing substrate that matches the shape of skin scars, and achieves automated production through drying and sealing in a closed environment.
This achieves a high degree of matching between the applicator and the shape of the lesion, improving the accuracy of irradiation and the safety of treatment, shortening the preparation cycle, reducing radiation risks, and ensuring product consistency and standardization.
Smart Images
Figure CN122124396A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of medical device manufacturing, radionuclide therapy, and automated control technology, specifically to an automated dressing manufacturing method and equipment. Background Technology
[0002] Skin scars, especially hypertrophic scars and keloids, are a common result of abnormal skin tissue repair after injuries such as trauma, surgery, burns, or infections. These lesions not only manifest as localized thickening, elevation, or tumor-like changes in the skin, often accompanied by discomfort symptoms such as itching and pain, but also have a significant negative impact on patients' mental health, social interactions, and quality of life. With social development and the public's increasing demand for aesthetics, the demand for treatment of various skin diseases, especially those involving cosmetic repair, continues to grow, making scar treatment an important topic that combines clinical and humanistic care.
[0003] Currently, radionuclide patch therapy has become one of the effective methods for treating keloids and other refractory skin lesions, especially phosphorus-32 (P-32) and strontium-90 / yttrium-90 (90Sr-90Y) patches. Both utilize the strong ionizing ability and short intra-tissue range of pure beta rays to achieve precise irradiation of the lesion area while avoiding damage to deep and surrounding normal tissues. However, the 90Sr-90Y patch is prefabricated with a long half-life, but its fixed shape and limited application area make it difficult to meet the treatment needs of multiple or large-area scars, and often requires multiple treatments, increasing the burden on patients. In contrast, P-32 patch therapy allows for on-site customization of the patch according to the shape and size of the scar, achieving "one patch per scar," with good conformability. It is particularly suitable for irregular, multiple, or large-area scars, and usually can be completed in a single treatment, reducing the number of hospital visits for patients and facilitating clinical promotion.
[0004] However, the current preparation of P-32 applicators relies on manual operation, with the general process being: manually outlining the scar, manually cutting the filter paper base, manually infusing the P-32 solution, and then drying and sealing. This traditional method has the following significant drawbacks: 1) Relying on manual drawing and cutting makes it difficult to accurately reproduce the complex boundaries of irregular scars, resulting in poor matching between the shape of the dressing and the actual contour of the lesion, affecting the accuracy of irradiation and treatment effect; 2) The active area and radioactivity of the patch rely heavily on empirical estimation and lack precise quantitative control, which can easily lead to uneven or deviated irradiation doses, affecting the consistency and safety of the therapeutic effect. 3) The entirely manual process is time-consuming and labor-intensive, with a long preparation cycle, making it difficult to meet the large clinical demand and exacerbating the problem of patients waiting in line for treatment; 4) In open, manual dripping, transfer, and handling processes, there is a risk of spillage, splashing, or aerosol diffusion of radioactive solutions, posing a potential radiation safety threat to operators and the environment. 5) The quality of preparation is highly dependent on the operator's personal skills and experience. There is poor consistency between different batches and even within the same batch. The lack of standardized procedures is not conducive to the standardization of treatment plans and the objective evaluation of efficacy.
[0005] The above reasons result in existing manual preparation methods for P-32 applicators having systemic deficiencies in terms of precision, efficiency, safety, and standardization, making it difficult to meet the growing clinical demand for personalized and precise treatment. Summary of the Invention
[0006] To address the above technical problems, this invention provides a technical solution for an automated applicator manufacturing method and equipment.
[0007] The technical problem solved by this invention can be achieved by the following technical solutions: An automated applicator manufacturing method includes: Step S1: Obtain the original image of the skin scar; Step S2: The original image is subjected to feature extraction and segmentation using an image recognition model to generate a feature map of the target region; Step S3: Generate the corresponding cutting path based on the feature map of the target region; Step S4: Based on the cutting path, the carrier material is cut to form an applicator substrate that matches the shape of the skin scar. Step S5: Apply phosphorus-32 solution to the applicator substrate according to the preset phosphorus-32 radioactivity parameters; Step S6: Post-process the applicator substrate after applying the phosphorus-32 solution to generate a phosphorus-32 applicator.
[0008] Preferably, in step S2, the image recognition model is a TransUNet network model, which includes a convolutional neural network layer and a Transformer coding layer b.
[0009] Preferably, the TransUNet network model is trained using a composite loss function based on Dice coefficient loss and binary cross-entropy loss.
[0010] Preferably, step S2 includes: Step S21: Extract local features from the original image using the convolutional neural network layer; Step S22: Perform global context modeling on the extracted local features through the Transformer encoding layer b; Step S23: Decode and upsample the modeled features to generate a target region feature map containing skin scar morphological features, including the shape, area and color of the scar.
[0011] Preferably, step S3 includes: Step S31: Based on the feature map of the target region, extract the closed contour vector information of the skin scar; Step S32: Calculate the actual physical area of the skin scar based on the closed contour vector information; Step S33: Based on the actual physical area, the closed contour vector information is converted into a motion command sequence based on the actual XY plane coordinate system to generate the cutting path.
[0012] Preferably, step S4 includes: Step S41: The carrier material is transferred to the cutting area; Step S42: Control the XY direction translation guide rail to move along the cutting path, and control the cutting drill to move along the Z direction to cut the carrier material.
[0013] Preferably, step S5 includes: Step S51: Calculate the required radioactivity of phosphorus-32 solution based on the scar area information in the target area feature map to obtain the phosphorus-32 radioactivity parameter. Step S52: The applicator substrate is transferred to the solution dispensing area; Step S53: Based on the phosphorus-32 radioactivity parameter, a phosphorus-32 solution with the corresponding activity is added dropwise to the applicator substrate.
[0014] Preferably, step S6 includes: Step S61: The applicator substrate after applying phosphorus-32 solution is transferred to the drying area for heating and drying, and the steam generated during heating and drying is removed. Step S62: The dried applicator substrate is coated and sealed to generate the phosphorus-32 applicator.
[0015] An automated applicator manufacturing apparatus for implementing an automated applicator manufacturing method as described above, comprising: Image acquisition module, used to acquire raw images of skin scars; An image processing module, connected to the image acquisition module, is used to extract and segment features from the original image using an image recognition model to generate a feature map of the target region. The cutting module, connected to the image processing module, is used to cut the carrier material according to the feature map of the target area to form an applicator substrate that matches the shape of the skin scar. A solution dripping module, connected to the cutting module, is used to apply phosphorus-32 solution to the applicator substrate according to a preset phosphorus-32 radioactivity parameter; A drying module, connected to the solution dripping module, is used to dry the patch substrate after the phosphorus-32 solution has been applied. The collection module, connected to the drying module, is used to coat the dried applicator substrate to generate a phosphorus-32 applicator.
[0016] Preferably, it further includes a performance monitoring and feedback module, which is connected to the image processing module, the cutting module, and the collection module, and is used for: The scar area in the feature map of the target region is compared with the manually calculated scar area; The fit between the patch substrate and the patient's scar was assessed by experts using visual evaluation. The uniformity of radioactivity distribution on the phosphorus-32 applicator was detected by radioactive phosphorus screen imaging.
[0017] Beneficial effects: This invention, through image recognition and automatic cutting technology, can accurately reproduce the contour of scars, achieving a high degree of matching between the applicator and the shape of the lesion, thus improving the accuracy of irradiation. At the same time, by controlling the application of radioactive solution through preset activity parameters, it ensures the quantification and consistency of active area and dose, improving the safety and stability of treatment efficacy. In addition, the automated process significantly shortens the preparation cycle and reduces manual intervention, which not only efficiently meets the large clinical demand and alleviates the waiting pressure on patients, but also avoids the risk of exposure to radioactive solution in manual operation, enhancing the radiation safety of operation and environment. Attached Figure Description
[0018] Figure 1 This is a flowchart of the preparation method of the present invention; Figure 2 This is a diagram of the TransUNet network model architecture of the present invention; Figure 3 This is a flowchart of the skin scar area calculation method of the present invention; Figure 4 This is a block diagram of the manufacturing equipment of the present invention; Figure 5 This is a schematic diagram of the manufacturing equipment of the present invention. Detailed Implementation
[0019] 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, and 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.
[0020] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.
[0021] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.
[0022] Reference Figure 1 This invention provides an automated applicator manufacturing method, comprising: Step S1: Obtain the original image of the skin scar; Step S2: The original image is subjected to feature extraction and segmentation using an image recognition model to generate a feature map of the target region; Step S3: Generate the corresponding cutting path based on the feature map of the target region; Step S4: Based on the cutting path, the carrier material is cut to form an applicator substrate that matches the shape of the skin scar. Step S5: Apply phosphorus-32 solution to the applicator substrate according to the preset phosphorus-32 radioactivity parameters; Step S6: Post-process the applicator substrate after applying the phosphorus-32 solution to generate a phosphorus-32 applicator.
[0023] Specifically, in this embodiment of the invention, in response to the systematic deficiencies of traditional manual preparation methods in terms of contour reproduction accuracy, dosage control consistency, preparation efficiency, and radiation safety, the invention integrates image recognition, automatic cutting, and quantitative liquid application technologies. This avoids errors from manual tracing and cutting, biases from experience estimation, and the risks of open operation. It achieves precise conformation of the phosphorus-32 patch to the lesion, controllable and repeatable radioactivity, highly efficient automation of the preparation process, and closed and safe operation, thereby improving the accuracy, safety, and standardization of treatment.
[0024] Specifically, in practical applications, medical staff or patients, under guidance, can use smartphones with high-resolution cameras or dedicated medical image acquisition equipment to photograph the patient's scarred skin area, obtaining clear photos or short video clips containing the scar and a small amount of surrounding normal skin as a reference. These digital image data can be transmitted via network or local transmission to the automated preparation process, thus serving as the source of the raw image data obtained in step S1.
[0025] This method makes full use of the widespread availability of mobile devices, facilitating the quick and easy initiation of the patch customization process in outpatient clinics, communities, and even remote settings, laying the foundation for subsequent automated processing.
[0026] In a preferred embodiment of the present invention, in step S2, the image recognition model is a TransUNet network model, which includes a convolutional neural network layer a and a Transformer coding layer b. The specific operation process of using the TransUNet network model to extract features and segment the original image is as follows: Local features are extracted from the original image through the convolutional neural network layer a; The extracted local features are modeled using the global context through the Transformer encoding layer b. The modeled features are decoded and upsampled to generate a target region feature map containing the morphological features of skin scars, including the shape, area and color of the scar.
[0027] Specifically, because skin scars, especially hypertrophic scars and keloids, often exhibit irregular boundaries, uneven elevations, and complex surface texture variations, traditional image segmentation methods or single-architecture convolutional neural network (CNN) models have limitations in accurately capturing such local details and long-range dependencies. Based on this, referring to... Figure 2 In this embodiment of the invention, a hybrid image recognition architecture is constructed using the TransUNet network to simultaneously achieve high-precision local feature extraction and global context information integration.
[0028] Specifically, the processing flow of the TransUNet network model is as follows: First, the original image of the scar taken by the camera of a mobile phone or device is input. Preliminary feature extraction is performed through three consecutive convolutional layers. These convolutional layers can effectively capture local details such as scar edges, textures and colors, and generate a series of multi-scale feature maps.
[0029] Next, these feature maps are flattened and linearly projected into a one-dimensional sequence, which is then input into an encoding module containing 12 Transformer layers. Through its self-attention mechanism, the Transformer establishes global dependencies between all pixels in the image, thereby integrating the overall morphological and structural information of the scar region. For example, it semantically associates the raised central region with the spreading edge to more accurately define the overall contour of the lesion.
[0030] Then, the TransUNet network performs feature restoration and refinement through the decoding path, which adopts a symmetrical upsampling structure and uses skip connections to fuse the rich spatial details from the early convolutional layers of the encoder with the high-level semantic information from the Transformer layers layer by layer.
[0031] In particular, to further enhance the discriminative power of feature representations, an attention mechanism module c (Convolutional Block Attention Module, CBAM) is introduced in this process. This module calculates attention weights sequentially along the channel and spatial dimensions, enabling the TransUNet network model to adaptively enhance key feature channels related to scars and focus on the spatial location of lesions, while suppressing background interference.
[0032] After multiple upsampling and feature fusion, the network finally outputs a pixel-level segmentation mask with the same resolution as the input image, which is an accurate target region feature map.
[0033] Through the above process, the TransUNet network not only achieves automatic and precise segmentation of scar lesions, but also directly calculates key parameters such as the specific shape, area, and color distribution of the lesions based on the segmentation results. This information provides a reliable data foundation for generating high-precision cutting paths in the subsequent step S3, thereby ensuring that the final prepared applicator substrate can achieve a high degree of conformal matching with the actual three-dimensional contour of the scar, effectively supporting personalized and precise clinical treatment.
[0034] In a preferred embodiment of the present invention, the TransUNet network model is trained using a composite loss function based on Dice coefficient loss and binary cross-entropy loss.
[0035] Specifically, considering that in skin scar image segmentation tasks, the area of the target region (i.e., the scar lesion) is usually much smaller than the background region, this class imbalance problem can easily lead to the model training process being overly biased towards the background, thereby reducing the segmentation accuracy of scar edge details. Based on this, in this embodiment of the invention, a composite loss function composed of a weighted combination of Dice coefficient loss and binary cross-entropy loss is used to optimize the TransUNet network.
[0036] Specifically, during training, the Dice coefficient loss measures the overlap between the predicted segmentation results and the ground truth labels, which can effectively mitigate the impact of class imbalance and enhance the model's sensitivity to small target regions; while the binary cross-entropy loss provides a stable gradient signal from the perspective of pixel classification, ensuring that the predicted probability distribution matches the ground truth labels.
[0037] This composite function combines the two, jointly guiding the model parameter update during backpropagation, enabling the network to not only focus on overall segmentation accuracy but also finely optimize the segmentation results of scar boundaries.
[0038] The TransUNet network model trained by this method can significantly improve the segmentation accuracy and robustness of irregular, small-area scars, providing a more reliable feature map basis for subsequent cutting path generation.
[0039] In a preferred embodiment of the present invention, step S3 includes: Step S31: Based on the feature map of the target region, extract the closed contour vector information of the skin scar; Step S32: Calculate the actual physical area of the skin scar based on the closed contour vector information; Step S33: Based on the actual physical area, the closed contour vector information is converted into a motion command sequence based on the actual XY plane coordinate system to generate the cutting path.
[0040] Specifically, to ensure that the shape of the final applicator substrate accurately conforms to the actual shape of the patient's scar, the pixel-level segmentation results in the target area feature map obtained in step S2 need to be converted into high-precision, executable physical cutting paths. Based on this, this embodiment of the invention further provides a method for converting and optimizing image contours into numerical control instructions, to improve the completeness and smoothness of contour extraction and achieve process adaptation.
[0041] Specifically, in order to eliminate artifacts (such as isolated scattered points or internal micro-holes) caused by image noise or small errors in the segmentation process, and to smooth the region boundaries to obtain more accurate and coherent contour lines, firstly, erosion and dilation operations are performed on the feature map of the target region (i.e., the segmentation mask) representing the scar and reference object to eliminate small holes caused by image noise or segmentation errors, reduce interference from non-target regions, and smooth pixel-level irregular undulations at the edges.
[0042] Next, refer to Figure 3 The coordinates of all outer boundary pixels of the skin scar in the processed target region feature map are extracted to obtain the closed contour vector information of the skin scar. Based on the set of coordinate points of the closed contour vector information, the approximate outer dimension of the skin scar in the image pixel coordinate system can be calculated. Based on these dimension information, the pixel area of the skin scar can be calculated.
[0043] Meanwhile, a circle with a known diameter is preferably used as a reference object. After morphological processing of the circle to smooth its edges, the coordinates of all continuous pixels on its outer boundary are accurately determined by performing topological analysis on the binary image, thereby obtaining a clear and closed reference object pixel outline at this scale. Based on this outline, the corresponding pixel area is calculated.
[0044] Furthermore, based on the ratio between the pixel area of the skin scar and the pixel area of the reference object, as well as the preset physical size of the reference object, the actual physical size of the skin scar can be calculated.
[0045] In this way, the actual cutting path can be obtained based on the closed contour vector information of the skin scar and its actual physical size.
[0046] Specifically, considering the physical properties of the applicator carrier material (such as the fiber structure and flexibility of filter paper) and the process parameters of the selected cutting device (such as the focusing spot diameter of a laser cutting head or the blade radius of a mechanical cutter head), path offset compensation and process adaptability optimization can be performed on the smoothed contour. For example, based on the heat-affected zone or blade radius, the design contour can be compensated for by equidistant expansion or contraction to ensure that the actual shape and size of the cut are consistent with the design contour.
[0047] Finally, the optimized final contour coordinate sequence, combined with preset process parameters such as cutting speed, entry and exit points, and idle stroke, is compiled into a standard CNC instruction sequence (such as G-code format) to drive the two-dimensional precision motion platform, thereby generating a cutting path file that can directly control the cutting device to perform precise forming operations.
[0048] In a preferred embodiment of the present invention, step S4 includes: Step S41: The carrier material is transferred to the cutting area; Step S42: Control the XY direction translation guide rail to move along the cutting path, and control the cutting drill to move along the Z direction to cut the carrier material.
[0049] Specifically, in this embodiment of the invention, a customized cutting scheme based on a CNC motion platform is adopted. This scheme uses the motion control principle of a CNC drilling machine to achieve contour cutting of flexible carrier materials through precise positioning in the X and Y axes and vertical feed motion in the Z axis.
[0050] Specifically, firstly, the carrier material is laid flat on the bottom conveying device, which (for example, a roller or precision belt conveying mechanism driven by a servo motor) automatically and continuously conveys the carrier material to the cutting area defined by the XY direction translation guide rail, and precisely stops at the preset starting processing position.
[0051] Next, the intelligent control system drives the XY-direction translation guide rail to move in coordination, causing the cutting drill bit mounted on the Z-axis moving mechanism to track the cutting path strictly in the horizontal plane.
[0052] At the same time, the Z-axis moving mechanism drives the cutting drill bit to move in the vertical direction, presses down when cutting is needed to complete the penetrating or semi-cutting processing of the filter paper, and lifts up in the non-cutting section, thereby achieving continuous and closed cutting along the scar contour.
[0053] The entire cutting process is automatically controlled by a program, requiring no manual intervention. After cutting, the bottom conveyor automatically transfers the formed filter paper substrate, which matches the shape of the scar, to the next station (such as the P-32 solution drop processing station). This solution ensures the customized, high-precision, and batch preparation of the applicator substrate, providing a physical carrier with a consistent morphology for the precise application of the subsequent radioactive solution.
[0054] In a preferred embodiment of the present invention, the carrier material is filter paper.
[0055] Specifically, phosphorus-32 patch therapy requires that the radionuclide be stably and uniformly loaded onto a carrier and continuously release beta rays to act on the lesion during treatment. Simultaneously, the carrier material must possess good biocompatibility, appropriate liquid absorption and retention capabilities, and ease of processing and sealing. Therefore, in this embodiment of the invention, medical-grade filter paper is selected as the ideal carrier material for the P-32 solution.
[0056] Specifically, filter paper, as a porous fibrous material, has a uniformly distributed microporous structure inside, which can quickly and uniformly adsorb and fix P-32 radioactive solution through capillary action, effectively preventing the solution from accumulating or spreading at the edges on the carrier surface, thereby ensuring the uniformity of dose distribution on the active surface of the applicator.
[0057] Meanwhile, the filter paper is soft and can closely conform to the irregular scar shape on the skin surface, improving the treatment fit; its thickness is moderate, which can ensure sufficient solution load without excessively increasing the overall thickness of the applicator, affecting the convenience and sealing of use.
[0058] In addition, filter paper has good cutting adaptability, which makes it easy to be precisely processed into substrates with any complex contour through automated cutting processes in the preceding steps.
[0059] Therefore, filter paper was chosen as the carrier, taking into account the physicochemical requirements of radionuclide loading, the biological requirements of clinical treatment, and the feasibility of automated manufacturing processes.
[0060] In a preferred embodiment of the present invention, step S5 includes: Step S51: Calculate the required radioactivity of phosphorus-32 solution based on the scar area information in the target area feature map to obtain the phosphorus-32 radioactivity parameter. Step S52: The applicator substrate is transferred to the solution dispensing area; Step S53: Based on the phosphorus-32 radioactivity parameter, a phosphorus-32 solution with the corresponding activity is added dropwise to the applicator substrate.
[0061] Specifically, in order to achieve precise and quantitative control of the radioactivity of phosphorus-32 on the dressing substrate, ensure accurate matching of treatment dose and scar area, and overcome the problems of traditional manual dripping relying on experience and uneven dosage, this embodiment of the invention sets up a solution dripping module based on precision motion control.
[0062] Specifically, the core drive unit of this module consists of a stepper motor, a coupling, and a precision ball screw. The stepper motor receives pulse commands from the control system, and its output shaft is directly connected to the precision ball screw via the coupling; the ball screw converts the motor's rotational motion into precise linear displacement of the micro-injection pump piston. By precisely controlling the stepper motor's step angle and speed, nanometer-level precision control of the injection pump piston's advance displacement and speed can be achieved, thereby ensuring highly accurate and repeatable dispensing volumes of P-32 solution.
[0063] Based on the calculated radioactivity parameters, the system automatically calculates the required volume of solution to be added and drives the liquid application mechanism to precisely add the corresponding volume of P-32 solution to the pre-positioned filter paper in a single-point or multi-point manner.
[0064] Because the filter paper material has uniform porous hydrophilic properties, the added solution will automatically, rapidly, and evenly diffuse and spread on its surface through capillary action until it covers the entire area that is essentially the same shape as the substrate. This process effectively adsorbs and confines the solution inside the filter paper, eliminating the risk of leakage and ensuring that the total radioactivity actually loaded on the filter paper strictly matches the preset value.
[0065] This mechatronics-based closed-loop quantitative dripping method not only automates, standardizes, and ensures consistent dosage of radiopharmaceutical delivery, greatly improving the safety and predictability of treatment efficacy, but also ensures that the entire application process is completed in a controlled, closed, or shielded environment, significantly reducing the risk of radiation exposure for operators and the possibility of environmental contamination by the solution.
[0066] In a preferred embodiment of the present invention, step S6 includes: Step S61: The applicator substrate after applying phosphorus-32 solution is transferred to the drying area for heating and drying, and the steam generated during heating and drying is removed. Step S62: The dried applicator substrate is coated and sealed to generate the phosphorus-32 applicator.
[0067] Specifically, to ensure that the P-32 solution is stably fixed on the carrier material to prevent the escape of radioactive materials, and to ensure the stability and safety of the applicator before storage, transportation and use, an automated post-processing procedure integrating drying, venting and sealing is designed in this embodiment of the invention.
[0068] Specifically, after the solution is added, the carrier substrate is conveyed into a dedicated drying module by a conveying device. The core of this module consists of a heating device and a negative pressure device. The heating device precisely controls the local ambient temperature of the substrate within a suitable range (e.g., 50°C) to gently and efficiently promote the evaporation of moisture or other solvents adsorbed by the filter paper, thereby solidifying the radioactive material.
[0069] Meanwhile, the supporting negative pressure device (such as a vacuum pump and its extraction pipe) works continuously to form and maintain a micro-negative pressure environment in the drying area, which quickly and directionally extracts the vapor containing radioactive solvent molecules generated during the heating process, and filters or collects it through a dedicated waste gas treatment system, completely preventing the vapor from spreading or condensing inside the equipment. This not only improves drying efficiency, but also ensures radiation safety in the working environment and the health of operators.
[0070] After drying, the substrate is automatically conveyed to the collection module. In this module, the dried applicator substrate is coated and sealed by automated equipment. Typically, a thin and uniform impermeable protective film (such as polyethylene film) is covered on its active surface, and the edges are heat-pressed or ultrasonically sealed. Finally, a finished P-32 applicator with stable physical state, safe sealing of radioactive materials, and ready for direct use in clinical treatment is produced.
[0071] Through the integrated post-processing described above, not only is rapid and controllable drying and sealing of the applicator substrate achieved, ensuring the safe fixation and effective protection of the radiation source, but also the automation of the operation minimizes manual intervention and reduces the radiation exposure risk for operators. At the same time, it ensures a high degree of consistency and reliability of each applicator product in terms of dosage safety, physical integrity and clinical applicability, providing key technical support for the standardization and large-scale application of P-32 patch therapy.
[0072] Reference Figure 4 and Figure 5 The present invention also provides an automated applicator manufacturing apparatus for implementing an automated applicator manufacturing method as described above, comprising: Image acquisition module 100 is used to acquire raw images of skin scars; Image processing module 200, connected to image acquisition module 100, is used to extract and segment features from the original image using an image recognition model to generate a feature map of the target region; The cutting module 300, connected to the image processing module 200, is used to cut the carrier material according to the feature map of the target area to form an applicator substrate that matches the shape of the skin scar. The solution dripping module 400 is connected to the cutting module 300 and is used to apply phosphorus-32 solution to the applicator substrate according to the preset phosphorus-32 radioactivity parameters. The drying module 500 is connected to the solution dripping module 400 and is used to dry the patch substrate after the phosphorus-32 solution is applied. The collection module 600 is connected to the drying module 500 and is used to coat the dried applicator substrate to generate a phosphorus-32 applicator.
[0073] In a preferred embodiment of the present invention, a performance monitoring and feedback module 700 is further included. The performance monitoring and feedback module 700 is connected to the image processing module 200, the cutting module 300, and the collection module 600, and is used for: The scar area in the feature map of the target region is compared with the manually calculated scar area; The fit between the patch substrate and the patient's scar was assessed by experts using visual evaluation. The uniformity of radioactivity distribution on the phosphorus-32 applicator was detected by radioactive phosphorus screen imaging.
[0074] Specifically, in order to quantitatively evaluate and control the key aspects of the automated manufacturing process, and to ensure that the final applicator product meets standards in terms of morphological matching, dosage accuracy, and expected clinical efficacy, reference is made to... Figure 4 In this embodiment of the invention, a performance monitoring and feedback module 700 is also provided to establish a closed-loop evaluation system covering image recognition accuracy, processing adaptability and radioactivity distribution quality.
[0075] In the image recognition accuracy verification stage, the module first compares the scar area (based on the target region feature map) automatically calculated and output by the image processing module 200 with the scar area obtained by the clinician using manual operation methods. If the relative error between the two is less than 10%, the image recognition accuracy of this batch is deemed to be qualified.
[0076] Secondly, in the process adaptability verification stage, at least two clinical experts with scar treatment experience can be selected to visually compare and evaluate the dressing substrate produced by the cutting module 300 with the corresponding patient's actual scar or high-precision model. The evaluation should be based on the contour matching degree, edge fit degree, etc. If more than 90% of the automatic conformal cutting substrate is evaluated as "conformal fit", the processing stage can be deemed qualified.
[0077] Finally, in the radioactivity quality verification stage, the finished phosphorus-32 patch output by the collection module 600 was imaged at high resolution using a radioactive phosphorus screen imaging system. By quantitatively analyzing the grayscale distribution of the image, the uniformity of its radioactivity distribution was detected. If the imaging results of more than 90% of the finished products showed no local abnormal radioactivity concentration or sparse defects, the uniformity of distribution met the preset standard.
[0078] The monitoring data from the performance monitoring and feedback module 700 will be fed back to the central control system in real time for the optimization and calibration of production parameters and quality traceability, forming a closed loop of manufacturing-monitoring-optimization to continuously improve the reliability and consistency of products.
[0079] In summary, this invention integrates core technologies such as artificial intelligence image recognition, digital path planning, automated precision machining, quantitative liquid application control, and closed-loop quality monitoring to construct a fully automated intelligent manufacturing system that covers the entire process from patient scar image acquisition to the preparation of personalized P-32 patch products.
[0080] Compared with the prior art, the present invention has the following significant advantages: High precision and personalized fit: The AI image recognition model (such as TransUNet) is used to automatically and accurately segment the scar area, and the carrier material is cut in a conformal manner through a CNC system. This solves the problem that traditional manual tracing and cutting cannot reproduce complex boundaries, ensuring that the shape of the applicator is highly matched with the contour of the lesion and improving the accuracy of irradiation.
[0081] Quantitative and consistent dose control: The required radioactivity is automatically calculated based on image segmentation results, and the quantitative and automated application of P-32 solution is achieved through a precision micro-droplet system. This overcomes the problem of dose inconsistency or deviation caused by traditional reliance on experience-based estimation, and ensures the consistency and safety of treatment dose.
[0082] Significantly improved production efficiency: The entire process is automated and interconnected, greatly reducing manual intervention, shortening the production cycle of a single product, and efficiently meeting the needs of clinical batches, alleviating the pressure of patients waiting in line.
[0083] Enhanced operational safety and radiation protection: The entire process is carried out in a controlled or closed environment. In particular, the quantitative addition and treatment of radioactive solutions are automated and closed, which effectively reduces the risk of solution spillage, splashing and aerosol diffusion, and significantly improves the radiation safety level of operators and the environment.
[0084] Standardization and traceability: By using digital processes and unified process parameter control, the reliance on the operator's personal experience is reduced, ensuring the consistency of product quality across different batches. Furthermore, the data management system enables full traceability of the production process, providing a reliable foundation for the standardization of treatment plans and the objective evaluation of efficacy.
[0085] Technology Integration and Industry Leadership: The project is the first fully automated intelligent manufacturing system for P-32 applicators in China, filling a gap in the industry. It has built an intelligent closed loop through the deep integration of multiple technologies, and has significant technological originality and industry leadership value.
[0086] Clinical translation and policy alignment: Addressing the pain points of manual clinical preparation, the project has completed the development of a prototype and outlined a clear medical device registration pathway, demonstrating promising prospects for clinical translation. Simultaneously, the project aligns closely with national policy guidance on the application of nuclear technology and independent innovation in high-end medical equipment, which will contribute to the popularization of nuclear medicine diagnostic and treatment technologies and the improvement of grassroots service capabilities.
[0087] The above description is merely a preferred embodiment of the present invention and does not limit the implementation and protection scope of the present invention. Those skilled in the art should realize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for manufacturing an automated applicator, characterized in that, include: Step S1: Obtain the original image of the skin scar; Step S2: The original image is subjected to feature extraction and segmentation using an image recognition model to generate a feature map of the target region; Step S3: Generate the corresponding cutting path based on the feature map of the target region; Step S4: Based on the cutting path, the carrier material is cut to form an applicator substrate that matches the shape of the skin scar. Step S5: Apply phosphorus-32 solution to the applicator substrate according to the preset phosphorus-32 radioactivity parameters; Step S6: Post-process the applicator substrate after applying the phosphorus-32 solution to generate a phosphorus-32 applicator.
2. The method for manufacturing an automated applicator according to claim 1, characterized in that, In step S2, the image recognition model is a TransUNet network model, which includes a convolutional neural network layer and a Transformer coding layer.
3. The method for manufacturing an automated applicator according to claim 2, characterized in that, The TransUNet network model is trained using a composite loss function based on Dice coefficient loss and binary cross-entropy loss.
4. The method for manufacturing an automated applicator according to claim 2, characterized in that, Step S2 includes: Step S21: Extract local features from the original image using the convolutional neural network layer; Step S22: Perform global context modeling on the extracted local features through the Transformer encoding layer; Step S23: Decode and upsample the modeled features to generate a target region feature map containing skin scar morphological features, including the shape, area and color of the scar.
5. The method for manufacturing an automated applicator according to claim 1, characterized in that, Step S3 includes: Step S31: Based on the feature map of the target region, extract the closed contour vector information of the skin scar; Step S32: Calculate the actual physical area of the skin scar based on the closed contour vector information; Step S33: Based on the actual physical area, the closed contour vector information is converted into a motion command sequence based on the actual XY plane coordinate system to generate the cutting path.
6. The method for manufacturing an automated applicator according to claim 1, characterized in that, Step S4 includes: Step S41: The carrier material is transferred to the cutting area; Step S42: Control the XY direction translation guide rail to move along the cutting path, and control the cutting drill to move along the Z direction to cut the carrier material.
7. The method for manufacturing an automated applicator according to claim 1, characterized in that, Step S5 includes: Step S51: Calculate the required radioactivity of phosphorus-32 solution based on the scar area information in the target area feature map to obtain the phosphorus-32 radioactivity parameter. Step S52: The applicator substrate is transferred to the solution dispensing area; Step S53: Based on the phosphorus-32 radioactivity parameter, a phosphorus-32 solution with the corresponding activity is added dropwise to the applicator substrate.
8. The method for manufacturing an automated applicator according to claim 1, characterized in that, Step S6 includes: Step S61: The applicator substrate after applying phosphorus-32 solution is transferred to the drying area for heating and drying, and the steam generated during heating and drying is removed. Step S62: The dried applicator substrate is coated and sealed to generate the phosphorus-32 applicator.
9. An automated applicator manufacturing device, characterized in that, A method for implementing an automated applicator manufacturing method as described in any one of claims 1-8, comprising: Image acquisition module, used to acquire raw images of skin scars; An image processing module, connected to the image acquisition module, is used to extract and segment features from the original image using an image recognition model to generate a feature map of the target region. The cutting module, connected to the image processing module, is used to cut the carrier material according to the feature map of the target area to form an applicator substrate that matches the shape of the skin scar. A solution dripping module, connected to the cutting module, is used to apply phosphorus-32 solution to the applicator substrate according to a preset phosphorus-32 radioactivity parameter; A drying module, connected to the solution dripping module, is used to dry the patch substrate after the phosphorus-32 solution has been applied. The collection module, connected to the drying module, is used to coat the dried applicator substrate to generate a phosphorus-32 applicator.
10. An automated applicator manufacturing device according to claim 9, characterized in that, It also includes a performance monitoring and feedback module, which is connected to the image processing module, the cutting module, and the collection module, and is used for: The scar area in the feature map of the target region is compared with the manually calculated scar area; The fit between the patch substrate and the patient's scar was assessed by experts using visual evaluation. The uniformity of radioactivity distribution on the phosphorus-32 applicator was detected by radioactive phosphorus screen imaging.