A medical device intelligent identification method, device, equipment and medium

Abstract

This invention relates to the fields of computer vision and medical device management technology. Addressing the problem of insufficient accuracy in surgical instrument recognition, it discloses a method, device, equipment, and medium for intelligent medical device recognition. Upon receiving a recognition command, a first image is acquired promptly to obtain comprehensive initial visual information. Combined with dual-polarization acquisition and pixel difference technology, specular highlight and diffuse reflection areas are accurately separated, improving image clarity. A first analysis model quickly determines confirmed and uncertain recognition areas, focusing on acquiring a second image of the uncertain areas to reduce redundant acquisition and improve data efficiency and accuracy. The image, after glare removal and optimization, possesses clearer feature information, and further analysis significantly improves the recognition accuracy of uncertain areas. Finally, the first and second confirmed recognition results are fused, comprehensively improving the accuracy and reliability of medical device recognition, thereby effectively solving the technical problem of insufficient accuracy in existing intelligent medical device recognition methods.

Description

Technical Field

[0001] This invention relates to the fields of computer vision and medical device management technology, and in particular to a method, device, equipment and medium for intelligent identification of medical devices. Background Technology

[0002] In the existing closed-loop management process for surgical instruments, medical devices typically follow a flow path of "issued by the sterilization supply center—used in the operating room—retrieved by the sterilization supply center." The specific process includes preoperative preparation in the supply room, distribution of sterilization packs, instrument counting during surgery, and verification after retrieval by the supply room. Each of these steps heavily relies on human experience to verify the type, quantity, and integrity of the instruments. Especially during the operating room counting and supply room verification stages, each instrument typically needs to be checked individually, which is tedious, time-consuming, and prone to human error leading to omissions or errors, potentially impacting surgical safety and the accuracy of traceability management. With the development of image recognition and detection technologies, some medical device management processes have begun to explore the introduction of assisted identification schemes based on intelligent analysis models (such as deep neural network models) to correct omissions, redundancies, and incorrect placement of medical devices, thereby improving the accuracy and efficiency of instrument counting and verification.

[0003] However, in the existing automated instrument inventory, traceability and deep learning annotation process based on image recognition, different instrument materials react differently to complex ambient light, and it is difficult to predict accurately. For example, instruments made of stainless steel, titanium alloy and ceramic often have strong specular highlights and diffuse reflection under varying lighting conditions, which can easily lead to overexposure in local areas, obscuring the surface texture information of the instrument. As a result, the target detection model may miss detections, affecting the accuracy and stability of recognition, and making it difficult to meet the high standards of clinical management. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and medium for intelligent identification of medical devices, in order to solve the technical problem of low accuracy in intelligent identification of medical devices.

[0005] Firstly, a method for intelligent identification of medical devices is provided, including:

[0006] Upon receiving a medical device identification command from the user, the system controls the image acquisition unit to acquire the first image of the medical device.

[0007] The first image is input into a pre-determined first analysis model and judged according to a pre-set confidence threshold to obtain a first determined image recognition result and a first uncertain image recognition result.

[0008] For the local target area corresponding to the first uncertain image recognition result, the image acquisition unit is controlled to acquire the second image of the medical device, and the second image is subjected to anti-glare optimization processing to obtain an anti-glare image;

[0009] The glare-free image is input into the first analysis model to obtain the second definitive image recognition result of the medical device.

[0010] Based on the first determined image recognition result and the second determined image recognition result, the target medical device corresponding to the recognition instruction is determined.

[0011] Secondly, a medical device intelligent identification device is provided, comprising:

[0012] The first image acquisition module is used to control the image acquisition unit to acquire the first image of the medical device after receiving the medical device identification command issued by the user.

[0013] The image recognition module is used to input the acquired first image into a predetermined first analysis model and judge it according to a preset confidence threshold to obtain a first certain image recognition result and a first uncertain image recognition result;

[0014] The image optimization processing module is used to control the image acquisition unit to acquire a second image of the medical device for the local target area corresponding to the first uncertain image recognition result, and to perform anti-glare optimization processing on the second image to obtain a glare-free image.

[0015] The second image recognition result analysis module is used to input the glare-free image into the first analysis model to obtain the second image recognition result of the medical device.

[0016] A medical device identification module is used to identify the target medical device corresponding to the identification command based on the first image recognition result and the second image recognition result.

[0017] Thirdly, a medical device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described intelligent identification method for medical devices.

[0018] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described intelligent identification method for medical devices.

[0019] This invention acquires a first image promptly upon receiving a recognition command, obtaining comprehensive initial visual information. Combining dual-polarization acquisition and pixel difference technology, it accurately separates specular highlight and diffuse reflection areas, improving image clarity. A first analysis model quickly determines confirmed and uncertain recognition areas, focusing on acquiring a second image of the uncertain areas to reduce redundant acquisition and improve data efficiency and accuracy. The image, after glare removal and optimization, possesses clearer feature information, and further analysis significantly improves the recognition accuracy of uncertain areas. Finally, by fusing the first and second confirmed recognition results, the accuracy and reliability of medical device recognition are comprehensively improved, effectively solving the technical problem of insufficient accuracy in existing intelligent medical device recognition. Attached Figure Description

[0020] To facilitate the explanation of the technical solutions of the embodiments of the present invention, the accompanying drawings referenced in the embodiments are briefly described below. It should be understood that the following drawings only show some typical embodiments of the present invention, and those skilled in the art can obtain other forms of drawings based on this without creative effort.

[0021] Figure 1 This is a schematic diagram of an application environment for the intelligent identification method for medical devices according to an embodiment of the present invention;

[0022] Figure 2 This is a flowchart illustrating a medical device intelligent identification method according to an embodiment of the present invention;

[0023] Figure 3 This is a schematic diagram of the dual-camera and polarization hardware assembly in one embodiment of the present invention;

[0024] Figure 4a This is a schematic diagram of an image acquired with a polarization angle of 0° in one embodiment of the present invention;

[0025] Figure 4b This is a schematic diagram of an image acquisition with a polarization angle of 90° in one embodiment of the present invention;

[0026] Figure 4c This is a schematic diagram of the acquired image fused with anti-glare technology in one embodiment of the present invention;

[0027] Figure 5a This is a schematic diagram of a dual-polarization image acquisition in one embodiment of the present invention;

[0028] Figure 5b This is a schematic diagram of the image after the glare removal effect in one embodiment of the present invention;

[0029] Figure 6 This is a schematic diagram of the final verification set in one embodiment of the present invention;

[0030] Figure 7 This is a schematic diagram of a normalized confusion matrix in one embodiment of the present invention;

[0031] Figure 8 This is a schematic diagram of a medical device area marking in one embodiment of the present invention;

[0032] Figure 9a This is a schematic diagram illustrating the consistency between the instrument photograph recognition and the count difference verification interface in one embodiment of the present invention;

[0033] Figure 9b This is a schematic diagram illustrating the inconsistency between the instrument photograph recognition and the count difference verification interface in one embodiment of the present invention;

[0034] Figure 10 This is a schematic diagram of the structure of a medical device intelligent identification device according to an embodiment of the present invention;

[0035] Figure 11 This is a schematic diagram of the structure of a medical device according to one embodiment of the present invention;

[0036] Figure 12 This is another structural schematic diagram of the medical device in one embodiment of the present invention. Detailed Implementation

[0037] 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, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] The intelligent identification method for medical devices provided in this invention can be applied to, for example... Figure 1In this application environment, the client communicates with the server via a network. Upon receiving a medical device identification command from the client, the server controls the image acquisition unit to acquire a first image of the medical device. The acquired first image is input into a pre-defined first analysis model and judged according to a pre-set confidence threshold, yielding a first confirmed image recognition result and a first uncertain image recognition result. For the local target area corresponding to the first uncertain image recognition result, the server controls the image acquisition unit to acquire a second image of the medical device and performs glare reduction optimization processing on the second image to obtain a glare-free image. Subsequently, the glare-free image is input into the first analysis model to obtain a second confirmed image recognition result of the medical device. Based on the first and second confirmed image recognition results, the target medical device corresponding to the user's medical device identification command is finally determined, and the recognition result is returned to the client. This invention improves image clarity by promptly acquiring the first image after receiving the identification command, obtaining comprehensive initial visual information, and accurately separating specular highlight and diffuse reflection areas by combining dual-polarization acquisition and pixel difference technology. The first analysis model quickly determines confirmed and uncertain recognition areas, focusing on acquiring a second image of the uncertain area, reducing redundant acquisition and improving data efficiency and accuracy. The image, after glare reduction and optimization, possesses clearer feature information, and further analysis significantly improves the accuracy of identifying uncertain areas. Finally, the first and second identification results are fused to comprehensively improve the accuracy and reliability of medical device identification, thereby effectively solving the technical problem of insufficient accuracy in existing intelligent medical device identification. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will be described in detail below through specific embodiments.

[0039] Please see Figure 2 As shown, Figure 2 A flowchart illustrating the intelligent identification method for medical devices provided in this embodiment of the invention includes the following steps:

[0040] S1. After receiving the medical device identification command from the user, control the image acquisition unit to acquire the first image of the medical device.

[0041] A typical application scenario for this invention is the current closed-loop management process for surgical instruments, where instrument circulation follows a chain of "issued by the sterilization supply center—used in the operating room—retrieved by the sterilization supply center." At each stage, the type, quantity, and condition of instruments are primarily checked manually, which carries risks of low efficiency, omissions, and errors, making it difficult to guarantee the accuracy of instrument identification. Therefore, this invention proposes to use an image acquisition unit in conjunction with an intelligent recognition algorithm to assist in the identification and verification of medical instruments. For example, during preoperative preparation in the supply room, the supply room nurse checks the model, quantity, and condition of each instrument against a paper "Surgical Instrument Inventory Sheet," and then completes packaging and sterilization. During the turnover and distribution process, the sterilized packages are placed in a sterile temporary storage warehouse and then delivered to the corresponding operating room according to the surgical schedule. The operating room conducts three counts: the scrub nurse and circulating nurse count and sign each instrument before surgery, before closure of the cavity, and after surgery. Taking a 4-6 hour tertiary surgery as an example, it usually takes two nurses a total of 30-40 minutes to complete the count. The loss of any instrument may lead to delayed closure of the cavity or a second exploration. The supply room re-checks the instruments after they are returned to the contaminated area, recording any missing or repairable items, and then proceeds to the next round of cleaning-packaging-sterilization. All of the above processes rely on manual verification, which is inefficient, prone to omissions and errors, and makes it difficult to guarantee the accuracy of instrument identification. Therefore, it is necessary to use medical equipment to achieve intelligent identification and assisted verification of medical instruments.

[0042] In this embodiment of the invention, dual cameras are installed at a sterile safety height of 40-50cm above the instrument (fixed by a sterile magnetic bracket or disposable plastic clamp to ensure that they do not contact the sterile area). This provides a convenient intelligent medical device identification solution for clinical use without the need to modify the operating table or counting table.

[0043] In practice, users can actively issue medical device identification commands through any of the following methods:

[0044] Through preset physical controls on the front-end image acquisition unit;

[0045] Through the preset touch controls on the display interface of the front-end image acquisition unit;

[0046] Commands are issued through the human-computer interaction interface on the control terminal via the backend server.

[0047] Upon receiving the identification command, the image acquisition unit is controlled to acquire the first image of the medical device. The image acquisition unit can be installed in the medical device inspection area, the inventory area of ​​the sterilization supply center, or the medical device placement area of ​​the operating room. It is fixed 40-50cm above the sterile area using a sterile magnetic bracket or disposable clamp, facilitating deployment without disrupting the sterile environment. This first image is the valid medical device image exposed and output by the acquisition unit.

[0048] In this embodiment of the invention, the image acquisition unit is controlled by a terminal and interacts with the server to issue a shooting command. The step of controlling the image acquisition unit to acquire a first image of the medical device after receiving a medical device identification command from the user includes:

[0049] The target recognition area is determined based on user instructions;

[0050] Send shooting instructions to the image acquisition unit in the area;

[0051] The camera unit is controlled according to instructions to acquire the first image of the medical device.

[0052] In detail, the medical device identification area refers to the physical space range that the image acquisition unit needs to focus on capturing. It directly affects the integrity and identification efficiency of the medical devices in the acquired images, such as the medical device inspection area, the sterilization supply center inventory area, or the medical device placement area in the operating room. To improve efficiency, when the identification command involves multiple areas and multiple devices, the system can simultaneously trigger multiple camera devices to acquire data in parallel, reducing operation latency.

[0053] Specifically, a camera_trigger topic message is sent via the MQTT protocol, carrying the area ID and shooting parameters, such as {"area_id":"OR-1-01", "mode":"high_res"}. This calls the camera SDK's capture() interface, directly triggering the image sensor to start frame acquisition, thereby obtaining the image corresponding to the medical device within the identified area.

[0054] Furthermore, the accuracy of medical device identification is directly related to surgical safety; failure to identify the device could lead to serious consequences such as leaving the device inside the body. Therefore, a "zero-tolerance" policy for identification errors is necessary in medical settings. This invention reduces the risk of missed or incorrect detections and improves identification reliability through precise data collection and identification.

[0055] S2. Input the acquired first image into a predetermined first analysis model and judge it according to a preset confidence threshold to obtain the first determined image recognition result and the first uncertain image recognition result.

[0056] In this embodiment of the invention, the first analysis model is a deep learning model trained on massive amounts of medical device images. It has a fixed network structure and weight parameters, typically using a lightweight YOLO-v8 network, and is usually trained through supervised learning, for example, using a cross-entropy loss function. This model is used for feature extraction and classification of medical device images. Its backbone is based on a simplified CSPDarknet structure, capable of extracting features such as edges and textures from images through multi-layer convolutional networks, and fusing feature maps of different scales to adapt to the detection needs of medical devices of different sizes. Finally, the model normalizes through fully connected layers and softmax layers, outputting the probability distribution vector of each medical device category corresponding to the first image. The first definitive image recognition result refers to the high-confidence recognition result output by the model, which can be directly used as a valid recognition result. The first uncertain image recognition result refers to the low-confidence result output by the model with ambiguous category distinctions, which cannot be used directly and requires manual review or a secondary recognition process.

[0057] In this embodiment of the invention, the step of inputting the acquired first image into a pre-determined first analysis model and judging it according to a pre-set confidence threshold to obtain a first determined image recognition result and a first uncertain image recognition result includes:

[0058] The first image is input into the first analysis model, and the probability distribution vector corresponding to the first image is output.

[0059] When the highest probability value in the probability distribution vector is greater than or equal to the preset confidence threshold, the medical device category corresponding to the highest probability value is determined as the target device category corresponding to the first image, and a first determined image recognition result is generated based on the target device category.

[0060] Calculate the difference between the highest probability value and the second highest probability value in the probability distribution vector. When the highest probability value is less than the preset confidence threshold and the difference is less than the preset difference threshold, determine the medical device category corresponding to the highest probability value as the candidate device category corresponding to the first image, and generate a first uncertain image recognition result based on the candidate device category.

[0061] In detail, the first analysis model refers to a deep learning model pre-trained with massive amounts of medical device images. It typically uses a lightweight YOLO-v8 network, whose backbone is based on a simplified CSPDarknet structure. Its core is to efficiently extract image features, that is, to have efficient image feature extraction capabilities. The image tensor of the first image is shaped and then the image edge and texture features are extracted through convolutional layers. That is, after the first image enters the network through the input layer, it first goes through multiple convolutional blocks to gradually compress the spatial dimension and increase the number of channels to capture low-level features such as edges and textures. The high-level semantic features output by the backbone are passed down and fused with the mid-level and low-level features to compensate for the feature loss caused by the lightweight structure. Finally, three fused feature maps are generated, which are suitable for the detection needs of medical devices of different sizes. After the feature maps are mapped by fully connected layers, they are normalized by softmax layers to output the probability distribution of the medical device category corresponding to the image.

[0062] Specifically, both the preset reliability threshold and the difference threshold are system presets and can be determined through experience or experimentation based on the distribution characteristics of the training data and the accuracy requirements of actual medical scenarios. The preset reliability threshold is typically set between 0.6 and 0.9 to distinguish the reliability of the model output results, and the confidence threshold can be dynamically adjusted according to the risk tolerance and validation set data distribution of different business scenarios. For example, in scenarios with extremely high safety requirements, such as "terminal counting before cavity closure," the confidence threshold is usually increased to 0.85–0.90 to minimize the risk of missed detections; while in scenarios with relatively low risk and high efficiency requirements, such as batch warehousing in the sterilization supply center, the threshold can be set between 0.60 and 0.70 to achieve faster turnover. The final value of the threshold can be determined by combining the true positive rate-false positive rate curve analysis, on-site experimental results, and business error tolerance requirements. Results higher than the threshold can be directly used for inventory management or counting records, while results lower than the threshold need to enter a manual review or secondary identification process. The difference threshold is used to help determine the model's discriminative power when the probabilities of multiple categories are close, avoiding forced determination under low confidence levels and improving safety and accuracy.

[0063] Furthermore, if the highest probability output by the first analysis model is greater than or equal to the preset confidence threshold, and the category distinction is clear, it is determined as a definite result. That is, the maximum value in the probability distribution and its corresponding category are queried, and the highest probability in the probability distribution vector is verified to be greater than or equal to the confidence threshold. If it is satisfied, the corresponding category is determined as the target device category, and the image corresponding to the target device category is determined as the first definite image recognition result. This recognition result can be directly used as effective data for business processes such as inventory management and inventory records, reducing manual intervention and improving work efficiency. If the highest probability value is lower than the threshold and the difference between it and the second highest probability value is small, it indicates that the model's feature discrimination ability for the image is insufficient, which may be due to similar device appearances (such as different models of hemostats) or poor image quality (such as glare obscuring texture). At this point, by filtering using a preset probability difference threshold, inaccurate judgments can be avoided, reducing medical risks. For example, if the highest probability value in the probability distribution vector is greater than 0.8 and the category is obvious, it can be directly judged as the first certain image recognition result and used for automatic registration, improving counting efficiency. If the highest probability value is lower than the preset confidence threshold and is close to the second highest probability value, it indicates that the model's discrimination ability is insufficient (such as similar appearance of instruments or poor image quality), and it is judged as the first uncertain image recognition result, which needs to enter the next step of image optimization and recognition process to reduce medical risks.

[0064] Furthermore, for the first uncertain image recognition result, subsequent image optimization and re-analysis are needed to obtain a more accurate medical device recognition result, ensuring the accuracy and reliability of the medical device recognition process and avoiding omissions or misjudgments.

[0065] S3. For the local target area corresponding to the first uncertain image recognition result, control the image acquisition unit to acquire the second image of the medical device, and perform anti-glare optimization processing on the second image to obtain a glare-free image.

[0066] In this embodiment of the invention, the local target area refers to the actual physical storage space of the medical device in the clinical environment, such as a fixed area (e.g., an instrument rack in the supply room) or a dynamic area (e.g., a movable instrument table in the operating room). The second image refers to an image that is collected again to supplement identification information after the first uncertain result is generated. It contains at least two viewing angles with different polarization angles or viewing angles with different physical positions. The main body of the image is the same as the first image, but the brightness distribution of the specular reflection area is different due to the different polarization angles or positions.

[0067] In this embodiment of the invention, controlling the image acquisition unit to acquire a second image of the medical device for the local target region corresponding to the first uncertain image recognition result includes:

[0068] Identify image acquisition units at different locations within the local target region;

[0069] The shooting direction of the image acquisition unit at different positions is determined based on the polarization angle of the preset rotation angle;

[0070] By controlling the image acquisition units at different positions according to the shooting direction, images can be captured to obtain second images of the medical device from different shooting directions.

[0071] In detail, multiple image acquisition units can be deployed in a local target area, and two acquisition units can be controlled to take pictures using polarizers at preset angles to each other. For example, when the polarization angles of the two acquisition directions differ by 90°: θ1=0° (horizontal polarization) mainly captures the diffuse reflection component, and θ2=90° (vertical polarization) mainly captures the specular reflection component. Utilizing the physical properties of polarized light, images Iθ1 and Iθ2 are acquired through 0° and 90° polarizers respectively, and the diffuse reflection component and specular reflection component can be separated by pixel-level differential imaging.

[0072] Specifically, such as Figure 3 As shown, two cameras with a resolution ≥1920×1080 can be selected. 0° / 90° CPL polarizers are fixed in front of the lenses respectively, and they are installed side-by-side 45 cm above the local target area. The cameras are connected to the host via a USB-3.0 hub, and automatic exposure and automatic white balance are turned off. When the first uncertain result is generated, the system starts the second image acquisition process, sets the polarization angles to θ1 and θ2 (such as 0° and 90°) for the corresponding acquisition units respectively, and sends a synchronization trigger signal to ensure consistent acquisition timing, outputting Iθ1 and Iθ2.

[0073] In addition, such as Figure 4a The image shown is an image acquired under conditions where the polarization angle is 0°; as shown Figure 4b The image shown is an image acquired under a polarization angle of 90°; as shown Figure 4c The image shown is a schematic diagram of the result after fusing two images to remove glare.

[0074] Furthermore, in order to ensure that the image optimization processing algorithm is effective for partial light reflection, enhance the adaptability to local reflection optimization of different materials, and control the consumption of algorithm resources while enhancing the processing of image detail features, an effective balance between recognition accuracy and computational resource consumption can be achieved.

[0075] In this embodiment of the invention, the glare-free image refers to a high-quality image that meets specific application requirements after performing glare-reduction optimization processing on the second image.

[0076] In this embodiment of the invention, performing glare reduction optimization processing on the second image to obtain a glare-free image includes:

[0077] Identify the first-view image and the second-view image in the second image that correspond to the polarization angle;

[0078] Calculate and normalize the pixel-level differences between the two viewpoint images to obtain the mirror probability map;

[0079] An adaptive glare suppression weight is calculated based on the refractive index and specular reflectivity of the medical device material, and the glare reduction image is generated by weighting the specular probability map.

[0080] The color attributes of the deglare image are adjusted in multiple preset color spaces to obtain the target deglare image corresponding to each color space;

[0081] Calculate the overall reflectivity of the deglare image and generate fusion weights for each color space based on the adaptive light suppression weights;

[0082] The deglare images of each target are weighted and fused according to the fusion weights to obtain the optimized image;

[0083] Local refinement is performed on the high specular probability regions of the optimized image to obtain a glare-free image.

[0084] In detail, after the first uncertain image recognition result appears, the system acquires two frames of images to be fused, Iθ1 and Iθ2 (e.g., 0° and 90°) at two preset polarization angles within the same local target area. For example, the pixel-level difference between the two frames of images to be fused, Iθ1 and Iθ2, can be calculated according to equation (1): Then, according to equation (2), perform the difference map corresponding to the pixel-level difference between Iθ1 and Iθ2. Normalize: (2), of which It is the minimum value. The specular probability maps corresponding to Iθ1 and Iθ2 are given. max D represents the maximum value of the difference map D across the entire image (i.e., max{i,j} D(i,j)). The value of w ranges from [0,1]. After normalization, a specular probability map (i.e., a pixel-level glare intensity mask) in the range of 0 to 1 is obtained, which is used for subsequent weight allocation and local refinement. Furthermore, to adapt to the reflective characteristics of instruments made of different materials, the refractive index recorded in the instrument material library can be used. With specular reflectivity R s Calculate the adaptive weight α.

[0085] For example, it can be calculated according to formula (3): (3), among which, , Weighting coefficients obtained through experience or training, refractive index With specular reflectivity R sAll data originates from a pre-defined material parameter library: this library is built upon publicly available optical manuals, instrument manufacturers' technical specifications, and a small number of sampling calibration results, and can be continuously maintained according to instrument category (stainless steel, titanium alloy, ceramic, etc.); if a corresponding material is temporarily missing from the library, the system automatically uses the default average value for that category of instruments, ultimately... The adaptive weight α is obtained by normalizing the mapping to the interval (0,1).

[0086] Furthermore, the mirror probability diagram (Obtained by normalizing the difference graph D using equation (2)) and adaptive weights By combining and weighting the glare reduction fusion, a glare reduction image F1 is generated, thereby achieving a uniform glare reduction effect across materials such as stainless steel, titanium alloy, and ceramic. Figure 5a As shown, the original image acquired by dual polarization exhibits obvious specular highlights; as Figure 5b As shown, the deglare image F1 obtained after the above differential normalization and material adaptive weighting process basically eliminates glare.

[0087] Specifically, brightness compression and / or local equalization can be performed on the deglare image F1 in at least two preset color spaces (such as RGB, HSV, LAB, YUV) to obtain the processing results for each color space. Subsequently, the proportion of bright pixels H_ratio,c and the proportion of overexposed pixels O_ratio,c of F1 after processing in each color space are statistically analyzed, and the reflectivity of the color space is calculated: R_(intensity,c) = H_ratio,c + 2·O_ratio,c (4). The coefficient 2 is an empirical weight used to emphasize the influence of overexposure on glare. R_(intensity,c) is then compared with preset thresholds T1 and T2 to determine the reflectivity level (slight / medium / heavy) of the color space, and the fusion weight w_c of the color space is dynamically generated accordingly.

[0088] For example, let the processing result of color space c (such as RGB) be F_c, and the processing result of color space d (such as HSV) be F_d: in areas of strong reflection (pixel-level glare weight close to 1), F_d is preferred (HSV's V channel suppresses light more thoroughly), while in areas of weak reflection (pixel-level glare weight close to 0), F_c is preferred (RGB has high color fidelity). Based on this, the processing results of multiple color spaces are weighted and fused to obtain the optimized image U; where the reflectivity R_(intensity,c) of the c-th color space and the adaptive light suppression weight α_c of the corresponding material are used together for weight calculation to obtain the fusion weight w_c of the c-th color space, and its calculation formula is:

[0089]

[0090] in Indicates the first A color space, To sum the index, The number of color spaces used; , R_k is a weighting coefficient used to jointly adjust the influence of reflectivity and material light-suppressing adaptability; R_(intensity,k) is the weighting coefficient of the first element. The reflectivity of each color space For the first Each color space corresponds to an adaptive light suppression weight for the material. The above-mentioned multi-color space brightness compression and dynamic weight fusion mechanism can automatically adjust the output under different surgical lamp brightness and color temperature conditions, maintain the brightness uniformity of the glare-free image, and reduce the need for recalibration.

[0091] Furthermore, while multi-color space fusion has significantly reduced glare, residual bright spots may still exist in highly reflective areas on the surface of metal instruments (such as the tip of surgical forceps), affecting recognition accuracy. Therefore, based on the mirror probability map, local optimization processing can be performed only on pixels with a mirror probability exceeding a preset threshold.

[0092] In this embodiment of the invention, in order to accurately identify the image in the highly reflective area of ​​the metal instrument surface, the step of performing local refinement on the high specular probability area of ​​the optimized image to obtain a glare-free image includes:

[0093] Based on the mirror probability map, a local filtering weight map is generated by combining the adaptive light suppression weight;

[0094] Mirror pixels with a probability greater than a preset probability threshold are extracted from the mirror probability map and used as a high-probability pixel set.

[0095] Based on the local filter weight map, edge-preserving smoothing is performed on the pixels in the high-probability pixel set to output a glare-free image.

[0096] In detail, the mirror probability diagram This is a grayscale image with the same dimensions as the original image, where each pixel value represents the probability that the pixel belongs to a specular reflection region. The probability can be calculated using pixel-level differences. Calculations show that a larger D value corresponds to a higher probability of mirroring. Subsequently, guided filtering is used to perform edge-preserving smoothing on the mirrored regions to optimize the image U or its gradient map. As a guide map G, it smooths areas with high mirror probability while preserving instrument texture and edges to avoid overall blurring.

[0097] Specifically, the mirror probability diagram is as follows: Let the probability threshold be T. When w(i,j)>T, the pixel is recorded as a high-probability pixel (where i and j represent the pixel's probability in the mirror image). The coordinate indexes in the inner row and column directions (unrelated to color channels) can be adjusted downwards for the edge areas of the instrument. Based on Generate local filter weight map ,in , A larger value indicates stronger specular reflection, which in turn corresponds to a smaller weight and a stronger filtering intensity. For In the region where filtering is applied, only weak filtering or no filtering is performed. The weight map W is used to adjust the filtering strength and is correlated with the guiding map G (which can be U or ...). The combined effect of these technologies performs pixel-level corrections on the optimized image, enabling targeted optimization of regions with high specular probability.

[0098] Furthermore, in the exemplary implementation, polarization difference and multi-color space processing are implemented using a Python+OpenCV CPU prototype at a resolution of 1280 × 720, with an acceptable average processing time. Replacing the interpretation layer implementation with C++ / SIMD optimization, retaining only the two primary color spaces, and disabling debug logging can further reduce latency and meet the requirements of real-time applications. After the aforementioned glare removal and color optimization, key image features are preserved or enhanced; therefore, the glare-free image needs to be input again into the first analysis model to obtain more accurate recognition results.

[0099] S4. Input the glare-free image into the first analysis model to obtain the second image recognition result of the medical device.

[0100] In this embodiment of the invention, the second determined image recognition result is the recognition result obtained after analyzing the glare-free image. By optimizing the image quality, recognition ambiguity is reduced, and the result is more reliable.

[0101] In detail, this step is consistent with the judgment process described in S2: the results are still divided based on the preset confidence threshold and probability difference threshold, so it will not be described again.

[0102] Specifically, after taking a picture, the system inputs the glare-free image into the lightweight YOLO-v8 network and instantly outputs the category, quantity, and preset reliability of each device. For detection results with the highest probability value lower than the preset reliability threshold, the system will generate a review prompt on the user interface for nurses to manually confirm as needed. The review record will be archived and can be used as reference data for subsequent model iterations and performance tuning.

[0103] For example, 1500 self-acquired device images were selected (training:validation = 80:20), and training was performed from scratch for 100 epochs on the YOLO-v8-s lightweight backbone. The validation set results are as follows. Figure 6 As shown, mAP50 = 0.993, mAP50-95 = 0.933, Precision = 0.998, and Recall = 0.995. The normalized confusion matrix is ​​as follows. Figure 7 As shown, the accuracy of Hemolok-L (clamping device - large) and Hemolok-XL (clamping device - extra large) is 100%, the accuracy of Hemolok-ML (clamping device - medium) is 97%, and the overall false negative rate is 1.9%. The above data is only one example to illustrate the feasibility of this method and does not limit the scope of protection of this invention.

[0104] Furthermore, the system firstly identifies that image recognition results may be misjudged due to factors such as ambient light and occlusion, and secondly, that image recognition results may deviate due to improper glare removal processing parameter settings. Therefore, the system can fuse and compare the two results: if they match, the result is directly confirmed; if the categories or quantities are inconsistent, the result with higher confidence is prioritized, or manual review is triggered to compensate for the error. For example, if the first result misclassifies the device category, the high-confidence output of the second result can be used to correct the error.

[0105] S5. Based on the first determined image recognition result and the second determined image recognition result, determine the target medical device corresponding to the recognition instruction.

[0106] In this embodiment of the invention, a more reliable recognition result is generated by the global coverage of the first determined image recognition result and the detailed precision complementarity of the second determined image recognition result.

[0107] In this embodiment of the invention, determining the medical device corresponding to the user's medical device identification command based on the first determined image recognition result and the second determined image recognition result includes:

[0108] The first determined image recognition result and the second determined image recognition result are fused at the result level to generate a fused recognition result;

[0109] Parse the target device requirements from the user's identification instructions;

[0110] The fusion recognition results are matched with the target requirements. When the matching degree of category and quantity reaches a preset threshold, the target medical device corresponding to the recognition instruction is determined based on the fusion recognition results; otherwise, a difference prompt or manual review process is triggered.

[0111] In detail, the first determined image recognition result and the second determined image recognition result are weighted and superimposed at the result level to generate a fused recognition result. This result includes instrument category, quantity and location information to integrate the global information of the original image and the detail accuracy of the deglare image, thereby reducing the probability of misjudgment caused by a single model. The recognition command input by the user may contain text description (e.g., "scalpel") or reference image, which needs to be converted into a computable feature representation (feature vector). The target instrument requirement is the target image / region specified in the instrument list or command.

[0112] Specifically, the similarity (such as cosine similarity) between the fusion result and the user-instructed feature vector can be calculated. When the similarity is greater than or equal to a preset matching threshold, the two are determined to be consistent. If the threshold is not reached, processes such as reshooting or manual review can be triggered. The above-mentioned dual-result fusion mechanism can further reduce the risk caused by a single error.

[0113] Furthermore, to achieve automatic omission correction, redundancy correction, and placement position correction functions, the method also includes:

[0114] The image regions corresponding to the first image acquired by the control image acquisition unit are marked sequentially to generate the actual identification result and actual marking sequence of the medical device.

[0115] Identify the actual mapping relationship data between the actual identification result and the actual label sequence;

[0116] Target mapping relationship data is generated based on the target results and target label sequences in the pre-determined medical device inventory list;

[0117] Analyze whether there are any differences between the actual mapping relationship data and the target mapping relationship data;

[0118] If a discrepancy exists, the pre-determined reminder unit will output the reminder information corresponding to the discrepancy.

[0119] In detail, firstly, a target mapping sequence is generated based on the selected instrument inventory list. Simultaneously, all identifiable medical instruments (such as surgical forceps, syringes, gauze, etc.) in the first image are marked. Specifically, the bounding box of each instrument is located using an object detection algorithm, and the marking order can be determined according to either of the following strategies: First, spatial coordinate sorting, with the upper left corner of the image as the origin, numbered in a "left to right, top to bottom" order, suitable for fixed instrument table scenarios; Second, functional usage order, numbered according to the order of use in the surgical procedure, suitable for dynamic operating room scenarios. For example, for an image containing three instruments, the actual marking sequence can be obtained: M001-surgical forceps [50,80,150,180], M002-hemostat [200,80,300,180], M003-tweezers [350,80,450,180] (see...). Figure 8 Based on the recognition output of the first analysis model, combined with the aforementioned tag sequence, a one-to-one mapping relationship can be formed between the actual recognition result and the actual tag sequence. This actual mapping relationship is used to record the "tag ID". The information corresponding to "instrument category (and its location)" is a quantitative description of the current instrument status. Correspondingly, the target mapping relationship is pre-set by clinical guidelines or surgical procedures, including the sequence mapping of the target instrument category / quantity and its expected placement location.

[0120] Specifically, the system compares the actual mapping relationship with the target mapping relationship by comparing the quantity and sequence relationship of the two: ① Actual quantity less than target → omission; ② Actual quantity more than target → redundancy; ③ Same category but position deviation → position error. This determines whether a discrepancy exists, including but not limited to: 1. Omission: The actual sequence lacks the instrument category or quantity specified in the target sequence; 2. Redundancy: The actual sequence contains instrument categories or quantities not specified in the target sequence; 3. Position error: The category is the same, but the placement position deviates too much from the target position. For position determination, indicators such as bounding box overlap and center point distance can be used and compared with preset thresholds to determine if there is a deviation. When a discrepancy is detected, the system control reminder unit outputs a reminder message for that discrepancy. For example, the system can list four columns on the difference comparison result interface: "Instrument Name / Expected Quantity / Actual Detection / Status". The status column is visually marked with √ / × icons to indicate consistency or discrepancy. After the nurse confirms, they click "Return to Take Another Photo" or close the prompt to complete one counting cycle.

[0121] Furthermore, such as Figure 9aAs shown, after taking a photo, the system automatically pops up the recognition results interface. The left side displays a preview and operation controls such as "Recognize Instruments," while the right side displays "No Difference" information in a green prompt bar and lists the difference details table (instrument name, expected quantity, actual detection, status). Nurses can choose "Confirm" to end this round of counting or "Retake Photo" to trigger the next recognition. Confirmation information can be written to the log for subsequent model iterations. Figure 9b As shown, when a difference is detected, a red (or other highlighted color) difference indicator bar can be displayed on the right side of the interface, and the specific abnormal item will be marked in the difference table. Nurses can also choose to confirm or retake the photo, and the confirmation record can be archived for model optimization.

[0122] Furthermore, the system can output different levels of alerts based on the type of discrepancy: 1. Omission: A high-level warning is displayed, and the target location can be marked with a dashed box on the interface; 2. Redundancy: The bounding box of the redundant instrument is highlighted in a general alert format; 3. Position Error: A comparison overlay image of the actual position and the target position is displayed, along with the offset. The alert content can be presented through text / image annotations on the display screen or flashing LED indicators on the instrument station. In terms of medical risk level, omission correction has the highest priority, followed by redundancy correction, and then position correction.

[0123] Furthermore, to prevent errors in the second determined image recognition result and to prevent misjudgment caused by extreme cases, the method further includes:

[0124] Based on the second determined image recognition result, a report on the association labeling between the image and the recognition result is generated;

[0125] Send the generated association tag report to the preset terminal and obtain the review result;

[0126] Based on the review results, the second image recognition result is determined or corrected, and the final result is generated.

[0127] In detail, the association tagging report serves as a bridge connecting machine recognition results with human review. It presents recognition details visually, facilitating quick judgment by reviewers. The association tagging report includes image acquisition time, device ID, the recognition category of the second-determined image recognition result, confidence level, and location bounding box. For example, the association tagging report could be: {"Report ID": "REP-20250627-001", "Recognition Result": {"Category": "Hemostat", "Confidence Level": 0.96, "Bounding Box": [210, 180, 320, 290]} Risk label: "Low risk (no obvious occlusion)"}, Related data: {"Standard template ID": "TPL-Hemostat-003", Feature matching degree: {"Overall": 0.91, "Handle": 0.89, "Handle": 0.94}}, Visualization image path: " / data / reports / vis_20250627-001.png"}.

[0128] Specifically, the report is sent to the review terminal. The interface displays a visual image of the report on the left and a comparison with a standard template on the right. 1. Review Passed: The system records the reviewer's ID and time, and marks the second confirmed image recognition result as final and valid. 2. Review Correction: The reviewer can modify fields such as category or bounding box. The system updates the result accordingly and triggers model feedback. 3. Review Questionable: Marked as pending review and pushed to senior reviewers for processing. The final result is written to the database and synchronized to downstream systems such as inventory records / inventory management.

[0129] As can be seen, the present invention significantly improves the accuracy and reliability of medical device identification by timely acquisition of the first image, dual polarization differential glare removal, and secondary recognition fusion, thus solving the problem of insufficient accuracy in the prior art.

[0130] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0131] In one embodiment, a medical device intelligent identification device is provided, which corresponds one-to-one with the medical device intelligent identification method described in the above embodiments. For example... Figure 10 As shown, the intelligent medical device identification device includes a first image acquisition module 101, an image recognition module 102, an image optimization and processing module 103, a second image recognition result analysis module 104, and a medical device identification module 105. Detailed descriptions of each functional module are as follows:

[0132] The first image acquisition module 101 is used to control the image acquisition unit to acquire the first image of the medical device after receiving a medical device identification command from the user.

[0133] Image recognition module 102 is used to input the acquired first image into a predetermined first analysis model and judge it according to a preset confidence threshold to obtain a first certain image recognition result and a first uncertain image recognition result;

[0134] The image optimization processing module 103 is used to control the image acquisition unit to acquire a second image of the medical device for the local target area corresponding to the first uncertain image recognition result, and to perform anti-glare optimization processing on the second image to obtain a glare-free image.

[0135] The second image recognition result analysis module 104 is used to input the glare-free image into the first analysis model to obtain the second image recognition result of the medical device.

[0136] The medical device identification module 105 is used to identify the target medical device corresponding to the identification instruction based on the first identification image recognition result and the second identification image recognition result.

[0137] In one embodiment, the image recognition module 102, when executing the process of inputting the acquired first image into a predetermined first analysis model and determining it according to a preset confidence threshold to obtain a first certain image recognition result and a first uncertain image recognition result, is configured to:

[0138] The first image is input into the first analysis model, and the probability distribution vector corresponding to the first image is output.

[0139] When the highest probability value in the probability distribution vector is greater than or equal to the preset confidence threshold, the medical device category corresponding to the highest probability value is determined as the target device category corresponding to the first image, and a first determined image recognition result is generated based on the target device category.

[0140] Calculate the difference between the highest probability value and the second highest probability value in the probability distribution vector. When the highest probability value is less than the preset confidence threshold and the difference is less than the preset difference threshold, determine the medical device category corresponding to the highest probability value as the candidate device category corresponding to the first image, and generate a first uncertain image recognition result based on the candidate device category.

[0141] In one embodiment, the image optimization processing module 103, when controlling the image acquisition unit to acquire a second image of the medical device for the local target region corresponding to the first uncertain image recognition result, is used to:

[0142] Identify image acquisition units at different locations within the local target region;

[0143] The shooting direction of the image acquisition unit at different positions is determined based on the polarization angle of the preset rotation angle;

[0144] By controlling the image acquisition units at different positions according to the shooting direction, images can be captured to obtain second images of the medical device from different shooting directions.

[0145] In one embodiment, the image optimization processing module 103, when performing glare optimization processing on the second image to obtain a glare-free image, is used to:

[0146] Identify the first-view image and the second-view image in the second image that correspond to the polarization angle;

[0147] Calculate and normalize the pixel-level differences between the two viewpoint images to obtain the mirror probability map;

[0148] An adaptive glare suppression weight is calculated based on the refractive index and specular reflectivity of the medical device material, and the glare reduction image is generated by weighting the specular probability map.

[0149] The color attributes of the deglare image are adjusted in multiple preset color spaces to obtain the target deglare image corresponding to each color space;

[0150] Calculate the overall reflectivity of the deglare image and generate fusion weights for each color space based on the adaptive light suppression weights;

[0151] The deglare images of each target are weighted and fused according to the fusion weights to obtain the optimized image;

[0152] Local refinement is performed on the high specular probability regions of the optimized image to output a glare-free image.

[0153] In one embodiment, the image optimization processing module 103, when performing local refinement on high specular probability regions of the optimized image and outputting a glare-free image, is further configured to:

[0154] Based on the mirror probability map, a local filtering weight map is generated by combining the adaptive light suppression weight;

[0155] Mirror pixels with a probability greater than a preset probability threshold are extracted from the mirror probability map and used as a high-probability pixel set.

[0156] Based on the local filter weight map, edge-preserving smoothing is performed on the pixels in the high-probability pixel set to output a glare-free image.

[0157] In one embodiment, during the intelligent identification of a medical device, the method is further used to:

[0158] The image regions corresponding to the first image acquired by the control image acquisition unit are marked sequentially to generate the actual identification result and actual marking sequence of the medical device.

[0159] Identify the actual mapping relationship data between the actual identification result and the actual label sequence;

[0160] Target mapping relationship data is generated based on the target results and target label sequences in the pre-determined medical device inventory list;

[0161] Analyze whether there are any differences between the actual mapping relationship data and the target mapping relationship data;

[0162] If a discrepancy exists, the pre-determined reminder unit will output the reminder information corresponding to the discrepancy.

[0163] In one embodiment, during the intelligent identification of a medical device, the method is further used to:

[0164] Based on the second determined image recognition result, a report on the association labeling between the image and the recognition result is generated;

[0165] Send the generated association tag report to the preset terminal and obtain the review result;

[0166] Based on the review results, the second image recognition result is determined or corrected, and the final result is generated.

[0167] This invention provides an intelligent medical device identification device. Upon receiving an identification command, it promptly acquires a first image to obtain comprehensive initial visual information. Combining dual-polarization acquisition and pixel difference technology, it accurately separates specular highlight and diffuse reflection areas, improving image clarity. A first analysis model quickly determines confirmed and uncertain identification areas, focusing on acquiring a second image of the uncertain areas to reduce redundant acquisition and improve data efficiency and accuracy. The image, after glare removal and optimization, possesses clearer feature information, and further analysis significantly improves the identification accuracy of uncertain areas. Finally, by fusing the first and second confirmed identification results, the accuracy and reliability of medical device identification are comprehensively improved, effectively solving the technical problem of insufficient accuracy in existing intelligent medical device identification systems.

[0168] Specific limitations regarding the intelligent identification device for medical devices can be found in the limitations of the intelligent identification method for medical devices described above, and will not be repeated here. Each module in the aforementioned intelligent identification device for medical devices can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the medical device in hardware form or independently of it, or they can be stored in the memory of the medical device in software form, so that the processor can call and execute the operations corresponding to each module.

[0169] In one embodiment, a medical device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 11 As shown, the medical device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a medical device intelligent identification method on the server side.

[0170] In one embodiment, a medical device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 12 As shown, the medical device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a medical device intelligent identification method on the client side.

[0171] In one embodiment, a medical device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0172] Upon receiving a medical device identification command from the user, the system controls the image acquisition unit to acquire the first image of the medical device.

[0173] The first image is input into a pre-determined first analysis model and judged according to a pre-set confidence threshold to obtain a first determined image recognition result and a first uncertain image recognition result.

[0174] For the local target area corresponding to the first uncertain image recognition result, the image acquisition unit is controlled to acquire the second image of the medical device, and the second image is subjected to anti-glare optimization processing to obtain an anti-glare image;

[0175] The glare-free image is input into the first analysis model to obtain the second definitive image recognition result of the medical device.

[0176] Based on the first determined image recognition result and the second determined image recognition result, the target medical device corresponding to the recognition instruction is determined.

[0177] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0178] Upon receiving a medical device identification command from the user, the system controls the image acquisition unit to acquire the first image of the medical device.

[0179] The first image is input into a pre-determined first analysis model and judged according to a pre-set confidence threshold to obtain a first determined image recognition result and a first uncertain image recognition result.

[0180] For the local target area corresponding to the first uncertain image recognition result, the image acquisition unit is controlled to acquire the second image of the medical device, and the second image is subjected to anti-glare optimization processing to obtain an anti-glare image;

[0181] The glare-free image is input into the first analysis model to obtain the second definitive image recognition result of the medical device.

[0182] Based on the first determined image recognition result and the second determined image recognition result, the target medical device corresponding to the recognition instruction is determined.

[0183] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or medical device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0184] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0185] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0186] It should be noted that if any software tools or components not belonging to our company appear in the embodiments of this application, they are merely for illustrative purposes and do not represent actual use.

[0187] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for intelligent identification of medical devices, characterized in that, include: Upon receiving a medical device identification command from the user, the system controls the image acquisition unit to acquire the first image of the medical device. The first image is input into a pre-determined first analysis model and judged according to a pre-set confidence threshold to obtain a first determined image recognition result and a first uncertain image recognition result. For the local target area corresponding to the first uncertain image recognition result, the image acquisition unit is controlled to acquire the second image of the medical device, and the second image is subjected to anti-glare optimization processing to obtain an anti-glare image; The glare-free image is input into the first analysis model to obtain the second definitive image recognition result of the medical device. Based on the first determined image recognition result and the second determined image recognition result, the target medical device corresponding to the recognition instruction is determined.

2. The intelligent identification method for medical devices as described in claim 1, characterized in that, The step of inputting the acquired first image into a pre-determined first analysis model and judging it according to a pre-set confidence threshold to obtain a first determined image recognition result and a first uncertain image recognition result includes: The first image is input into the first analysis model, and the probability distribution vector corresponding to the first image is output. When the highest probability value in the probability distribution vector is greater than or equal to the preset confidence threshold, the medical device category corresponding to the highest probability value is determined as the target device category corresponding to the first image, and a first determined image recognition result is generated based on the target device category. Calculate the difference between the highest probability value and the second highest probability value in the probability distribution vector. When the highest probability value is less than the preset confidence threshold and the difference is less than the preset difference threshold, determine the medical device category corresponding to the highest probability value as the candidate device category corresponding to the first image, and generate a first uncertain image recognition result based on the candidate device category.

3. The intelligent identification method for medical devices as described in claim 1, characterized in that, The step of controlling the image acquisition unit to acquire a second image of the medical device for the local target region corresponding to the first uncertain image recognition result includes: Identify image acquisition units at different locations within the local target region; The shooting direction of the image acquisition unit at different positions is determined based on the polarization angle of the preset rotation angle; By controlling the image acquisition units at different positions according to the shooting direction, images can be captured to obtain second images of the medical device from different shooting directions.

4. The intelligent identification method for medical devices as described in claim 3, characterized in that, The step of performing glare reduction optimization on the second image to obtain a glare-free image includes: Identify the first-view image and the second-view image in the second image that correspond to the polarization angle; Calculate and normalize the pixel-level differences between the two viewpoint images to obtain the mirror probability map; An adaptive glare suppression weight is calculated based on the refractive index and specular reflectivity of the medical device material, and the glare reduction image is generated by weighting the specular probability map. The color attributes of the deglare image are adjusted in multiple preset color spaces to obtain the target deglare image corresponding to each color space; Calculate the overall reflectivity of the deglare image and generate fusion weights for each color space based on the adaptive light suppression weights; The deglare images of each target are weighted and fused according to the fusion weights to obtain the optimized image; Local refinement is performed on the high specular probability regions of the optimized image to output a glare-free image.

5. The intelligent identification method for medical devices as described in claim 4, characterized in that, The step of performing local refinement on the high specular probability regions of the optimized image to output a glare-free image includes: Based on the mirror probability map, a local filtering weight map is generated by combining the adaptive light suppression weight; Mirror pixels with a probability greater than a preset probability threshold are extracted from the mirror probability map and used as a high-probability pixel set. Based on the local filter weight map, edge-preserving smoothing is performed on the pixels in the high-probability pixel set to output a glare-free image.

6. The intelligent identification method for medical devices as described in claim 1, characterized in that, The method further includes: The image regions corresponding to the first image acquired by the control image acquisition unit are marked sequentially to generate the actual identification result and actual marking sequence of the medical device. Identify the actual mapping relationship data between the actual identification result and the actual label sequence; Target mapping relationship data is generated based on the target results and target label sequences in the pre-determined medical device inventory list; Analyze whether there are any differences between the actual mapping relationship data and the target mapping relationship data; If a discrepancy exists, the pre-determined reminder unit will output the reminder information corresponding to the discrepancy.

7. The intelligent identification method for medical devices as described in claim 1, characterized in that, The method further includes: Based on the second determined image recognition result, a report on the association labeling between the image and the recognition result is generated; Send the generated association tag report to the preset terminal and obtain the review result; Based on the review results, the second image recognition result is determined or corrected, and the final result is generated.

8. A medical device intelligent identification device, characterized in that, include: The first image acquisition module is used to control the image acquisition unit to acquire the first image of the medical device after receiving the medical device identification command issued by the user. The image recognition module is used to input the acquired first image into a predetermined first analysis model and judge it according to a preset confidence threshold to obtain a first certain image recognition result and a first uncertain image recognition result; The image optimization processing module is used to control the image acquisition unit to acquire a second image of the medical device for the local target area corresponding to the first uncertain image recognition result, and to perform anti-glare optimization processing on the second image to obtain a glare-free image. The second image recognition result analysis module is used to input the glare-free image into the first analysis model to obtain the second image recognition result of the medical device. A medical device identification module is used to identify the target medical device corresponding to the identification command based on the first image recognition result and the second image recognition result.

9. A medical device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it is used to implement the intelligent identification method for medical devices as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it is used to implement the intelligent identification method for medical devices as described in any one of claims 1 to 7.

Images