Medical device implant inspection system and inspection method thereof

Abstract

The invention discloses a medical device implant inspection system and an inspection method thereof, and relates to the technical field of medical device implant inspection. The medical device implant inspection system comprises the steps of obtaining multi-modal data of a to-be-inspected product, wherein the multi-modal data comprises first two-dimensional image data and three-dimensional point cloud data of the to-be-inspected product; performing feature fusion on the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector; matching the high-dimensional feature vector with a standard feature vector stored in a standard database so as to identify specification and model information of the to-be-detected product; calling standard product data corresponding to the specification and model information from the standard database; and inspecting the to-be-inspected product according to the standard product data, and generating an inspection result of the to-be-inspected product. The technical scheme provided by the invention can adapt to the inspection operation of medical instrument implants of multiple specifications and models.

Description

Technical Field

[0001] This invention relates to the field of medical device implant testing technology, and in particular to a medical device implant testing system and testing method. Background Technology

[0002] Medical device implants (hereinafter referred to as "implants"), such as dental implants or cardiac stents, are high-value, high-risk Class III medical devices, and their quality is directly related to patient safety. Therefore, factory inspection is a crucial part of the production process, which usually includes visual inspection (whether there are scratches, burrs, stains, and whether the shape is complete) and dimensional inspection (whether the key dimensions are within the tolerance range).

[0003] Due to individual patient differences and diverse clinical needs, implantable products are often characterized by a wide variety and complex specifications, resulting in a large number of SKUs (Stock Keeping Units). In flexible production models with multiple product types and small batches, a single production line often needs to frequently switch between producing different specifications and models. Traditional inspection methods rely entirely on experienced inspectors using the naked eye, magnifying glasses, or simple optical measuring tools for judgment. While some companies are gradually adopting machine vision solutions with technological advancements, these still have significant limitations, such as a lack of adaptability; when switching production lines to new specifications and models, inspectors must perform tedious system calibration and parameter adjustments, making rapid and automated switching between multiple specifications and models of medical device implants impossible.

[0004] It should be noted that the above content is only used to help understand the technical solution of the present invention, and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to propose a medical device implant testing system and method, which aims to enable testing of medical device implants of various specifications and models.

[0006] To achieve the above objectives, this invention proposes a method for testing medical device implants, comprising the following steps: Acquire multimodal data of the product to be inspected, wherein the multimodal data includes first two-dimensional image data and three-dimensional point cloud data of the product to be inspected; The first two-dimensional image data and the three-dimensional point cloud data are fused to form a high-dimensional feature vector; The high-dimensional feature vector is matched with the standard feature vector stored in the standard database to identify the specification and model information of the product to be inspected. Retrieve standard product data corresponding to the specification and model information from the standard database; The product to be inspected is tested based on the standard product data, and the test results of the product to be inspected are generated.

[0007] In one embodiment, the step of fusing features from the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector includes the following steps: A first set of geometric features is extracted from the first two-dimensional image data, wherein the first set of geometric features includes the shape contour, aspect ratio and hole distribution features of the product to be inspected; A second set of geometric features is extracted from the three-dimensional point cloud data. The second set of geometric features includes the overall size, volume and surface area features of the product to be inspected. The first set of geometric features and the second set of geometric features are concatenated and fused to form the high-dimensional feature vector.

[0008] In one embodiment, the step of matching the high-dimensional feature vector with pre-stored standard feature vectors in a standard database to identify the specification and model information of the product to be inspected includes the following steps: The similarity between the high-dimensional feature vector and the standard feature vector is calculated using a Siamese neural network model, and the standard feature vector with the highest similarity is identified as the specification and model information of the product to be inspected.

[0009] In one embodiment, the step of inspecting the product to be inspected based on the standard product data and generating the inspection result of the product to be inspected includes the following steps: Acquire second two-dimensional image data of the key surface area of ​​the product to be inspected, wherein the resolution of the second two-dimensional image data is higher than that of the first two-dimensional image data; The second two-dimensional image data is input into a deep learning segmentation model based on the U-Net architecture for analysis. The deep learning segmentation model performs pixel-level segmentation on the defects in the second two-dimensional image data, automatically identifies the defect type, and generates its location information. The defect type includes any one or more of scratches, dents, and rust spots. The defect type and location information are compared with the defect acceptance range pre-stored in the standard product data to generate the inspection result of the appearance defect inspection of the product to be inspected.

[0010] In one embodiment, the step of inspecting the product to be inspected based on the standard product data and generating the inspection result of the product to be inspected includes the following steps: The three-dimensional point cloud data is registered with the standard three-dimensional model in the standard product data by using an iterative nearest point algorithm. After registration, two key points corresponding to the size to be measured are located in the three-dimensional point cloud data, and the Euclidean distance between the two key points is calculated to obtain the actual measured size. The actual measured dimensions are compared with the pre-stored dimensional tolerance range in the standard three-dimensional model to generate the inspection results of the three-dimensional dimensional inspection of the product to be inspected.

[0011] In one embodiment, after the steps of inspecting the product to be tested based on the standard product data and generating the test results for the product to be tested, the medical device implant testing method includes the following steps: The coding information on the product to be inspected is read using optical character recognition technology; The encoded information is associated and bound with the test result and then output.

[0012] To achieve the above objectives, the present invention proposes a medical device implant testing system, the medical device implant testing system comprising: A multimodal perception module is used to acquire multimodal data of the product to be inspected, wherein the multimodal data includes first two-dimensional image data and three-dimensional point cloud data of the product to be inspected; The central processing module includes a memory storing the standard database; the central processing module is used to perform feature fusion on the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector; match the high-dimensional feature vector with the standard feature vectors pre-stored in the standard database to identify the specification and model information of the product to be inspected; retrieve standard product data corresponding to the specification and model information from the standard database; perform appearance defect inspection and three-dimensional dimension inspection on the product to be inspected based on the standard product data, and generate the inspection result of the product to be inspected.

[0013] In one embodiment, the system includes a global vision unit, a macro vision unit, and a three-dimensional measurement unit, wherein the global vision unit is used to acquire first two-dimensional image data of the product under inspection, the macro vision unit is used to acquire second two-dimensional image data of the product under inspection, and the three-dimensional measurement unit is used to acquire three-dimensional point cloud data of the product under inspection.

[0014] In one embodiment, the medical device implant testing system includes a robotic arm motion module, wherein the multimodal sensing module is integrated at the end of the robotic arm motion module; and wherein the robotic arm motion module includes a six-axis robotic arm.

[0015] In one embodiment, the medical device implant testing system includes an illumination module, which is used to automatically adjust the brightness, angle, and spectrum of the light source according to the specifications and model information of the product to be tested, wherein the specifications and model information includes the material and reflective properties of the product to be tested.

[0016] The technical solution of this invention achieves inspection operations for medical device implants of various specifications and models through automatic identification based on multimodal data fusion. Specifically, this invention can automatically identify the specifications and models of the products to be inspected and adaptively call the corresponding inspection standards, realizing an intelligent inspection process of "inspecting what comes in," minimizing manual intervention during production line changeovers, greatly improving production line efficiency, and perfectly matching the production mode of multiple varieties and small batches.

[0017] Secondly, by integrating two-dimensional image data and three-dimensional point cloud data for identification and inspection, this invention significantly improves the comprehensiveness and accuracy of the inspection. It can not only complete the inspection of appearance defects in two-dimensional vision, but also perform three-dimensional dimension inspection on complex three-dimensional dimensions, making the detection dimensions more comprehensive and the inspection results more reliable, thus reducing the risk of misjudgment caused by a single inspection method.

[0018] Furthermore, this invention integrates the testing of medical device implants into a seamless, automated switching process. The standard feature vectors of all products to be tested and their corresponding standard product data are pre-stored in the standard database, enabling the standard product data of the products to be tested to be automatically obtained from the standard database, thus achieving automated switching of the testing process. At the same time, it avoids the tedium and errors of manual testing, effectively improving the reliability of the test results. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of one embodiment of the medical device implant testing method provided by the present invention; Figure 2 A second flowchart illustrating the steps of an embodiment of the medical device implant testing method provided by the present invention; Figure 3 A third flowchart of an embodiment of the medical device implant testing method provided by the present invention; Figure 4 The fourth step of an embodiment of the medical device implant testing method provided by the present invention is shown in the flowchart below. Figure 5 Fifth step flowchart of an embodiment of the medical device implant testing method provided by the present invention; Figure 6 The sixth step flowchart of an embodiment of the medical device implant testing method provided by the present invention.

[0021] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] The technical solutions of this invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a portion of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0023] It should be noted that if the embodiments of the present invention involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0024] Furthermore, it should be noted that the descriptions involving "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed by this invention.

[0025] Medical implants (hereinafter referred to as "implants"), such as dental implants or cardiac stents, are high-value, high-risk Class III medical devices, and their quality directly affects patient safety. Therefore, factory inspection is a crucial step in the production process, typically including visual inspection (for scratches, burrs, stains, and whether the shape is intact) and dimensional inspection (whether critical dimensions are within tolerance). Due to individual patient differences and diverse clinical needs, implant products are often characterized by a wide variety and complex specifications, i.e., a large number of SKUs (Stock Keeping Units). In a flexible production model with multiple varieties and small batches, a single production line often needs to frequently switch between producing different specifications and models. Traditional inspection methods rely entirely on experienced inspectors to judge by sight, magnifying glasses, or simple optical measuring tools. With technological advancements, some companies are gradually adopting machine vision technology solutions, but these still have significant limitations, such as a lack of adaptability; when the production line switches to a new specification or model, inspectors must perform tedious system calibration and parameter adjustments, making it impossible to achieve rapid and automatic switching between multiple specifications and models of medical implants.

[0026] To address the aforementioned technical problems, this invention proposes a method for testing medical device implants.

[0027] Please see Figure 1 In one embodiment of the present invention, the medical device implant testing method includes the following steps: Step S10: Obtain multimodal data of the product to be inspected, wherein the multimodal data includes the first two-dimensional image data and the three-dimensional point cloud data of the product to be inspected; Step S20: Perform feature fusion on the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector; Step S30: Match the high-dimensional feature vector with the standard feature vector stored in the standard database to identify the specification and model information of the product to be inspected; Step S40: Retrieve standard product data corresponding to the specification and model information from the standard database; Step S50: Inspect the product to be inspected according to the standard product data, and generate the inspection result of the product to be inspected.

[0028] The technical solution of this invention achieves inspection operations for medical device implants of various specifications and models through automatic identification based on multimodal data fusion. Specifically, this invention can automatically identify the specifications and models of the products to be inspected and adaptively call the corresponding inspection standards, realizing an intelligent inspection process of "inspecting what comes in," minimizing manual intervention during production line changeovers, greatly improving production line efficiency, and perfectly matching the production mode of multiple varieties and small batches.

[0029] Secondly, by integrating two-dimensional image data and three-dimensional point cloud data for identification and inspection, this invention significantly improves the comprehensiveness and accuracy of the inspection. It can not only complete the inspection of appearance defects in two-dimensional vision, but also perform three-dimensional dimension inspection on complex three-dimensional dimensions, making the detection dimensions more comprehensive and the inspection results more reliable, thus reducing the risk of misjudgment caused by a single inspection method.

[0030] Furthermore, this invention integrates the testing of medical device implants into a seamless, automated switching process. The standard feature vectors of all products to be tested and their corresponding standard product data are pre-stored in the standard database, enabling the standard product data of the products to be tested to be automatically obtained from the standard database, thus achieving automated switching of the testing process. At the same time, it avoids the tedium and errors of manual testing, effectively improving the reliability of the test results.

[0031] As a preferred embodiment of the above embodiments, refer to Figure 2 Step S20 includes the following steps: Step S21: Extract a first set of geometric features from the first two-dimensional image data, wherein the first set of geometric features includes the shape contour, aspect ratio, and hole distribution features of the product to be inspected; Step S22: Extract a second set of geometric features from the three-dimensional point cloud data. The second set of geometric features includes the overall size, volume, and surface area features of the product to be inspected. Step S23: The first geometric feature set and the second geometric feature set are concatenated and fused to form the high-dimensional feature vector.

[0032] This setup extracts macroscopic morphological features such as shape contours and aspect ratios from the first-dimensional image data, and spatial geometric features such as overall size and volume from the three-dimensional point cloud data. These two features are then combined to construct a high-dimensional feature vector that comprehensively represents the product's identity. This multi-source heterogeneous feature fusion strategy significantly improves the system's ability to represent complex implant products, laying a solid foundation for subsequent accurate identification and effectively avoiding the risk of misidentification due to single features.

[0033] As a preferred embodiment of the above embodiments, refer to Figure 3 Step S30 includes the following steps: Step S31: Calculate the similarity between the high-dimensional feature vector and the standard feature vector using a Siamese neural network model, and identify the standard feature vector with the highest similarity as the specification and model information of the product to be inspected.

[0034] This setup, limiting similarity matching to a Siamese neural network model, is key to achieving high-precision and rapid recognition. Through deep metric learning, this model maintains excellent discriminative ability even with small sample sizes, making it particularly suitable for the multi-specification, small-batch production characteristics of medical devices. Its end-to-end computation mode not only ensures recognition efficiency but also captures subtle feature differences through nonlinear mapping, enabling the system to maintain extremely high recognition accuracy even when faced with products of different specifications and similar shapes.

[0035] Among them, the method for calculating the similarity between the features of the product to be inspected and the features in the database using the Siamese neural network model is as follows: Similarity = 1 - D(F_current, F_database_i); Where F_current is the current product feature vector, F_database_i is the feature vector of the i-th specification in the database, and D is the distance function (such as cosine distance). The system selects the specification with the highest similarity as the recognition result. This process achieves what you see is what you measure, without requiring manual specification of the product model.

[0036] As a preferred embodiment of the above embodiments, refer to Figure 4 Step S50 includes the following steps: Step S51: Acquire second two-dimensional image data of the key surface areas of the product to be inspected, wherein the resolution of the second two-dimensional image data is higher than that of the first two-dimensional image data; In the above steps, a high-resolution macro camera is precisely moved to each key surface area of ​​the product to be inspected, and second two-dimensional image data with a resolution much higher than that of the images used in the recognition stage is acquired. Step S52: The second two-dimensional image data is input into a deep learning segmentation model based on the U-Net architecture for analysis. The deep learning segmentation model performs pixel-level segmentation on the defects in the second two-dimensional image data, automatically identifies the defect type, and generates its location information. The defect types include any one or more of scratches, dents, and rust spots. In the above steps, the acquired second two-dimensional image data is input into a deep learning segmentation model based on the U-Net architecture for analysis. The power of this model lies in its pixel-level segmentation, enabling it to accurately delineate the contour, location, and extent of each tiny defect and automatically identify its type (such as scratches, dents, and rust spots). This achieves a leap from qualitative judgment to quantitative analysis, with accuracy and consistency far exceeding manual visual inspection. The U-Net architecture is a convolutional neural network based on deep learning, primarily used for image segmentation tasks.

[0037] Step S53: Compare the defect type and location information with the pre-stored defect acceptance range in the standard product data to generate an inspection result for the appearance defect of the product to be inspected. In the above step, the identified defect type and location information are automatically compared with the pre-stored, quantified defect acceptance range in the standard database (e.g., scratches with a length not exceeding 50 micrometers and a depth not exceeding 5 micrometers are allowed). Based on this, the system generates an objective "pass / fail" inspection result. This step replaces the decision-making process that relies on personal experience and subjective feelings in manual inspection, ensuring absolute uniformity of judgment standards and data traceability, and completely eliminating misjudgments and omissions caused by personnel fatigue or experience differences.

[0038] With this setup, this embodiment mainly performs appearance defect inspection on the product to be inspected. By adopting a deep learning segmentation model based on the U-Net architecture, pixel-level accurate positioning and classification of surface defects are achieved, transforming traditional subjective qualitative judgment into objective quantitative analysis, which significantly improves the accuracy and consistency of appearance inspection.

[0039] As a preferred embodiment of the above embodiments, refer to Figure 5 Step S50 includes the following steps: Step S54: Register the 3D point cloud data with the standard 3D model in the standard product data using an iterative nearest-point algorithm. In the above step, the iterative nearest-point algorithm precisely rotates and translates the 3D point cloud data of the product to be inspected, which is acquired in real time and has a random spatial position, with the ideal standard 3D model in the standard database until the two match optimally in 3D space. This step is equivalent to perfectly superimposing a real part with its design blueprint in virtual space, eliminating measurement errors caused by inconsistent product placement positions and angles, and establishing a unified and accurate spatial coordinate system benchmark for subsequent precise measurements. Step S55: Locate two key points corresponding to the dimension to be measured in the registered 3D point cloud data, and calculate the Euclidean distance between the two key points to obtain the actual measured dimension. In the above steps, on the registered point cloud model, the system automatically locates two key points corresponding to the dimension to be measured according to the preset measurement task (for example, to measure the screw diameter, locate the points opposite to the two sides of the thread; to measure the length, locate the points at the beginning and end). Subsequently, the system calculates the straight-line distance between these two points using the Euclidean distance formula to obtain the actual measured dimension. This essentially realizes the function of a vernier caliper or coordinate measuring machine in the physical world in a digital 3D model, achieving true non-contact 3D measurement. Among them, the spatial values ​​of two key points are defined as P1(x1,y1,z1) and P2(x2,y2,z2), and the Euclidean distance formula is: actual measured size = √[(x2-x1)²+(y2-y1)²+(z2-z1)²]; Step S56: Compare the actual measured dimensions with the pre-stored dimensional tolerance range in the standard 3D model to generate the inspection result of the 3D dimensional inspection of the product to be inspected. In the above step, the system automatically compares the calculated actual measured dimensions with the pre-stored dimensional tolerance range in the standard 3D model. Based on the comparison result, the system generates an objective conclusion of "qualified / unqualified". This step replaces the visual errors and subjective judgments that may occur when manually reading and comparing tolerance zones, ensuring the accuracy, consistency, and traceability of the dimensional inspection results, and providing crucial data assurance for the functional safety of medical device implants.

[0040] With this setup, this embodiment mainly performs three-dimensional dimensional inspection on the product under inspection. It achieves accurate registration of three-dimensional point cloud data with standard three-dimensional models through an iterative nearest point algorithm, and performs dimensional calculations based on Euclidean distance. This solves the problem of accurate measurement of complex three-dimensional implant structures, enabling the measurement results to reach metrological-grade accuracy.

[0041] As a preferred embodiment of the above embodiments, refer to Figure 6 Step S50 is followed by the following steps: Step S60: Read the coding information on the product to be inspected using optical character recognition technology. In the above steps, after the product to be inspected completes physical inspection, the system controls the camera to aim at the coding area (such as the laser-engraved serial number or batch number) on the product surface or label, and automatically reads the character information using optical character recognition technology. This step replaces the tedious, error-prone, and inefficient process of inspectors manually recording or typing codes after inspection, achieving contactless and automated collection of product identification information and ensuring the accuracy and efficiency of raw data acquisition.

[0042] Step S70: Associate and bind the encoded information with the inspection results and output them. In the above steps, the system automatically binds the automatically read encoded information (the product's unique identification card) with all the inspection results generated in the previous steps (including specific data such as appearance and dimensions), and automatically uploads it as a complete data packet to the factory's MES (Manufacturing Execution System) or QMS (Quality Management System) via industrial communication protocols (such as OPCUA, ModbusTCP). This ensures that each inspection report can be accurately mapped to each specific product, thereby constructing a complete and reliable traceability chain from a single product to all its production and inspection data.

[0043] This setup, through the added automatic identification and association steps of coding information, achieves automatic binding of product identity information and inspection results, completely eliminating the manual data entry step. This not only improves efficiency but also ensures the integrity and accuracy of the data chain, providing a reliable guarantee for quality traceability throughout the entire lifecycle.

[0044] To address the aforementioned technical problems, this invention proposes a medical device implant testing system.

[0045] In one embodiment of the present invention, the medical device implant testing system includes: A multimodal perception module is used to acquire multimodal data of the product to be inspected, wherein the multimodal data includes first two-dimensional image data and three-dimensional point cloud data of the product to be inspected; The central processing module includes a memory storing the standard database; the central processing module is used to perform feature fusion on the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector; match the high-dimensional feature vector with the standard feature vectors pre-stored in the standard database to identify the specification and model information of the product to be inspected; retrieve standard product data corresponding to the specification and model information from the standard database; perform appearance defect inspection and three-dimensional dimension inspection on the product to be inspected based on the standard product data, and generate the inspection result of the product to be inspected.

[0046] The specific steps for the testing method of medical device implants can be referred to the above embodiments. Since this medical device implant testing system adopts all the technical solutions of all the above embodiments, it possesses at least all the beneficial effects brought about by the technical solutions of the above embodiments, and will not be elaborated further here.

[0047] As a preferred embodiment of the above, the multimodal perception module includes a global vision unit, a macro vision unit, and a 3D measurement unit. The global vision unit acquires first two-dimensional image data of the product under inspection, the macro vision unit acquires second two-dimensional image data of the product under inspection, and the 3D measurement unit acquires 3D point cloud data of the product under inspection. With this configuration, the global vision unit is responsible for acquiring the first two-dimensional image data of the product. Its core task is to quickly capture the overall outline, basic shape, and general layout of the product under inspection for a preliminary and rapid judgment of its specifications and model, providing crucial information for subsequent specification and model identification and preliminary positioning. The macro vision unit acquires high-resolution second two-dimensional image data. Under the precise guidance of the robotic arm motion module, it approaches key surface areas (such as thread grooves and joint contact surfaces) of the product under inspection, capturing microscopic surface details. This is equivalent to using a high-definition magnifying glass to carefully examine potential flaws, serving as the sole data source for subsequent identification of scratches, dents, and other microscopic defects. The 3D measurement unit (such as a lidar) acquires the 3D point cloud data of the product under inspection. This allows the product under inspection to form a digital shell composed of millions of three-dimensional coordinate points, accurately reflecting the surface irregularities of the object. It endows the product with precise three-dimensional spatial perception capabilities, enabling the quantification of all spatial geometric features of the product, such as length, width, height, volume, and surface curvature, thereby achieving accurate, non-contact measurement of complex three-dimensional dimensions.

[0048] As a preferred embodiment of the above embodiments, the medical device implant inspection system includes a robotic arm motion module, and the multimodal sensing module is integrated at the end of the robotic arm motion module; wherein the robotic arm motion module includes a six-axis robotic arm. This configuration integrates the multimodal sensing module (including a global camera, a macro camera, and a 3D measurement unit) at the end of the six-axis robotic arm, enabling it to move and rotate at any angle in three-dimensional space. This allows the multimodal sensing module integrated at the end to no longer passively wait for the product to be inspected to move to a fixed position, but rather to actively and from multiple angles approach the product—whether it's performing a macroscopic scan around the product to identify the model, precisely aiming the macro camera at a tiny threaded surface for defect detection, or guiding the 3D measuring instrument to scan along an optimal path. This proactive sensing method completely eliminates the reliance of traditional testing systems on precision fixtures and fixed tooling, thereby greatly improving the system's adaptability to medical device implants of different specifications, sizes, and testing requirements. A single system can complete the entire complex process from "macroscopic identification" to "microscopic detection," which is the core guarantee for achieving efficient quality inspection under the flexible production model of multiple varieties and small batches.

[0049] As a preferred embodiment of the above, the medical device implant inspection system includes an illumination module. This module automatically adjusts the brightness, angle, and spectrum of the light source based on the specifications and model information of the product under inspection, including the material and reflective properties of the product. This design takes into account that stable lighting is as important as sufficient and suitable lighting for human observation in visual inspection. However, industrial environments are complex, especially since medical device implants often contain reflective materials such as metals and vary in shape. Fixed lighting conditions can easily cause overexposure, reflection, or shadows, leading to image feature loss or distortion, directly affecting the accuracy of subsequent identification and inspection. The core value of this illumination module lies in its dynamic adjustment capability: once the system identifies and determines the specifications and model of the product under inspection, it can immediately retrieve the material and surface characteristic data of that model from the database (such as whether it is a high-reflectivity titanium alloy, matte coating, or ceramic), and automatically and accurately adjust the brightness (avoiding excessive brightness and whitening or excessive darkness and noise), angle (avoiding specular reflection spots and highlighting contours or textures) and spectrum (selecting the most suitable light color for the material to enhance image contrast) of the light source. This is equivalent to equipping the system with a pair of "eyes" that can automatically adapt to any observation environment. No matter how the product changes, it can provide the most suitable lighting, thereby always obtaining the clearest and most distinctive images. This provides a reliable guarantee for all subsequent vision-based accurate identification, defect detection, and dimensional measurement, improving the robustness and detection accuracy of the entire system from the source.

[0050] It should be noted that other aspects of the medical device implant testing system and testing method disclosed in this invention are prior art and will not be elaborated here.

[0051] The above are merely optional embodiments of the present invention and do not limit the patent scope of the present invention. Any application of the present invention directly or indirectly in other related technical fields is included within the patent protection scope of the present invention.

Claims

1. A method for testing medical device implants, characterized in that: Includes the following steps: Acquire multimodal data of the product to be inspected, wherein the multimodal data includes first two-dimensional image data and three-dimensional point cloud data of the product to be inspected; The first two-dimensional image data and the three-dimensional point cloud data are fused to form a high-dimensional feature vector; The high-dimensional feature vector is matched with the standard feature vector stored in the standard database to identify the specification and model information of the product to be inspected. Retrieve standard product data corresponding to the specification and model information from the standard database; The product to be inspected is tested based on the standard product data, and the test results of the product to be inspected are generated.

2. The medical device implant testing method as described in claim 1, characterized in that: The step of fusing features from the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector includes the following steps: A first set of geometric features is extracted from the first two-dimensional image data, wherein the first set of geometric features includes the shape contour, aspect ratio and hole distribution features of the product to be inspected; A second set of geometric features is extracted from the three-dimensional point cloud data. The second set of geometric features includes the overall size, volume and surface area features of the product to be inspected. The first set of geometric features and the second set of geometric features are concatenated and fused to form the high-dimensional feature vector.

3. The medical device implant testing method as described in claim 1, characterized in that: The step of matching the high-dimensional feature vector with the standard feature vectors pre-stored in the standard database to identify the specification and model information of the product to be inspected includes the following steps: The similarity between the high-dimensional feature vector and the standard feature vector is calculated using a Siamese neural network model, and the standard feature vector with the highest similarity is identified as the specification and model information of the product to be inspected.

4. The medical device implant testing method as described in claim 1, characterized in that: The steps of inspecting the product to be inspected based on the standard product data and generating the inspection result of the product to be inspected include the following steps: Acquire second two-dimensional image data of the key surface area of ​​the product to be inspected, wherein the resolution of the second two-dimensional image data is higher than that of the first two-dimensional image data; The second two-dimensional image data is input into a deep learning segmentation model based on the U-Net architecture for analysis. The deep learning segmentation model performs pixel-level segmentation on the defects in the second two-dimensional image data, automatically identifies the defect type, and generates its location information. The defect types mentioned include any one or more of scratches, pits, and rust spots; The defect type and location information are compared with the defect acceptance range pre-stored in the standard product data to generate the inspection result of the appearance defect inspection of the product to be inspected.

5. The medical device implant testing method as described in claim 1, characterized in that: The steps of inspecting the product to be inspected based on the standard product data and generating the inspection result of the product to be inspected include the following steps: The three-dimensional point cloud data is registered with the standard three-dimensional model in the standard product data by using an iterative nearest point algorithm. After registration, two key points corresponding to the size to be measured are located in the three-dimensional point cloud data, and the Euclidean distance between the two key points is calculated to obtain the actual measured size. The actual measured dimensions are compared with the pre-stored dimensional tolerance range in the standard three-dimensional model to generate the inspection results of the three-dimensional dimensional inspection of the product to be inspected.

6. The method for testing medical device implants as described in any one of claims 2 to 5, characterized in that: After the steps of inspecting the product to be tested based on the standard product data and generating the test results for the product to be tested, the medical device implant testing method includes the following steps: The coding information on the product to be inspected is read using optical character recognition technology; The encoded information is associated and bound with the test result and then output.

7. A medical device implant testing system, characterized in that: The medical device implant testing system includes: A multimodal perception module is used to acquire multimodal data of the product to be inspected, wherein the multimodal data includes first two-dimensional image data and three-dimensional point cloud data of the product to be inspected; The central processing module includes a memory storing a standard database. The central processing module is used to perform feature fusion on the first two-dimensional image data and the three-dimensional point cloud data to form a high-dimensional feature vector; match the high-dimensional feature vector with pre-stored standard feature vectors in the standard database to identify the specification and model information of the product to be inspected; retrieve standard product data corresponding to the specification and model information from the standard database; perform appearance defect inspection and three-dimensional dimension inspection on the product to be inspected based on the standard product data, and generate the inspection result of the product to be inspected.

8. The medical device implant testing system as described in claim 7, characterized in that: The multimodal perception module includes a global vision unit, a macro vision unit, and a three-dimensional measurement unit. The global vision unit is used to acquire first two-dimensional image data of the product under inspection, the macro vision unit is used to acquire second two-dimensional image data of the product under inspection, and the three-dimensional measurement unit is used to acquire three-dimensional point cloud data of the product under inspection.

9. The medical device implant testing system as described in claim 7, characterized in that: The medical device implant testing system includes a robotic arm motion module, and the multimodal sensing module is integrated at the end of the robotic arm motion module; wherein the robotic arm motion module includes a six-axis robotic arm.

10. The medical device implant testing system according to any one of claims 7 to 9, characterized in that: The medical device implant testing system includes an illumination module, which is used to automatically adjust the brightness, angle, and spectrum of the light source according to the specifications and model information of the product to be tested, wherein the specifications and model information includes the material and reflective properties of the product to be tested.

Images