Automatic sterilization equipment for medical orthopedic instruments and sterilization method thereof

Abstract

This invention discloses an automated disinfection device and method for medical orthopedic instruments. By acquiring a three-dimensional digital model of the instrument to be disinfected and identifying complex structural feature areas on its surface, the disinfection process can precisely locate key parts such as threads and blind holes, achieving a shift from extensive processing to structural adaptation. The disinfection process is modeled and optimized based on fluid dynamics simulation, generating customized initial disinfection parameter sets for different feature areas. Through synchronous acquisition and fusion analysis of multimodal sensor data, a real-time evaluation model of the cleaning status is constructed, enabling synchronous quantitative monitoring of the thoroughness of cleaning and the risk of instrument damage in each area. An adaptive and precise disinfection is achieved by dynamically adjusting the motion and spray parameters at the disinfection execution end through a model predictive control algorithm based on quantitative scoring. By introducing fluorescent labeling and image analysis technology, objective and quantitative evaluation of the disinfection effect and generation of a comprehensive disinfection report are realized.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and in particular to an automatic disinfection device and disinfection method for medical orthopedic devices. Background Technology

[0002] Sterilization of medical devices is crucial for ensuring surgical safety and controlling hospital infections. Orthopedic surgical instruments, such as bone drills, bone saws, bone plates, screws, and reduction forceps, generally have complex structures, often including fine threads, deep grooves, complex joint surfaces, and blind holes of various sizes. These structural characteristics make it easy for contaminants (such as blood, tissue, and bone fragments) to remain and form biofilms, posing a significant challenge to thorough sterilization. Traditional sterilization methods, such as manual scrubbing combined with ultrasonic cleaning, or fully automated cleaning and sterilization machines, have significant limitations. Manual cleaning is inefficient, inconsistent, and struggles to ensure thorough cleaning of complex, hard-to-reach areas; while existing automated cleaning equipment often uses fixed spray programs and parameters, lacking the ability to perceive and adapt to individual instrument differences and contamination levels. This can lead to cleaning blind spots or the use of excessive mechanical force (such as excessive water pressure) to achieve cleaning effects, causing microscopic damage to instruments, especially those with sharp edges or fine threads, affecting their lifespan and surgical safety.

[0003] Therefore, there is an urgent need to develop an automated disinfection device and method that can intelligently identify the characteristics of orthopedic instruments, dynamically sense the cleaning status, and adaptively adjust the disinfection strategy accordingly, so as to achieve efficient, thorough and non-destructive disinfection and meet the high standards of instrument safety requirements of modern orthopedic surgery. Summary of the Invention

[0004] In order to overcome the above-mentioned shortcomings of the prior art, the present invention provides an automatic disinfection device and method for medical orthopedic instruments, aiming to provide an automatic disinfection device and method that can intelligently identify the characteristics of orthopedic instruments, dynamically sense the cleaning status, and adaptively adjust the disinfection strategy accordingly, so as to achieve efficient, thorough and non-destructive automatic disinfection.

[0005] The technical solution adopted by this invention to solve its technical problem is: an automatic disinfection method for medical orthopedic instruments, comprising the following steps: S1: Obtain a three-dimensional digital model of the instrument to be disinfected and identify the characteristic areas of the complex structure on the surface of the instrument; S2: The disinfection process is modeled and optimized based on fluid dynamics simulation to generate an initial set of disinfection parameters for different characteristic regions; S3: Drive the disinfection execution end to perform initial disinfection on the instrument surface according to the initial disinfection parameter set, and simultaneously collect multimodal sensor data; S4: Build and run a real-time assessment model for cleaning status, perform fusion analysis on the collected multimodal sensor data, and output a quantitative score in real time that reflects the adequacy of cleaning in each area and the risk of instrument damage. S5: Based on the quantitative score, a model predictive control algorithm is used to dynamically adjust the motion parameters and spray parameters of the disinfection execution end to perform secondary disinfection on the instrument surface; S6: Calculate the disinfection coverage rate using fluorescent labeling and image analysis technology, and generate a comprehensive disinfection report.

[0006] As a further improvement of the present invention: step S2 includes: S21: Based on the aforementioned three-dimensional digital model, construct a fluid dynamics simulation model of the flow of the disinfection medium on the instrument surface; S22: Define the optimization target for instrument surface cleaning efficiency, wherein the optimization target is positively correlated with wall shear stress and cleaning agent concentration, and negatively correlated with process energy consumption and maximum shear stress; S23: Based on the optimization objective, the optimal initial spray pressure, incident angle and flow rate for different characteristic areas are determined through iterative simulation optimization, forming the initial disinfection parameter set.

[0007] As a further improvement of the present invention: In step S21, the fluid dynamics simulation model is constructed based on a numerical method coupled with Lagrange particles and Eulerian grids to simulate the microscopic interaction between droplets of the disinfection medium and the surface of orthopedic instruments; wherein, the governing equations of the fluid dynamics simulation model include: The momentum conservation equation for the Euler phase is: ; The equation of motion for the Lagrange particle is: ; Where ρ is the fluid density, u is the fluid velocity vector, p is the fluid pressure, μ is the fluid dynamic viscosity, and S... p For the momentum source term of the Lagrange particle; m p v is the particle mass. p F is the particle velocity vector. drag F is the fluid resistance experienced by the particle. pressure F is the pressure gradient force acting on the particle. surface This refers to the interaction force between the particle and the surface of the instrument.

[0008] As a further improvement of the present invention: in step S3, the multimodal sensor data includes pressure fluctuation time series data, fluid image sequence and acoustic emission signal.

[0009] As a further improvement of the present invention: In step S4, the real-time evaluation model of the cleaning status is a dual-channel long short-term memory network model, the construction and operation of which include: S41: Extract time-frequency domain features from pressure fluctuation time series data, extract visual flow field features from fluid image sequences, and extract waveform energy features from acoustic emission signals; S42: The extracted time-frequency domain features, visual flow field features, and waveform energy features are weighted and fused to form a fused feature vector; S43: Input the fused feature vector into the dual-channel long short-term memory network model. The first output channel of the dual-channel long short-term memory network model generates a cleaning adequacy score S. clean ∈[0,1], the second output channel of the dual-channel long short-term memory network model generates a damage risk warning score S. risk ∈[0,1].

[0010] As a further improvement of the present invention: In step S5, the dynamic adjustment of the motion parameters and spray parameters of the disinfection execution end using a model predictive control algorithm specifically includes: S51: Establish a discrete system prediction model with the end state of disinfection execution and spray parameters as state variables, and the optimization objectives being to improve the cleaning adequacy score and suppress the damage risk score. S52: In each control cycle k, the cleaning adequacy score S output by the model in real time is evaluated based on the cleaning status. clean (k) and damage risk warning score S risk (k) represents the observed value, and the optimal control sequence ΔU is solved in a rolling manner over the next N steps. (k)=[Δu(k),Δu(k+1),...,Δu(k+N 1)], where Δu is the adjustment amount including spray pressure, incident angle and flow rate; S53: Execute the first step Δu(k) of the optimal control sequence and proceed to the next cycle of rolling optimization.

[0011] As a further improvement of the present invention: step S6 specifically includes: S61: Apply a fluorescent marker to the surface of the sterilized instrument and obtain a fluorescent image of the instrument surface; S62: Based on the fluorescence image of the instrument surface, the fluorescence image is processed using a U-Net convolutional neural network based on the attention mechanism to calculate the disinfection coverage and generate a comprehensive disinfection report.

[0012] As a further improvement of the present invention: step S62 specifically includes: S621: An encoder based on an attention mechanism U-Net convolutional neural network extracts multi-scale features of fluorescence images, and a decoder gradually restores the spatial resolution. S622: Introduce a spatial attention gate module into the U-Net convolutional neural network, based on the encoder feature map F. enc With decoder feature map F dec The attention weight map Att is dynamically generated. The formula for calculating the attention weight map Att is: ; Where σ is the Sigmoid activation function, ψ is the convolution operation, [*,*] represents feature concatenation, ReLU is the rectified linear unit activation function, and Conv... 1×1 For a 1×1 convolution operation, F enc For encoder feature map, F dec This is the decoder feature map; S623: The U-Net convolutional neural network outputs a disinfection coverage probability map for each pixel. After thresholding, the overall and regional disinfection coverage rates are calculated to generate a comprehensive disinfection report.

[0013] This invention also provides an automatic disinfection device for medical orthopedic instruments, used to implement the aforementioned automatic disinfection method for medical orthopedic instruments, comprising: The 3D scanning module is used to acquire a 3D digital model of the instrument to be disinfected and to identify complex structural feature areas on the surface of the instrument. A multi-degree-of-freedom robotic arm and a disinfection execution module located at its end; A multimodal sensor array is used to acquire multimodal sensor data. Control host, the control host includes: The simulation optimization unit is used to run the fluid dynamics simulation and generate an initial disinfection parameter set; The status assessment unit is used to run a real-time assessment model of the cleaning status. Adaptive control unit, used to run model predictive control algorithms; The image analysis unit is used to run the U-Net convolutional neural network based on the attention mechanism.

[0014] As a further improvement to the present invention, it also includes: An edge computing module is connected to a multimodal sensor array and is used to perform preprocessing, feature extraction and fusion of multimodal sensor data in real time, and send the obtained fused feature vector to the status evaluation unit of the control host.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This invention acquires a three-dimensional digital model of the instrument to be disinfected and identifies complex structural features on its surface, enabling precise positioning of key areas such as threads and blind holes during the disinfection process, thus shifting from extensive processing to structural adaptation. By modeling and optimizing the disinfection process based on fluid dynamics simulation, customized initial disinfection parameter sets are generated for different feature areas, overcoming the inconsistencies and limitations of traditional empirical parameters. Through synchronous acquisition and fusion analysis of multimodal sensor data, a real-time cleaning status assessment model is constructed, achieving simultaneous quantitative monitoring of the thoroughness of cleaning and the risk of instrument damage in each area. By dynamically adjusting the motion and spray parameters at the disinfection execution end using a model predictive control algorithm based on quantitative scoring, adaptive and precise disinfection is achieved, moving from fixed-program execution to a closed-loop perception-decision-execution system. Finally, by introducing independent verification steps using fluorescent labeling and image analysis technology, objective and quantitative evaluation of the disinfection effect and generation of a comprehensive disinfection report are realized. Attached Figure Description

[0016] Figure 1 This is a flowchart of an automated sterilization method for a medical orthopedic device according to the present invention.

[0017] Figure 2 This is a structural block diagram of an automatic disinfection device for medical orthopedic instruments according to the present invention. Detailed Implementation

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is 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. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0019] It should be understood that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or device.

[0020] It should be understood that specific details are provided in the following description to facilitate a complete understanding of the exemplary embodiments. However, those skilled in the art will understand that the exemplary embodiments can be implemented without these specific details. For example, the system may be shown in block diagrams to avoid obscuring the example with unnecessary details. In other instances, well-known processes, structures, and techniques may be shown without unnecessary details to avoid obscuring the exemplary embodiments. The invention will now be further described in conjunction with the accompanying drawings and embodiments: Implementation Case 1: Please see Figure 1 An automated sterilization method for medical orthopedic instruments includes the following steps: S1: Obtain a three-dimensional digital model of the instrument to be disinfected and identify the characteristic areas of the complex structure on the surface of the instrument; S2: The disinfection process is modeled and optimized based on fluid dynamics simulation to generate an initial set of disinfection parameters for different characteristic regions; S3: Drive the disinfection execution end to perform initial disinfection on the instrument surface according to the initial disinfection parameter set, and simultaneously collect multimodal sensor data; S4: Build and run a real-time assessment model for cleaning status, perform fusion analysis on the collected multimodal sensor data, and output a quantitative score in real time that reflects the adequacy of cleaning in each area and the risk of instrument damage. S5: Based on the quantitative score, a model predictive control algorithm is used to dynamically adjust the motion parameters and spray parameters of the disinfection execution end, and to perform targeted secondary disinfection on the instrument surface; S6: Calculate the disinfection coverage rate using fluorescent labeling and image analysis technology, and generate a comprehensive disinfection report.

[0021] By acquiring a 3D digital model of the instruments to be disinfected and identifying complex structural features on their surfaces, the disinfection process can precisely locate key areas such as threads and blind holes, achieving a shift from extensive processing to structural adaptation. Modeling and optimizing the disinfection process based on fluid dynamics simulation generates customized initial disinfection parameter sets for different feature areas, overcoming the inconsistencies and limitations of traditional empirical parameters. Through synchronous acquisition and fusion analysis of multimodal sensor data, a real-time cleaning status assessment model is constructed, enabling simultaneous quantitative monitoring of the thoroughness of cleaning and the risk of instrument damage in each area. A model predictive control algorithm based on quantitative scoring dynamically adjusts the motion and spray parameters at the disinfection execution end, achieving adaptive and precise disinfection from fixed program execution to a closed-loop perception-decision-execution system. Finally, by introducing independent verification steps using fluorescent labeling and image analysis technology, objective and quantitative evaluation of the disinfection effect and generation of a comprehensive disinfection report are achieved, forming a complete quality control closed loop.

[0022] In some implementations, step S2 includes: S21: Based on the aforementioned three-dimensional digital model, construct a fluid dynamics simulation model of the flow of the disinfection medium on the instrument surface; S22: Define the optimization target for instrument surface cleaning efficiency, wherein the optimization target is positively correlated with wall shear stress and cleaning agent concentration, and negatively correlated with process energy consumption and maximum shear stress; S23: Based on the optimization objective, the optimal initial spray pressure, incident angle and flow rate for different characteristic areas are determined through iterative simulation optimization, forming the initial disinfection parameter set.

[0023] By constructing a fluid dynamics simulation model based on a 3D digital model, high-fidelity prediction of the flow of disinfection media on instrument surfaces was achieved in a virtual environment, overcoming the drawbacks of high cost and long cycle time of traditional trial-and-error methods. By defining an optimization target for instrument surface cleaning efficiency, which is positively correlated with wall shear stress and cleaning agent concentration, and negatively correlated with process energy consumption and maximum shear stress, the parameter optimization process can scientifically balance the two key requirements of thorough cleaning and instrument protection, avoiding insufficient cleaning or instrument damage that may result from optimizing a single target. Iterative simulation optimization determined the optimal initial spray pressure, incident angle, and flow rate for different characteristic areas, forming the initial disinfection parameter set. This represents a leap from a one-size-fits-all fixed parameter to differentiated and refined parameter settings based on different characteristic areas, providing scientific and accurate benchmark operating parameters for subsequent automated disinfection operations.

[0024] In some embodiments, in step S21, the fluid dynamics simulation model is constructed based on a numerical method coupling Lagrange particles and Eulerian grids to simulate the microscopic interaction between droplets of the disinfection medium and the surface of orthopedic instruments; wherein, the governing equations of the fluid dynamics simulation model include: The momentum conservation equation for the Euler phase is: ; Where ρ is the fluid density, u is the fluid velocity vector, p is the fluid pressure, μ is the fluid dynamic viscosity, and S... p This is the momentum source term for Lagrange particles (used to achieve bidirectional momentum exchange between the particle phase and the fluid phase). The equation of motion for the Lagrange particle is: ; Where, m p v is the particle mass. p F is the particle velocity vector. drag F is the fluid resistance experienced by the particle. pressure F is the pressure gradient force acting on the particle. surfaceThis refers to the interaction force between particles and the instrument surface (used to simulate droplet impact dynamics, surface wetting, and dynamic contact angle changes).

[0025] A fluid dynamics simulation model was constructed using a numerical method coupling Lagrange particles and Eulerian meshes. This model achieved high-resolution simulation of the microscopic impact, wetting, and splashing processes between disinfectant droplets and the surface of orthopedic instruments, overcoming the limitations of traditional pure Eulerian methods in simulating complex phase interface motions. A momentum source term S from Lagrange particles was introduced into the momentum conservation equation of the Eulerian phase. p Mathematically, this achieves bidirectional momentum exchange between the discrete particle phase and the continuous fluid phase, enabling the fluid dynamics simulation model to more realistically reflect the disturbance and influence of the disinfection medium droplet swarm on the overall flow field. This is achieved by defining the microscopic interaction force F between the particle and the instrument surface in the Lagrange particle motion equation. surface This significantly improves the cleaning efficiency and the accuracy and reliability of mechanical load prediction for complex structures on instrument surfaces (such as threads and blind holes).

[0026] In some implementations, in step S3, the multimodal sensor data includes pressure fluctuation time-series data, fluid image sequences, and acoustic emission signals.

[0027] In some implementations, in step S4, the real-time evaluation model for the cleaning status is a dual-channel long short-term memory network model, the construction and operation of which include: S41: Extract time-frequency domain features from pressure fluctuation time series data, extract visual flow field features from fluid image sequences, and extract waveform energy features from acoustic emission signals; S42: The extracted time-frequency domain features, visual flow field features, and waveform energy features are weighted and fused to form a fused feature vector; S43: Input the fused feature vector into the dual-channel long short-term memory network model. The first output channel of the dual-channel long short-term memory network model generates a cleaning adequacy score S. clean ∈[0,1], the second output channel of the dual-channel long short-term memory network model generates a damage risk warning score S. risk ∈[0,1].

[0028] By simultaneously acquiring three types of multimodal sensor data—pressure fluctuation time-series data, fluid image sequences, and acoustic emission signals—a comprehensive process perception system is achieved, encompassing fluid system state, macroscopic flow coverage, and material microscopic response, overcoming the limitations of single-signal dimension assessment. By constructing a dual-channel long short-term memory (LSTM) network model and extracting and weighting multimodal sensor data to form a fused feature vector, information from time-frequency domain features, visual flow field features, and waveform energy features can be comprehensively utilized to accurately uncover deep temporal correlations and dynamic evolution patterns in the cleaning process. Through the unique dual-channel output structure of the dual-channel LSTM network model, for the first time, two key quantitative scores—cleaning adequacy and damage risk warning—are output in parallel and in real-time during the disinfection process. This allows the control system to simultaneously make clear and quantifiable decisions based on maximizing cleaning effectiveness and minimizing instrument damage, achieving intelligent, refined, and safe closed-loop management of the complex disinfection process.

[0029] In some implementations, step S5, which involves dynamically adjusting the motion parameters and spray parameters of the disinfection execution end using a model predictive control algorithm, specifically includes: S51: Establish a discrete system prediction model with the end state of disinfection execution and spray parameters as state variables, and the optimization objectives being to improve the cleaning adequacy score and suppress the damage risk score. S52: In each control cycle k, the cleaning adequacy score S output by the model in real time is evaluated based on the cleaning status. clean (k) and damage risk warning score S risk (k) represents the observed value, and the optimal control sequence ΔU is solved in a rolling manner over the next N steps. (k)=[Δu(k),Δu(k+1),...,Δu(k+N 1)], where Δu is the adjustment amount including spray pressure, incident angle and flow rate; S53: Execute the first step Δu(k) of the optimal control sequence and proceed to the next cycle of rolling optimization.

[0030] By establishing a discrete system prediction model with the optimization objectives of improving cleaning adequacy scores and suppressing damage risk scores, the dynamic control problem of the disinfection process is transformed into a rolling optimization mathematical problem, making the control decision forward-looking and systematic. In each control cycle, the cleaning adequacy score and damage risk warning score output by the real-time evaluation model of the cleaning status are used as observations, and the optimal control sequence for multiple future steps is solved rollingly. This achieves rapid adaptation and precise response of the control strategy to the cleaning status. By executing the first step of the rolling optimization sequence and continuously iterating, the optimal balance between improving cleaning effectiveness and instrument safety is continuously and automatically approached in a dynamically changing environment.

[0031] In some implementations, step S6 specifically includes: S61: Apply a fluorescent marker to the surface of the sterilized instrument and obtain a fluorescent image of the instrument surface; S62: Based on the fluorescence image of the instrument surface, the fluorescence image is processed using a U-Net convolutional neural network based on the attention mechanism to calculate the disinfection coverage and generate a comprehensive disinfection report.

[0032] In some implementations, step S62 specifically includes: S621: An encoder based on an attention mechanism U-Net convolutional neural network extracts multi-scale features of fluorescence images, and a decoder gradually restores the spatial resolution. S622: Introduce a spatial attention gate module into the U-Net convolutional neural network, based on the encoder feature map F. enc With decoder feature map F dec The attention weight map Att is dynamically generated. The formula for calculating the attention weight map Att is: ; Where σ is the Sigmoid activation function, ψ is the convolution operation, [*,*] represents feature concatenation, ReLU is the rectified linear unit activation function, and Conv... 1×1 For a 1×1 convolution operation, F enc For encoder feature map, F dec This is the decoder feature map; S623: The U-Net convolutional neural network outputs a disinfection coverage probability map for each pixel. After thresholding, it calculates the overall and zone (such as the threaded area and the hole area) disinfection coverage rate and generates a comprehensive disinfection report.

[0033] By applying fluorescent markers to the surfaces of sterilized instruments and acquiring fluorescence images of these surfaces, the qualitative assessment of sterilization effectiveness is transformed into an objective and quantitative analysis of fluorescence signals, enabling measurable and traceable evaluation results. A U-Net convolutional neural network based on an attention mechanism is employed to process the fluorescence images. Its encoder-decoder structure fully extracts multi-scale features, and a spatial attention gating module dynamically focuses on complex structural feature regions on the instrument surface, significantly improving the accuracy of fluorescence coverage area segmentation for irregular areas such as threads and holes. Finally, by outputting pixel-level sterilization coverage probability maps and calculating overall and zoned sterilization coverage rates, a detailed and objective quantitative sterilization report is generated, providing reliable data support for sterilization quality verification and closed-loop management.

[0034] Implementation Case 2: Please refer to Figure 2. An automatic sterilization device for medical orthopedic instruments, used to implement the aforementioned automatic sterilization method for medical orthopedic instruments, includes: The 3D scanning module is used to acquire a 3D digital model of the instrument to be disinfected and to identify complex structural feature areas on the surface of the instrument. A multi-degree-of-freedom robotic arm and a disinfection execution module located at its end; A multimodal sensor array is used to acquire multimodal sensor data. Control host, the control host includes: The simulation optimization unit is used to run the fluid dynamics simulation and generate an initial disinfection parameter set; The status assessment unit is used to run a real-time assessment model of the cleaning status. Adaptive control unit, used to run model predictive control algorithms; The image analysis unit is used to run the U-Net convolutional neural network based on the attention mechanism.

[0035] By integrating a 3D scanning module, a multi-degree-of-freedom robotic arm, a multimodal sensor array, and a control host including a simulation optimization unit, a state evaluation unit, an adaptive control unit, and an image analysis unit, a highly collaborative hardware execution platform was constructed. This platform realizes a complete automated closed loop from physical perception and intelligent decision-making to precise execution, ensuring the feasibility of the described automatic disinfection method for medical orthopedic instruments in the physical system. It achieves efficient, precise, and verifiable fully automated intelligent disinfection of orthopedic instruments.

[0036] In some implementations, it also includes: An edge computing module is connected to a multimodal sensor array and is used to perform preprocessing, feature extraction and fusion of multimodal sensor data in real time, and send the obtained fused feature vector to the status evaluation unit of the control host.

[0037] By deploying the preprocessing, feature extraction, and fusion functions of multimodal sensor data on the edge computing module connected to the multimodal sensor array, the raw data can undergo high-speed, low-latency preliminary intelligent processing near the source of acquisition. This significantly reduces the amount of data transmitted to the central control host and the communication latency, enabling the status assessment unit of the control host to receive the refined fused feature vector more quickly. This effectively reduces the real-time computing load of the host, ensures the operating efficiency of the real-time cleaning status assessment model, and greatly improves the real-time response speed and control closed-loop reliability of the entire system to the disinfection process. This provides a key guarantee for subsequent real-time assessment and adaptive closed-loop control.

[0038] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0039] The above 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 or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. An automated sterilization method for medical orthopedic instruments, characterized in that: Includes the following steps: S1: Obtain a three-dimensional digital model of the instrument to be disinfected and identify the characteristic areas of the complex structure on the surface of the instrument; S2: The disinfection process is modeled and optimized based on fluid dynamics simulation to generate an initial set of disinfection parameters for different characteristic regions; S3: Drive the disinfection execution end to perform initial disinfection on the instrument surface according to the initial disinfection parameter set, and simultaneously collect multimodal sensor data; S4: Build and run a real-time assessment model for cleaning status, perform fusion analysis on the collected multimodal sensor data, and output a quantitative score in real time that reflects the adequacy of cleaning in each area and the risk of instrument damage. S5: Based on the quantitative score, a model predictive control algorithm is used to dynamically adjust the motion parameters and spray parameters of the disinfection execution end to perform secondary disinfection on the instrument surface; S6: Calculate the disinfection coverage rate using fluorescent labeling and image analysis technology, and generate a comprehensive disinfection report.

2. The automatic disinfection method for medical orthopedic instruments according to claim 1, characterized in that: Step S2 includes: S21: Based on the aforementioned three-dimensional digital model, construct a fluid dynamics simulation model of the flow of the disinfection medium on the instrument surface; S22: Define the optimization target for instrument surface cleaning efficiency, wherein the optimization target is positively correlated with wall shear stress and cleaning agent concentration, and negatively correlated with process energy consumption and maximum shear stress; S23: Based on the optimization objective, the optimal initial spray pressure, incident angle and flow rate for different characteristic areas are determined through iterative simulation optimization, forming the initial disinfection parameter set.

3. The automatic disinfection method for medical orthopedic instruments according to claim 2, characterized in that: In step S21, the fluid dynamics simulation model is constructed based on a numerical method coupling Lagrange particles and Eulerian grids to simulate the microscopic interaction between droplets of the disinfection medium and the surface of orthopedic instruments; wherein, the governing equations of the fluid dynamics simulation model include: The momentum conservation equation for the Euler phase is: ； The equation of motion for the Lagrange particle is: ； Where ρ is the fluid density, u is the fluid velocity vector, p is the fluid pressure, μ is the fluid dynamic viscosity, and S... p For the momentum source term of the Lagrange particle; m p v is the particle mass. p F is the particle velocity vector. drag F is the fluid resistance experienced by the particle. pressure F is the pressure gradient force acting on the particle. surface This refers to the interaction force between the particle and the surface of the instrument.

4. The automatic disinfection method for medical orthopedic instruments according to claim 1, characterized in that: In step S3, the multimodal sensor data includes pressure fluctuation time series data, fluid image sequences, and acoustic emission signals.

5. The automatic disinfection method for medical orthopedic instruments according to claim 4, characterized in that: In step S4, the real-time evaluation model for the cleaning status is a dual-channel long short-term memory network model, the construction and operation of which include: S41: Extract time-frequency domain features from pressure fluctuation time series data, extract visual flow field features from fluid image sequences, and extract waveform energy features from acoustic emission signals; S42: The extracted time-frequency domain features, visual flow field features, and waveform energy features are weighted and fused to form a fused feature vector; S43: Input the fused feature vector into the dual-channel long short-term memory network model. The first output channel of the dual-channel long short-term memory network model generates a cleaning adequacy score S. clean ∈[0,1], the second output channel of the dual-channel long short-term memory network model generates a damage risk warning score S. risk ∈[0,1].

6. The automatic disinfection method for medical orthopedic instruments according to claim 5, characterized in that: In step S5, the dynamic adjustment of the motion parameters and spray parameters of the disinfection execution end using a model predictive control algorithm specifically includes: S51: Establish a discrete system prediction model with the end state of disinfection execution and spray parameters as state variables, and the optimization objectives being to improve the cleaning adequacy score and suppress the damage risk score. S52: In each control cycle k, the cleaning adequacy score S output by the model in real time is evaluated based on the cleaning status. clean (k) and damage risk warning score S risk (k) represents the observed value, and the optimal control sequence ΔU is solved in a rolling manner over the next N steps. (k)=[Δu(k),Δu(k+1),...,Δu(k+N 1)], where Δu is the adjustment amount including spray pressure, incident angle and flow rate; S53: Execute the first step Δu(k) of the optimal control sequence and proceed to the next cycle of rolling optimization.

7. The automatic disinfection method for medical orthopedic instruments according to claim 1, characterized in that: Step S6 specifically includes: S61: Apply a fluorescent marker to the surface of the sterilized instrument and obtain a fluorescent image of the instrument surface; S62: Based on the fluorescence image of the instrument surface, the U-Net convolutional neural network based on the attention mechanism is used to process the fluorescence image, calculate the disinfection coverage, and generate a comprehensive disinfection report.

8. The automatic disinfection method for medical orthopedic instruments according to claim 7, characterized in that: Step S62 specifically includes: S621: An encoder based on an attention mechanism U-Net convolutional neural network extracts multi-scale features of fluorescence images, and a decoder gradually restores the spatial resolution. S622: Introduce a spatial attention gate module into the U-Net convolutional neural network, based on the encoder feature map F. enc With decoder feature map F dec The attention weight map Att is dynamically generated. The formula for calculating the attention weight map Att is: ； Where σ is the Sigmoid activation function, ψ is the convolution operation, [*,*] represents feature concatenation, ReLU is the rectified linear unit activation function, and Conv... 1×1 For a 1×1 convolution operation, F enc For encoder feature map, F dec This is the feature map of the decoder; S623: The U-Net convolutional neural network outputs a disinfection coverage probability map for each pixel. After thresholding, the overall and regional disinfection coverage rates are calculated to generate a comprehensive disinfection report.

9. An automatic sterilization device for medical orthopedic instruments, used to implement the automatic sterilization method for medical orthopedic instruments as described in any one of claims 1-8, characterized in that: include: The 3D scanning module is used to acquire a 3D digital model of the instrument to be disinfected and to identify complex structural feature areas on the surface of the instrument. A multi-degree-of-freedom robotic arm and a disinfection execution module located at its end; A multimodal sensor array is used to acquire multimodal sensor data. Control host, the control host includes: The simulation optimization unit is used to run the fluid dynamics simulation and generate an initial disinfection parameter set; The status assessment unit is used to run a real-time assessment model of the cleaning status. Adaptive control unit, used to run model predictive control algorithms; The image analysis unit is used to run the U-Net convolutional neural network based on the attention mechanism.

10. An automatic sterilization device for medical orthopedic instruments according to claim 9, characterized in that: Also includes: An edge computing module is connected to a multimodal sensor array and is used to perform preprocessing, feature extraction and fusion of multimodal sensor data in real time, and send the obtained fused feature vector to the status evaluation unit of the control host.

Images