A method for intelligent management of neurosurgical instruments

Abstract

The present invention relates to an intelligent management method for neurosurgical instruments, comprising the following steps: obtaining specification parameters, performance indicators and material information of neurosurgical instruments from different manufacturers, and establishing an instrument attribute feature library; based on the instrument attribute feature vector, clustering analysis is performed on similar instruments from different manufacturers; if a temporary replacement of an instrument is required, the instrument attribute feature vector of a spare instrument is obtained from the instrument attribute feature library; if the functional matching degree of the instrument to be replaced is lower than a preset risk threshold, an instrument interchange risk reminder is issued to the doctor; after the instrument replacement is confirmed, the specification parameters of the instrument to be replaced are obtained, and compared with the specification parameters of the surgical operation specification; during the operation, the use status of the target instrument is monitored in real time, and the remaining number of uses of the target instrument is predicted. The present invention improves the accuracy and efficiency of surgical instrument replacement, reduces surgical risks, and optimizes instrument management.

Description

Technical Field

[0001] The present invention relates to the technical field of medical instruments, and in particular to an intelligent management method for neurosurgery instruments. Background Art

[0002] In the intelligent management of neurosurgical surgical instruments, the instrument adaptability evaluation system faces a key technical challenge: how to quickly and accurately judge the interchangeability and functional matching between instruments from different manufacturers when it is necessary to temporarily replace spare instruments during surgery. The complexity of this problem lies in the fact that instruments produced by different manufacturers differ in design, material, specifications, etc. Even for instruments of the same type, their performance parameters and scope of application may not be the same. The evaluation system needs to analyze the similarity between the instrument to be replaced and the alternative instrument in real time in the massive instrument data, and comprehensively consider the accuracy requirements of the surgical operation, the doctor's personal preferences and other factors to give the best replacement plan. At the same time, the accuracy and efficiency of the instrument adaptability judgment directly affect the continuity and stability of the surgical process. If the replacement plan given by the evaluation system is inappropriate, it may cause the instrument to be unstable during surgery, damage to tissues, and other problems, prolong the operation time and increase the risk of infection. Frequent instrument replacement will also interrupt the rhythm of the operation, affect the doctor's operation, and may even endanger the patient's life safety.

[0003] Therefore, the instrument adaptability evaluation system also needs to have an in-depth understanding of the working mechanisms of different instruments in surgical operations, establish a correlation model between instrument performance and surgical procedures, optimize the instrument replacement strategy through intelligent algorithms, and maximize the continuity and smooth progress of the operation. Summary of the invention

[0004] In order to solve the problems existing in the above-mentioned prior art, the present invention aims to provide an intelligent management method for neurosurgical surgical instruments. The intelligent management method for neurosurgical surgical instruments improves the accuracy and efficiency of surgical instrument replacement, reduces surgical risks, and optimizes instrument management.

[0005] The present invention provides a method for intelligent management of neurosurgical instruments, comprising the following steps:

[0006] S1. Obtain the specification parameters, performance indicators and material information of neurosurgical instruments from different manufacturers, establish an instrument attribute feature library, extract the movement precision and instrument material indicators of the instruments, form a numerical feature vector and use it as an instrument attribute feature vector, summarize the instrument attribute feature vectors, and build an instrument attribute feature library;

[0007] S2. Based on the instrument attribute feature vector, cluster analysis is performed on similar instruments from different manufacturers, and the similarity between different instruments is determined according to the motion precision of the surgical instruments and the clustering results, and the instrument interchangeability mapping relationship is obtained, and an instrument interchangeability mapping table is constructed;

[0008] S3. If it is necessary to temporarily replace the instrument, the instrument attribute feature vector of the spare instrument is obtained from the instrument attribute feature library, and the instrument with the highest similarity to the spare instrument is searched in the instrument interchangeability mapping relationship. The functional matching degree of the instrument to be replaced is obtained by combining the precision of the surgical operation and the frequency of surgical perspective conversion;

[0009] S4. If the functional matching degree of the device to be replaced is lower than the preset risk threshold, a device interchange risk reminder is issued to the doctor, and the brand, weight and handle of the device to be replaced are obtained according to the doctor's personal preference;

[0010] S5. After the device replacement is confirmed, the specification parameters of the device to be replaced are obtained and compared with the specification parameters of the surgical operation specification. If the device to be replaced meets the requirements of the surgical operation specification, the device to be replaced is determined to be the target device;

[0011] S6. During the operation, the usage status of the target instrument is monitored in real time, and the remaining usage times of the target instrument are predicted. When the predicted value is lower than the preset replenishment threshold, the instrument spare parts replenishment process is triggered.

[0012] Preferably, the step S1 specifically includes:

[0013] An optical scanner is used to obtain the three-dimensional morphological data of neurosurgical instruments, and the stress threshold and material hardness value of the instruments are obtained through a mechanical testing platform to obtain the original data set of the instruments;

[0014] According to the original data set of the device, a Gaussian filter is used to process the surface morphology of the device, and a characteristic spectrum of the device contour curve is obtained by Fourier transform to obtain a device surface feature data set;

[0015] For the device surface feature data set, a mapping relationship between device feature parameters and performance indicators is established through an artificial neural network, a comprehensive parameter matrix is ​​input, and a device performance prediction value is output;

[0016] According to the predicted value of the instrument performance, the random forest algorithm is used to process the instrument motion trajectory data, and the instrument attribute feature vector is obtained through a numerical calculation method to obtain an instrument attribute feature library.

[0017] Preferably, the step S2 specifically includes:

[0018] Acquire device attribute feature vectors from the device attribute feature library, and group the device attribute feature vectors using a hierarchical clustering algorithm to obtain an initial device classification data set;

[0019] A principal component decomposition method is used to perform a dimensionality reduction operation on the initial classification data set of the device, and a similarity quantification calculation is performed on the result of the dimensionality reduction operation by a multidimensional scaling method to obtain a device similarity data set;

[0020] The tensile strength of the device similarity dataset is measured using a tensile testing machine, the elastic modulus is obtained through a bending test, and the device mechanical property dataset is obtained based on the fatigue test data;

[0021] An interchangeability evaluation matrix is ​​constructed according to the instrument similarity data set and the instrument mechanical property data set, and an interchangeability matching relationship is extracted from the interchangeability evaluation matrix by a numerical calculation method to obtain an instrument interchangeability mapping table.

[0022] Preferably, the step S3 specifically includes:

[0023] The cosine similarity calculation method is used to calculate the feature vectors of the backup device and the original device, and the calculation of the feature vector is based on the feature data in the device attribute feature library;

[0024] Acquire a device similarity data set according to the feature vector, wherein the device similarity data set is generated by a feature matching algorithm between the backup device and the original device;

[0025] Recording the motion trajectory of the instrument in the instrument similarity data set by a motion capture device, wherein the motion trajectory of the instrument includes displacement data and three-axis posture angle data;

[0026] A convolutional neural network is used to process the instrument motion trajectory to obtain an instrument operation precision index, and a support vector machine prediction model is established based on the instrument operation precision index and the instrument similarity data set. The support vector machine prediction model outputs a function matching score.

[0027] Preferably, the step S4 specifically includes:

[0028] Compare the functional matching value of the device to be replaced with the preset risk threshold, obtain the basic risk value of the device through a numerical calculation method, and generate a risk level identification according to the preset risk classification rules;

[0029] Obtain the physician's historical selection data on device brand, weight range, and handle type from the physician habit record database, establish a physician habit prediction model through support vector machine, and obtain the physician's device adaptability score;

[0030] Matching the basic parameters of the device to be replaced with the physician's device adaptability score, determining the adaptability of the device to the physician's operating habits through a quantitative calculation method, and generating device adaptability score data;

[0031] A comprehensive risk assessment model is constructed using a neural network algorithm, the risk level identifier, the device adaptability score data and the storage adaptability data are input, the comprehensive risk value of device replacement is determined, and graded risk warning information is generated according to a preset risk threshold.

[0032] Preferably, the step S5 specifically includes:

[0033] Acquire the size data of the device to be replaced by an optical measuring instrument, acquire the material parameters of the device to be replaced by a mechanical sensor, and generate a device specification parameter set according to the size data and the material parameters;

[0034] Extracting the surgical type classification standard from the surgical specification database according to the instrument specification parameter set, the surgical type classification standard including the surgical site, the degree of trauma and the duration of the operation, and generating a specification requirement parameter set;

[0035] A support vector machine is used to construct a specification parameter evaluation model, and the specification standard compliance is calculated according to the instrument specification parameter set and the specification requirement parameter set, wherein the specification standard compliance indicates the degree of matching between the instrument specification and the surgical requirement;

[0036] A parameter deviation data set is obtained by performing Gaussian filtering on the degree of conformity to the specification standard, and a comprehensive deviation score is calculated based on the parameter deviation data set. If the comprehensive deviation score is less than a preset score threshold, the device to be replaced is determined to be the target device.

[0037] Preferably, the step S6 specifically includes:

[0038] A sensor array is used to obtain the force parameters and motion trajectory data of the target device, wherein the sensor array includes a mechanical sensor and a displacement sensor, and a photoelectric sensor is used to measure the surface wear value of the target device to obtain a device status monitoring data set;

[0039] Inputting the equipment status monitoring data set into a long short-term memory neural network for feature extraction to obtain a numerical value of the degree of wear of the target equipment;

[0040] Perform fatigue performance testing on the target device by using a stress strain tester, and establish a life decay curve based on the target device history database records;

[0041] The degree of wear and tear is cumulatively calculated according to the life decay curve, and the remaining number of uses of the target device is predicted through a deep learning algorithm. If the remaining number of uses is lower than a preset replenishment threshold, an equipment replenishment application form is generated.

[0042] The intelligent management method for neurosurgery instruments described in the present invention has the following advantages:

[0043] The intelligent management method for neurosurgery instruments of the present invention can quickly and accurately find the spare instrument with the highest similarity to the instrument to be replaced by establishing an instrument attribute feature library and performing cluster analysis, thereby reducing the waiting time during the operation and improving the operation efficiency; by constructing an instrument interchangeability mapping table, the doctor can more conveniently understand the similarity and interchangeability between different instruments, which helps to make more reasonable choices during the operation; when it is determined that the instrument needs to be replaced, the system can provide personalized instrument selection suggestions based on the doctor's personal preferences, enhancing the comfort of the operation and the doctor's satisfaction; real-time monitoring of the use status of the target instrument and prediction of the remaining number of uses help to discover potential problems of the instrument in advance, replenish spare parts in time, and avoid interruptions or delays in the operation due to instrument failure; through intelligent instrument management, the use and demand of the instrument can be more accurately grasped, which helps hospitals or medical institutions to more reasonably allocate and dispatch medical resources and improve resource utilization efficiency. The intelligent management method for neurosurgery instruments improves the accuracy and efficiency of surgical instrument replacement, reduces surgical risks, and optimizes instrument management. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of a method for intelligent management of neurosurgical instruments described in the present invention. DETAILED DESCRIPTION

[0045] like Figure 1 As shown, the method for intelligent management of neurosurgical instruments described in the present invention comprises the following steps:

[0046] S1. Obtain the specification parameters, performance indicators and material information of neurosurgical instruments from different manufacturers, establish an instrument attribute feature library, extract the movement precision and instrument material indicators of the instruments, form a numerical feature vector and use it as the instrument attribute feature vector, summarize the instrument attribute feature vectors, and build an instrument attribute feature library;

[0047] S2. Based on the instrument attribute feature vector, cluster analysis is performed on similar instruments from different manufacturers. According to the motion precision of the surgical instruments and the clustering results, the similarity between different instruments is determined, the instrument interchangeability mapping relationship is obtained, and an instrument interchangeability mapping table is constructed; wherein, cluster analysis of similar instruments from different manufacturers is performed using a clustering algorithm;

[0048] S3. If it is necessary to temporarily replace the instrument, the instrument attribute feature vector of the spare instrument is obtained from the instrument attribute feature library, and the instrument with the highest similarity to the spare instrument is found in the instrument interchangeability mapping relationship. The functional matching degree of the instrument to be replaced is obtained by combining the precision of the surgical operation and the frequency of surgical perspective conversion;

[0049] S4. If the functional matching degree of the device to be replaced is lower than the preset risk threshold, a device interchange risk reminder is issued to the doctor, and the brand, weight and handle of the device to be replaced are obtained according to the doctor's personal preference;

[0050] S5. After the device replacement is confirmed, the specification parameters of the device to be replaced are obtained and compared with the specification parameters of the surgical operation specification. If the device to be replaced meets the requirements of the surgical operation specification, the device to be replaced is determined to be the target device;

[0051] S6. During the operation, the usage status of the target instrument is monitored in real time, and the remaining number of uses of the target instrument is predicted. When the predicted value is lower than the preset replenishment threshold, the instrument spare parts replenishment process is triggered. Specifically, by obtaining the target instrument usage frequency data, combined with the instrument failure rate and service life, a machine learning algorithm is used to predict the remaining number of uses of the target instrument.

[0052] Furthermore, in this embodiment, step S1 specifically includes:

[0053] An optical scanner is used to obtain the three-dimensional morphological data of neurosurgical instruments, and the stress threshold and material hardness value of the instruments are obtained through a mechanical testing platform to obtain the original data set of the instruments;

[0054] According to the original data set of the device, the surface morphology of the device is processed by using a Gaussian filter, and the characteristic spectrum of the device contour curve is obtained by Fourier transform to obtain the device surface feature data set;

[0055] For the device surface feature data set, the mapping relationship between the device feature parameters and performance indicators is established through artificial neural network, the comprehensive parameter matrix is ​​input, and the device performance prediction value is output;

[0056] According to the predicted value of device performance, the random forest algorithm is used to process the device motion trajectory data, and the device attribute feature vector is obtained through numerical calculation method to obtain the device attribute feature library;

[0057] Specifically, the neurosurgical instruments are modeled and collected in three dimensions through optical scanners, the surface morphology and size parameters of the instruments are collected through laser sensors, and the stress threshold and material hardness values ​​of the instruments are obtained using a mechanical testing platform to generate the original data set of the instruments.

[0058] For the surface morphology features in the original data set of the device, the surface roughness value is extracted using a Gaussian filter, and the characteristic spectrum is obtained by processing the device contour curve through Fourier transform to establish the device surface feature data set;

[0059] According to the mechanical test data, the elastic coefficient and mechanical strength value of the material are extracted from the original data set of the device, and the comprehensive parameter matrix is ​​constructed by combining the device surface feature data set. The key characteristic parameters of the device are extracted by principal component analysis.

[0060] Fatigue testing equipment is used to obtain the wear resistance and fatigue life data of the device, and an artificial neural network is used to establish a mapping model between the key characteristic parameters of the device and the performance indicators. The input is a comprehensive parameter matrix and the output is the predicted value of the device performance.

[0061] The motion trajectory data of the equipment is collected based on the 3D motion capture system, and combined with the motion precision measurement value, a motion feature model is constructed through the random forest algorithm. The input is the motion trajectory data of the equipment, and the output is the motion precision index.

[0062] According to the predicted value of device performance and the motion precision index, the device attribute feature vector is generated by numerical calculation method, and the parameters of different dimensions are unified through normalization processing to establish the device attribute feature library;

[0063] The accuracy and safety of neurosurgical instruments are crucial to the success of surgery, so it is of great significance to conduct quantitative analysis of instrument attributes and characteristics to achieve standardized evaluation;

[0064] Here is an example:

[0065] In practical application, the needle holder is used as the test object to illustrate the specific implementation process. Its geometric shape data is collected through three-dimensional optical scanning, and the morphological characteristics of the surface roughness in the range of 0.2 to 0.8 microns are obtained by laser sensor scanning. The stress threshold is measured by the mechanical testing platform as 520 MPa and the material hardness value is 58HRC;

[0066] After the surface morphology of the needle holder was processed by Gaussian filtering, the surface roughness value was extracted to be 0.35 microns. Fourier transform analysis was applied to the instrument contour curve, and the main frequency component in the characteristic spectrum was obtained to be distributed in the range of 200 to 800 Hz, reflecting the level of instrument processing accuracy. On this basis, the elastic modulus of the material was measured to be 210 GPa and the tensile strength was 1100 MPa. The comprehensive parameter matrix constructed in combination with the surface characteristics was used to extract key features through principal component analysis, and the cumulative contribution rate reached 85%;

[0067] The fatigue test results show that the wear of the needle holder is 0.02 mm after 100,000 cycles, and the fatigue life is expected to reach 500,000 cycles. These data are input into the artificial neural network model, which adopts a three-layer structure, 20 hidden nodes, and 2,000 training samples. After the model converges, the prediction accuracy reaches 93%;

[0068] The suture action of the needle holder was recorded by a three-dimensional motion capture system with a sampling frequency of 200 Hz, and the deviation of the motion trajectory was within 0.1 mm;

[0069] The random forest algorithm uses 500 decision trees to build a motion feature model. The input variables include three motion parameters: position, speed, and acceleration. After training, the accuracy of evaluating motion precision reaches 91%.

[0070] The final attribute feature vector contains 30 dimensions, covering mechanical properties, material characteristics and motion characteristics. The maximum and minimum value normalization process is used to achieve dimensional unification. All feature values ​​are mapped to the range of 0 to 1, which facilitates horizontal comparison between different devices.

[0071] The established attribute feature library can also be used to evaluate other types of surgical instruments, such as microscissors and hemostatic forceps, to achieve standardized quantitative evaluation of instrument performance;

[0072] Based on the attribute feature library, we conducted a comparative analysis of similar instruments produced by different manufacturers and found that a certain brand of needle holders performed outstandingly in terms of motion precision, with a trajectory repeatability error of less than 0.05 mm, but had a relatively short fatigue life of only 300,000 cycles. Characteristic vector analysis revealed that this was related to the high hardness of its material and the large elastic modulus, which provided a direction for the optimization and improvement of the instrument.

[0073] Furthermore, step S1 further includes:

[0074] The instruments include microscissors and nerve forceps. The key attributes of microscissors include blade length, hardness and sharpness, and the key attributes of nerve forceps include jaw size and gripping force. Movement precision includes displacement accuracy of instrument operation and vibration accuracy of the instrument end.

[0075] An optical measuring instrument is used to scan and measure the blade of the micro scissors. The optical measuring instrument obtains the blade shape data and sharpness value through a laser sensor and an infrared spectrometer to obtain the geometric feature data set of the micro scissors.

[0076] The reference parameters are established according to the geometric characteristic data set of the microscissors, and the pressure distribution curve and stress state of the nerve clamp jaws are collected through the mechanical sensor and strain measurement device to obtain the geometric characteristic data set of the nerve clamp;

[0077] For the geometric feature data sets of microscissors and neural forceps, an artificial neural network is used to establish the instrument feature mapping relationship, and the instrument operation stability index is output from the mapping relationship;

[0078] According to the instrument operation stability index, the motion trajectory data is collected through the displacement sensor and acceleration sensor, and the motion precision prediction model is established using the random forest algorithm to obtain the instrument key attribute feature database;

[0079] Specifically, an optical measuring instrument is used to scan the blade of the micro-scissors, the blade shape and length data are obtained through a laser sensor, the blade hardness value is measured using a Vickers hardness tester, the blade sharpness value is obtained from an infrared spectrometer, and a geometric feature data set of the micro-scissors is generated;

[0080] The pressure distribution curve of the nerve clamp jaws during closing is obtained from the mechanical sensor, and the distance change data during opening and closing of the jaws is collected using the displacement sensor. The stress state of the jaw surface is measured using the strain gauge, and the size parameters of the jaws are measured using the digital micrometer to generate the nerve clamp geometric feature data set.

[0081] The mechanical sensor is used to collect the pressure data in the process of nerve clamping in real time, the force distribution state of the jaws is obtained through the strain measurement device, and the clamping force change curve of the instrument is recorded in combination with the pressure sensor array to construct the nerve clamp holding force characteristic data set;

[0082] For the geometric feature data sets of microscissors and nerve forceps, an artificial neural network is used to establish the instrument feature mapping relationship. The input parameters include geometric dimensions and material performance data, and the output parameters are instrument operation stability indicators.

[0083] A high-speed camera system is used to obtain the movement trajectory of the instrument end, and the displacement change data of the instrument is recorded through a displacement sensor. The vibration signal is collected by an acceleration sensor, and a motion precision prediction model is established using a random forest algorithm.

[0084] Based on the displacement accuracy value and vibration accuracy parameters output by the motion precision prediction model and the force distribution data in the nerve clamp gripping force feature data set, a database of key attribute characteristics of the instrument is generated;

[0085] Here is an example:

[0086] Microscissors are precision neurosurgery instruments. The accuracy of their blade length directly affects the standardization of surgical incisions. Scanned by a high-precision optical measuring instrument, the blade length is 25 mm, the radius of curvature is 0.15 mm, the surface roughness of the blade is 0.2 microns, the Vickers hardness value reaches 650HV, and infrared spectrum analysis shows that the blade sharpness coefficient is 0.92.

[0087] The jaw design of the nerve forceps plays a key role in the safety of tissue clamping. The mechanical sensor recorded that the pressure distribution during the closing process of the jaws showed a uniform and gradual trend. The maximum clamping pressure was 2.5 Newtons. The jaw opening and closing distance can be adjusted from 0 to 8 mm. The strain gauge measured that the stress distribution on the jaw surface was uniform, with a maximum stress value of 420 MPa. The digital micrometer measured that the jaw width was 3 mm and the length was 12 mm.

[0088] In the actual application scenario of the nerve clamp, the pressure sensor array is arranged at the key stress points of the jaws, and the real-time clamping force variation range is between 0.5 and 2 Newtons. The strain measurement device records that the stress difference on both sides of the jaws is less than 5%. The pressure distribution curve is bell-shaped, and the peak value appears in the center area of ​​the jaws, which confirms the uniformity of the clamping force.

[0089] In the process of establishing the device feature mapping relationship, the artificial neural network adopts a three-layer structure. The input layer contains 20 nodes corresponding to geometric features and material performance parameters, the number of hidden layer nodes is 40, and the 5 nodes of the output layer correspond to stability evaluation indicators. The number of training samples reaches 1,000 groups. After the model converges, the prediction accuracy reaches 94%;

[0090] The high-speed camera system records the movement trajectory of the instrument end at a sampling rate of 1000 frames per second. The displacement sensor measures the displacement accuracy better than 0.05 mm. The vibration frequency collected by the acceleration sensor is mainly distributed in the range of 20 to 200 Hz, and the amplitude is less than 0.02 mm.

[0091] The motion precision prediction model constructed by the random forest algorithm contains 200 decision trees, with parameters such as displacement, speed, and acceleration as feature variables, and the accuracy of motion precision evaluation reaches 92%;

[0092] Through comprehensive analysis, it was found that the sharpness of microscissors was positively correlated with the hardness of the blade, with a correlation coefficient of 0.85, while the uniformity of the clamping force of the nerve clamp was closely related to the uniformity of the stress distribution on the jaw surface, and the deviation coefficient of the two was less than 0.1;

[0093] The results of the instrument movement precision evaluation show that there is a trade-off between displacement accuracy and vibration accuracy. The optimized comprehensive performance indicators show that the operating stability of the microscissors reaches 0.95, and the clamping accuracy of the nerve forceps reaches 0.93, both of which meet the requirements of clinical application.

[0094] Furthermore, in this embodiment, step S2 specifically includes:

[0095] Obtaining device attribute feature vectors from a device attribute feature library, grouping the device attribute feature vectors using a hierarchical clustering algorithm, and obtaining an initial device classification data set;

[0096] The principal component decomposition method is used to reduce the dimension of the initial classification data set of the device, and the multidimensional scaling method is used to quantify the similarity of the results of the dimension reduction operation to obtain the device similarity data set;

[0097] The tensile strength of the device similarity dataset was measured using a tensile testing machine, the elastic modulus was obtained through a bending test, and the device mechanical property dataset was obtained based on the fatigue test data;

[0098] An interchangeability evaluation matrix is ​​constructed based on the device similarity data set and the device mechanical property data set, and the interchangeability matching relationship is extracted from the interchangeability evaluation matrix through a numerical calculation method to obtain a device interchangeability mapping table;

[0099] Specifically, neurosurgical instruments from different manufacturers were grouped using a hierarchical clustering algorithm based on the instrument attribute feature vectors in the instrument attribute feature library, and the Euclidean distance between instruments was calculated based on the instrument size parameters, shape features, and material properties to generate an initial instrument classification data set.

[0100] The principal component decomposition method was used to reduce the dimension of the initial classification data set of the device, and the device clustering coefficient and grouping identifier were extracted. The similarity of different device groups was quantitatively calculated by multidimensional scaling method to generate a device similarity data set.

[0101] Conduct mechanical property tests on each group of instruments in the instrument similarity dataset. Use a tensile testing machine to measure the tensile strength of the instruments. Use a bending test to obtain the elastic modulus of the instruments. Combine the fatigue test data to generate an instrument mechanical property dataset.

[0102] The displacement trajectory data of the instrument is collected through a three-dimensional motion capture device, the three-axis posture angle changes of the instrument are recorded based on an inertial sensor, and the deformation of the instrument is measured using a strain sensor to generate an instrument motion feature data set;

[0103] For the instrument motion feature data set, a deep learning algorithm is used to establish a motion feature recognition model. The input parameters include displacement trajectory, posture angle and deformation data, and the output parameter is the instrument operation accuracy index.

[0104] Combining the device similarity data set, mechanical property data set and operation accuracy index, a device interchangeability evaluation matrix is ​​constructed through numerical calculation method, from which the interchangeability matching relationship is extracted to generate a device interchangeability mapping table;

[0105] Here is an example:

[0106] The interchangeability assessment of neurosurgical instruments involves multi-dimensional feature analysis. Taking micro-needle holders as an example, similar products produced by 10 different manufacturers were grouped using a hierarchical clustering algorithm. The instrument size parameters showed that the length of the forceps head was distributed in the range of 10 to 15 mm, the diameter of the handle was 2 to 3 mm, the surface roughness was in the range of 0.2 to 0.5 μm, and the material was medical stainless steel. The calculated Euclidean distance matrix reflects the similarity between the instruments.

[0107] The principal component decomposition process showed that the cumulative contribution rate of the first three principal components reached 85%, of which the first principal component reflected the size characteristics of the device, accounting for 45%, the second principal component reflected the material performance, accounting for 25%, and the third principal component characterized the surface characteristics, accounting for 15%. Multidimensional scaling analysis classified the 10 devices into three similarity groups, and the intra-group similarity coefficient was greater than 0.85;

[0108] The results of mechanical properties tests showed that the tensile strength ranged from 800 to 1200 MPa, the elastic modulus was from 190 to 210 GPa, and the performance attenuation rate after 500,000 cycles of fatigue testing was less than 5%. The differences in mechanical properties of the devices in different groups were mainly reflected in fatigue life, with the maximum difference reaching 200,000 cycles.

[0109] During the motion feature acquisition process, the sampling frequency of the 3D motion capture device was 200 Hz, the pitch angle recorded by the inertial sensor was within the range of plus or minus 30 degrees, the roll angle was plus or minus 45 degrees, and the yaw angle was plus or minus 60 degrees. The maximum deformation measured by the strain sensor was 0.2 mm, and the displacement trajectory repeatability error was less than 0.1 mm.

[0110] The deep learning model uses a 5-layer convolutional neural network structure. The input layer contains 15 feature parameters, the number of hidden layer nodes is 128, 64, 32, and 16, and the output layer is 3 accuracy indicators, including position accuracy, posture accuracy, and deformation. The number of training samples is 2,000 groups, and the accuracy of the verification set reaches 92%;

[0111] In the final interchangeability mapping table, the first similarity group includes 4 instruments, the difference in motion accuracy index is less than 5%, and the deviation of mechanical performance parameters is within 10%. The second group includes 3 instruments, the difference in motion accuracy index is between 5% and 10%, and the deviation of mechanical performance parameters is within 15%. The third group includes 3 instruments, and the differences in various indicators are large, so it is not recommended to use them interchangeably;

[0112] The mapping table can be used to intuitively determine the interchangeability between instruments from different manufacturers. Instruments in the first similarity group can be directly interchanged, instruments in the second group need to undergo adaptability assessment, and instruments in the third group do not meet the interchangeability conditions.

[0113] Further, step S2 also includes:

[0114] Based on the results of cluster analysis, the similarity between the devices is quantified by calculating the distance or cosine similarity between the attribute feature vectors corresponding to the key attributes of different devices, and a device interchangeability mapping table is constructed based on the similarity scores. The device interchangeability mapping table includes elements representing the similarity between the devices.

[0115] According to the geometric size, material performance and motion parameters of the device, the maximum and minimum value standardization method is used to normalize the device attribute feature vector, and the device distance feature set is obtained through the Euclidean distance formula;

[0116] For the device attribute feature vector, the cosine similarity calculation method is used to analyze the projection relationship of the feature vector in the high-dimensional space to obtain the device similarity numerical matrix;

[0117] According to the device distance feature set and similarity value matrix, the hierarchical clustering method is used to group the devices, and the differences between the groups are judged by the minimum distance criterion to obtain the final device grouping data set;

[0118] For the final grouped data set, a support vector machine was used to establish an interchangeability prediction model, and an interchangeability mapping table was constructed through numerical matrix operations to obtain the device interchangeability matching score;

[0119] Specifically, according to the cluster analysis results, the key attributes of each group of devices, including geometric dimensions, material properties and motion parameters, are extracted. The device attribute feature vectors are normalized using the maximum and minimum value normalization method. The distance values ​​between different device feature vectors are calculated using the Euclidean distance formula to generate a device distance feature set.

[0120] The cosine similarity calculation method is used to perform directional analysis on the device attribute feature vectors, and the vector angle is calculated based on the projection relationship of the feature vectors in the high-dimensional space to generate a device similarity numerical matrix;

[0121] Based on the preset clustering threshold, the device distance feature set is quantitatively graded, and the similarity of the devices is scored in combination with the similarity value matrix to generate an initial device grouping data set;

[0122] The hierarchical clustering method is used to verify the similarity within the initial grouping data set of the device, and the minimum distance criterion is used to determine the difference between the groups to generate the final grouping data set of the device;

[0123] For the final grouped data set of instruments, an interchangeability prediction model was established based on support vector machine. The input parameters included instrument distance features and similarity values, and the output parameter was the interchangeability matching score.

[0124] The interchangeability of the same group of devices is quantitatively evaluated based on the interchangeability matching scores, and an interchangeability mapping table is constructed through numerical matrix operation methods to achieve quantitative characterization of the interchangeability relationship of the devices;

[0125] Here is an example:

[0126] The interchangeability evaluation of neurosurgical instruments involves multi-dimensional feature analysis. Taking nerve stripping forceps as an example, 15 similar products from different manufacturers were collected for analysis, and key attributes including forceps head length, jaw width, material hardness, surface roughness and operation accuracy parameters were extracted. All feature values ​​were mapped to the range of 0 to 1 through maximum and minimum value normalization processing, and the calculated Euclidean distance values ​​were distributed in the range of 0.15 to 0.85.

[0127] Cosine similarity analysis showed that the device feature vectors showed an obvious clustering trend in the 30-dimensional space. Device pairs with vector angles less than 30 degrees showed high similarity in performance. 90% of the element values ​​in the similarity value matrix were distributed in the range of 0.6 to 0.95, reflecting that most devices have a certain potential for interchangeability.

[0128] The clustering threshold was set to 0.75, and the nerve stripping forceps were initially grouped. Instruments with similarity higher than the threshold were classified into the same group. The initial grouping results showed that the 15 products could be divided into 4 groups, of which the first group contained 6 instruments, the second group contained 4 instruments, the third group contained 3 instruments, and the fourth group contained 2 instruments. The geometric parameters and performance indicators of the instruments in the groups showed strong consistency.

[0129] The minimum distance criterion was used in the hierarchical clustering validation process. When the minimum distance between groups was greater than 0.4, they were judged as different categories. The average similarities within the groups were calculated to be 0.82, 0.78, 0.73, and 0.69, respectively, and the maximum similarity between groups was 0.35, which confirmed the rationality of the grouping results.

[0130] The support vector machine uses radial basis kernel function to construct an interchangeability prediction model. The training sample contains 2000 sets of historical interchangeability verification data. The cross-validation accuracy rate reaches 91%. The model's prediction results of interchangeability of new devices are highly consistent with actual application performance.

[0131] In the interchangeability mapping table, the interchangeability matching scores of the first group of 6 instruments were all over 0.85, showing good interchangeability performance, the scores of the second group of 4 instruments were between 0.75 and 0.85, the scores of the third group of 3 instruments were between 0.65 and 0.75, and the scores of the fourth group of 2 instruments were lower than 0.65 due to large differences in characteristics;

[0132] The interchangeability between different devices is intuitively presented in matrix form, where the diagonal element value is 1, indicating a perfect match, and the off-diagonal element value reflects the degree of interchangeability between the two devices. The closer the value is to 1, the better the interchangeability.

[0133] Furthermore, in this embodiment, step S3 specifically includes:

[0134] The cosine similarity calculation method is used to calculate the feature vectors of the backup device and the original device. The feature vector calculation is based on the feature data in the device attribute feature library;

[0135] Acquire a device similarity data set according to the feature vector, where the device similarity data set is generated by a feature matching algorithm between the backup device and the original device;

[0136] The motion trajectory of the equipment in the equipment similarity dataset is recorded by a motion capture device, and the equipment motion trajectory includes displacement data and three-axis posture angle data;

[0137] The convolutional neural network is used to process the instrument motion trajectory to obtain the instrument operation precision index. A support vector machine prediction model is established based on the instrument operation precision index and the instrument similarity data set. The support vector machine prediction model outputs the function matching score.

[0138] Specifically, based on the feature vector data in the device attribute feature library, the cosine similarity calculation method is used to calculate the similarity between the backup device and the original device, and the device combination with the highest similarity is screened through the feature matching algorithm to generate a device similarity data set;

[0139] The displacement trajectory of the instrument during the operation is collected through motion capture equipment, the three-axis posture angle changes of the instrument are recorded using inertial sensors, and the instrument operation force is measured using strain sensors to generate an instrument motion feature data set;

[0140] The surgical field of view transformation process is recorded based on an optical tracking device, and the field of view transformation frequency data is obtained using an angle sensor. Combined with the field of view transformation amplitude parameters, a surgical field of view feature data set is generated.

[0141] The convolutional neural network is used to process the instrument motion feature data set. The input parameters include displacement trajectory and posture angle, and the output parameter is the instrument operation precision index.

[0142] The frequency score of visual field conversion is calculated based on the surgical visual field feature data set, and the instrument function adaptability data set is generated by combining the instrument operation precision index and similarity data;

[0143] For the instrument function adaptability data set, a support vector machine was used to establish a function matching prediction model. The input parameters included instrument similarity, operation precision, and visual field switching frequency, and the output parameter was the function matching score.

[0144] Here is an example:

[0145] Temporary replacement of neurosurgical instruments involves multi-dimensional matching evaluation. The replacement of microscissors is used as an example to illustrate the specific implementation process. The feature vector of the candidate microscissors is extracted from the instrument attribute feature library, including the head length of 25 mm, the blade thickness of 0.3 mm and the hardness value of 650 HV. The cosine similarity calculation shows that the similarity with the original microscissors is 0.92, indicating that the two have a high consistency in basic attributes.

[0146] The sampling frequency of the motion capture device was set to 200 Hz, and the displacement trajectory of the microscissors during the operation was recorded, with a maximum displacement amplitude of 30 mm. The inertial sensor measured the three-axis attitude angle change range as plus or minus 25 degrees for pitch angle, plus or minus 35 degrees for roll angle, and plus or minus 45 degrees for yaw angle. The strain sensor showed that the operating force was distributed in the range of 0.5 to 2 Newtons.

[0147] The optical tracking device records the field of view change data of the surgical microscope. Statistics show that the field of view changes 4 to 6 times per minute, with a single change angle between 15 and 45 degrees. The frequency of the instrument remaining stable during the field of view change is 75%, reflecting the high requirements for instrument control accuracy in the current surgical stage.

[0148] The convolutional neural network uses a 5-layer structure to process the motion characteristics of the instrument. The input layer contains the displacement trajectory and posture angle data. The spatiotemporal features are extracted through 3 convolutional layers. The fully connected layer outputs an operation precision score of 0.88, indicating that the alternative microscissors meet the requirements of surgical precision operation.

[0149] The frequency of visual field conversion was scored using a normalized method, taking into account the two dimensions of conversion times and amplitude, and the calculated comprehensive score was 0.82. Combining the instrument operation precision index and similarity data, the generated functional adaptability matrix showed that the alternative microscissors met the replacement requirements in all indicators;

[0150] The support vector machine uses the radial basis kernel function to build a function matching prediction model. The weights of the input parameters are similarity 0.4, operation precision 0.35, and field of view conversion frequency 0.25. The number of model training samples is 1000 groups, and the cross-validation accuracy rate reaches 93%;

[0151] The calculated functional matching score was 0.87, which was higher than the preset threshold of 0.8, proving that the alternative microscissors were fully capable of meeting the operational requirements of the current surgical stage;

[0152] In practical applications, this evaluation method is also applicable to the temporary replacement of other types of surgical instruments, such as microtweezers, nerve stripping forceps and other surgical instruments. Through a comprehensive evaluation of the three dimensions of similarity, operation precision and frequency of field of view conversion, the replacement of instruments is accurately matched with surgical needs.

[0153] Data show that the success rate of instrument replacement using this method reaches 95%, significantly improving the continuity of the surgical process;

[0154] Further, step S3 also includes:

[0155] According to the attribute feature vector of the backup instrument, the instrument interchangeability mapping table is searched row by row, and the similarity value between the backup instrument and each instrument is calculated. The instrument to be replaced is selected according to the similarity value. The similarity value represents the quantitative expression of the functional matching degree of the two instruments, and the matching degree represents the impact of the interchange of the two instruments on the surgery;

[0156] According to the device attribute feature vector, the data of the backup device is standardized by a feature matching algorithm to obtain a device feature data matrix;

[0157] The cosine similarity calculation method is used to traverse the rows and columns of the device feature data matrix, and the attribute distance between the backup device and each device in the interchangeability mapping table is calculated using the Euclidean distance formula to obtain the initial similarity data set;

[0158] The initial similarity data set is reduced in dimension by principal component analysis, and attribute weights are set according to feature contribution rates to obtain a device similarity score set.

[0159] A functional attribute hierarchical tree is constructed based on the device similarity score set, and a matching prediction model is established using the random forest algorithm. If the functional matching score is greater than the preset matching threshold, a device replacement priority table is generated;

[0160] Specifically, according to the device size parameters, material characteristics and performance indicators in the device attribute feature vector, the data of the spare devices is standardized through the feature matching algorithm, and the weight coefficient of each attribute is calculated by the hierarchical analysis method to generate the device feature data matrix;

[0161] The cosine similarity calculation method is used to traverse the rows and columns of the device feature data matrix, and the attribute distance between the backup device and each device in the interchangeability mapping table is calculated according to the Euclidean distance formula to generate an initial similarity data set;

[0162] The initial similarity data set is reduced in dimension by principal component analysis, the key characteristic parameters of the device are extracted, the attribute weights are set according to the characteristic contribution rate, and the device similarity score set is generated;

[0163] A functional attribute hierarchical tree is constructed based on the device similarity score set, and a matching prediction model is established using the random forest algorithm. The input parameters include similarity scores and attribute weights, and the output parameter is the functional matching score.

[0164] Arrange the devices in descending order according to the functional matching scores, use the priority sorting algorithm to generate the device replacement sequence, and select qualified devices based on the preset matching threshold to build the final replacement plan;

[0165] The final replacement plan is quantified through the numerical matrix calculation method, the device function matching mapping relationship is established, and the device replacement priority table is generated;

[0166] Here is an example:

[0167] The interchangeability evaluation of neurosurgical surgical instruments is described by taking the microscopic needle holder as an example. First, the attribute feature vector of the spare needle holder is extracted, including the clamp head length of 12 mm, the jaw width of 2.5 mm, the material hardness of 58 HRC and the surface roughness of 0.3 μm. The geometric dimension weight of 0.4, the material performance weight of 0.35 and the surface characteristic weight of 0.25 are calculated by the hierarchical analysis method. After standardization, the feature data matrix is ​​generated.

[0168] The similarity between the spare needle holder and 10 similar instruments in the interchangeability mapping table was calculated. The cosine similarity values ​​ranged from 0.65 to 0.95, and the similarity with 3 instruments exceeded 0.9, indicating that these instruments had high consistency in basic features.

[0169] The Euclidean distance calculation showed that the minimum distance value was 0.12, and the corresponding instrument had a deviation of less than 5% in each parameter;

[0170] The results of principal component analysis show that the cumulative contribution rate of the first three principal components reaches 88%, of which the first principal component reflects geometric characteristics, accounting for 45%, the second principal component reflects material properties, accounting for 28%, and the third principal component represents surface characteristics, accounting for 15%;

[0171] The weight coefficient is optimized according to the feature contribution rate, and the generated similarity score is distributed in the range of 0 to 1. The functional attribute hierarchy tree contains 3 layers of structure. The first layer is the geometric size class, which includes length, width and thickness nodes. The second layer is the material performance class, which includes hardness, strength and toughness nodes. The third layer is the surface property class, which includes roughness and friction coefficient nodes.

[0172] The random forest algorithm used 500 decision trees and 2,000 training samples, with a prediction accuracy of 94%. The priority ranking of replacement instruments showed that the needle holder had the highest functional matching score of 0.93, the maximum deviation in geometric dimensions from the original instrument was 0.2 mm, the material hardness difference was 2 HRC, and the surface roughness difference was 0.05 microns;

[0173] The function matching score of the second candidate device was 0.89, and the deviations of various parameters were slightly large but still within the acceptable range. The function matching threshold was set at 0.85, and a total of 3 candidate devices that met the requirements were screened out;

[0174] The replacement priority table finally generated is in the form of a matrix, with rows representing alternative devices and columns representing evaluation indicators, including similarity scores, functional matching scores, and comprehensive ranking information;

[0175] This table can be used to intuitively judge the pros and cons of each alternative instrument and select the most appropriate replacement plan based on specific surgical needs;

[0176] Actual application data showed that the success rates of device replacement for the top three priority groups were 98%, 95% and 92%, respectively, confirming the reliability of the evaluation method.

[0177] Furthermore, in this embodiment, step S4 specifically includes:

[0178] Compare the functional matching value of the device to be replaced with the preset risk threshold, obtain the basic risk value of the device through a numerical calculation method, and generate a risk level identification according to the preset risk classification rules;

[0179] Obtain the physician's historical selection data on device brand, weight range, and handle type from the physician habit record database, establish a physician habit prediction model through support vector machine, and obtain the physician's device adaptability score;

[0180] The basic parameters of the device to be replaced are matched with the physician's device adaptability score, and the adaptability of the device to the physician's operating habits is determined by quantitative calculation methods to generate device adaptability score data;

[0181] A comprehensive risk assessment model is constructed using a neural network algorithm, which inputs risk level identification, device adaptability score data, and storage adaptability data to determine the comprehensive risk value of device replacement and generate graded risk warning information based on preset risk thresholds;

[0182] Specifically, the functional matching value of the device to be replaced is compared with the preset threshold, the basic risk value of the device is evaluated by numerical calculation method, the risk level identification is generated according to the risk classification rules, and the initial data set of device risk is established;

[0183] Extract the physician's historical selection data on device brand, weight range and handle type from the physician habit record database, and combine it with the device usage frequency record to generate the physician device preference data set;

[0184] A support vector machine was used to establish a physician habit prediction model, with input parameters including brand preference, weight range, and handle type data, and the output parameter being the physician device adaptability score;

[0185] Data matching is performed based on the basic parameters of the device to be replaced and the physician's device adaptability score, and the matching degree between the device and the physician's operating habits is evaluated through quantitative calculation methods to generate a device adaptability score data set;

[0186] The data association algorithm is used to evaluate the storage space size and temperature and humidity requirements of the equipment, and the storage condition parameters of the equipment to be replaced are combined to generate a storage adaptability data set;

[0187] A comprehensive risk assessment model is constructed based on a neural network algorithm. The input parameters include initial risk data, adaptability score and storage adaptability data. The output parameter is the comprehensive risk value of device replacement.

[0188] Classify the comprehensive risk value of device replacement according to the preset risk threshold, generate graded risk warning information, and construct a device replacement risk assessment report;

[0189] Here is an example:

[0190] The risk assessment of neurosurgical instrument replacement is explained by taking microscissors as an example. The functional matching degree of the microscissors to be replaced is 0.82, which is lower than the preset threshold of 0.85. According to the risk grading rules, it is classified as medium risk, and the corresponding risk level is 2, indicating that there is a potential risk in the replacement of the instrument;

[0191] The initial risk data set contains information such as a matching degree deviation value of 0.03, a risk level coefficient of 0.6, and a warning level parameter of 0.7;

[0192] The physician's habit record showed that the physician used German brand microscissors 85% of the time in the past 6 months, preferred weights between 15 and 18 grams, and chose straight handles 90% of the time;

[0193] The preference data set also records the physician’s specific usage parameters of the instrument, including information on a holding angle of 75 degrees, an operating force of 1.2 Newtons, and a handle diameter of 8 mm;

[0194] The support vector machine uses the radial basis kernel function to build a habit prediction model. The input layer contains 15 feature parameters, the number of training samples is 1000 groups, and the cross-validation accuracy rate reaches 92%;

[0195] The model prediction showed that the brand matching of the microscissors to be replaced was 0.75, the weight matching was 0.88, the handle matching was 0.82, and the comprehensive adaptability score was 0.81;

[0196] The storage temperature of the microscissors to be replaced should be between 18 and 22 degrees Celsius, with a relative humidity of 40% to 60%, and the occupied space should be 250×50×30 mm;

[0197] The temperature of the existing equipment storage environment is maintained at 20±1 degrees Celsius, the relative humidity is 50%±5%, the remaining storage space is 300×60×40 mm, and the storage adaptability score is 0.95;

[0198] The neural network adopts a 5-layer structure, including 3 hidden layers, with 64, 32, and 16 nodes respectively. The input parameter weights are 0.4 for risk initial data, 0.35 for adaptability score, and 0.25 for storage adaptability;

[0199] After model calculation, the comprehensive risk value of device replacement is 0.78, corresponding to a medium risk level, and it is necessary to focus on brand matching issues;

[0200] The risk assessment report pointed out three key issues: the device function match was 0.03 percentage points below the threshold, which was significantly different from the physician’s brand preference, with a brand match of only 0.75;

[0201] Other aspects such as weight, handle shape and storage conditions basically meet the requirements. Based on the evaluation results, it is recommended to find more suitable alternative equipment or take additional adaptive training measures;

[0202] Comprehensive evaluation data show that although the microscissors have shortcomings in some aspects, the overall risk is within a controllable range and it still has a certain degree of replacement feasibility while ensuring surgical safety.

[0203] Furthermore, in this embodiment, step S5 specifically includes:

[0204] Acquire the size data of the device to be replaced through an optical measuring instrument, acquire the material parameters of the device to be replaced from a mechanical sensor, and generate a device specification parameter set based on the size data and material parameters;

[0205] Extract the surgical type classification standard from the surgical specification database according to the instrument specification parameter set. The surgical type classification standard includes the surgical site, degree of trauma and duration of the operation, and generate a specification requirement parameter set;

[0206] A support vector machine was used to construct a specification parameter evaluation model, and the specification standard compliance was calculated based on the instrument specification parameter set and the specification requirement parameter set. The specification standard compliance indicates the degree of match between the instrument specification and the surgical requirements.

[0207] Gaussian filtering is performed on the specification standard compliance to obtain a parameter deviation data set, and a comprehensive deviation score is calculated based on the parameter deviation data set. If the comprehensive deviation score is less than a preset score threshold, the device to be replaced is determined to be the target device;

[0208] Specifically, the information of the device to be replaced is extracted according to the device replacement confirmation number, the device size data is collected through an optical measuring instrument, the device material parameters are obtained from a mechanical sensor, and a device specification parameter set is generated using a digital measurement method;

[0209] Extract the surgical type classification standard from the surgical specification database, divide the surgical difficulty level according to the surgical site, degree of trauma, and duration of surgery, and generate a parameter set of specification requirements;

[0210] A support vector machine was used to construct a specification parameter evaluation model, with input parameters including device size, material characteristics, and performance indicators, and output parameters being the degree of conformity to specification standards;

[0211] Calculate the deviation between the specification parameter set of the device to be replaced and the parameter set required by the specification by using the numerical comparison method, classify each parameter according to the preset standard range, and generate a parameter deviation data set;

[0212] The parameter deviation data set is smoothed based on the Gaussian filtering algorithm to eliminate abnormal deviation values, and the comprehensive deviation score is calculated using the weighted average method;

[0213] A qualified judgment threshold is set for the comprehensive deviation score, and a multi-dimensional evaluation is performed in combination with the conformity to the specification standards. If each indicator meets the threshold requirements, the device to be replaced is determined to be the target device;

[0214] Here is an example:

[0215] The specification parameter verification process of neurosurgery surgical instruments takes the replacement confirmation of microscissors as an example. According to the confirmation number NE202502, the basic information of the microscissors to be replaced is obtained. The optical measuring instrument collects the scissors length of 125 mm, the blade length of 25 mm, the blade thickness of 0.3 mm, the blade angle of 45 degrees, the material hardness value of 58HRC, the tensile strength of 1100 MPa, and the surface roughness of 0.2 microns;

[0216] Microneurosurgical surgeries are finely graded in the surgical specification database. According to the surgical site, the surgeries are divided into three areas: brainstem, thalamus and skull base. The degree of trauma is divided into three levels: minimally invasive, small incision and craniotomy. The duration of the operation is divided into three levels: within 2 hours, 2 to 4 hours, and more than 4 hours. Different difficulty levels correspond to different instrument specification requirements.

[0217] The support vector machine uses Gaussian kernel function to build the specification evaluation model. The input layer contains 12 feature parameters, the number of training samples is 1500 groups, and the cross-validation accuracy rate reaches 95%;

[0218] The model calculation results show that the size specification conformity of the microscissors to be replaced is 0.92, the material performance conformity is 0.88, and the surface characteristics conformity is 0.94;

[0219] The parameter comparison results show that the length of the blade of the microscissors to be replaced is 0.2 mm different from the standard value, the thickness of the blade is 0.02 mm, the hardness value is 2HRC, and the surface roughness is 0.03 microns. The deviation values ​​of various parameters are all within the preset standard range;

[0220] The deviation values ​​are graded according to the preset intervals, with the deviation of the cutter head length being grade 1, the deviation of the blade thickness being grade 1, the deviation of the hardness value being grade 2, and the deviation of the surface roughness being grade 1;

[0221] The Gaussian filter uses a 5×5 filter window to process the deviation data. The comprehensive deviation score after filtering is 0.91, which is higher than the preset qualified threshold of 0.85. The weight coefficients of each parameter are 0.4 for size specification, 0.35 for material performance, and 0.25 for surface characteristics. The calculated weighted average score is 0.90;

[0222] The final evaluation results showed that all indicators of the microscissors to be replaced met the requirements of surgical specifications, among which the size specifications performed most outstandingly. Although there were slight deviations in material properties, they were still within the acceptable range, and the surface characteristics fully met the standards.

[0223] According to the evaluation data, the microscissors fully meet the conditions as a target instrument and are suitable for use in the intended surgery;

[0224] Actual application data show that the actual performance of instruments screened using this evaluation method during surgery is 96% consistent with the evaluation results.

[0225] Furthermore, in this embodiment, step S6 specifically includes:

[0226] A sensor array is used to obtain the force parameters and motion trajectory data of the target device. The sensor array includes a mechanical sensor and a displacement sensor. The surface wear value of the target device is measured by a photoelectric sensor to obtain a device status monitoring data set.

[0227] According to the equipment status monitoring data set, the long short-term memory neural network is input to extract features to obtain the target equipment wear degree value;

[0228] Perform fatigue performance test on target equipment through stress-strain tester, and establish life decay curve based on historical database records of target equipment;

[0229] The loss degree values ​​are accumulated and calculated according to the life decay curve, and the remaining usage times of the target device are predicted through the deep learning algorithm. If the remaining usage times are lower than the preset replenishment threshold, a device replenishment application form is generated;

[0230] Specifically, a sensor array is used to collect the use status of the target device in real time, the mechanical sensor is used to record the force parameters of the device, the displacement sensor is used to obtain the movement trajectory of the device, and the photoelectric sensor is used to measure the surface wear value of the device to generate a device status monitoring data set;

[0231] Extract equipment failure records, maintenance records and usage frequency data based on the equipment history database, perform fatigue performance tests on the equipment through a stress strain tester, and generate an equipment life parameter data set;

[0232] Long short-term memory neural network is used to extract features from the instrument status monitoring data set. The input parameters include force value, motion trajectory and surface wear, and the output parameter is the degree of instrument wear.

[0233] Establish a life decay curve based on the device life parameter data set, and use the sliding time window method to accumulate the loss degree to generate the device remaining life data set;

[0234] A device life prediction model is built based on a deep learning algorithm. The input parameters include the remaining life value and usage frequency data, and the output parameter is the predicted number of uses.

[0235] If the predicted number of times of use is lower than the preset replenishment threshold, the stock replenishment quantity is calculated based on the frequency of use and loss rate of the device, and a device replenishment application form is generated;

[0236] Here is an example:

[0237] The life monitoring of neurosurgical instruments is explained by taking the micro needle holder as an example. The sensor array contains 12 measurement points. The mechanical sensor records that the clamping force of the instrument varies between 0.5 and 2.5 Newtons. The displacement sensor measures the movement amplitude as plus or minus 30 degrees. The photoelectric sensor shows that the wear depth of the jaw surface reaches 0.02 mm. The comprehensive data reflects that the instrument has entered the normal use stage.

[0238] The historical database shows that among the fault records of this type of needle holder, poor jaw bite accounted for 45%, insufficient clamping force accounted for 35%, and surface scratches accounted for 20%. On average, there was one fault every 200 uses.

[0239] Maintenance records show that the instrument return rate is 5%, the replacement rate is 3%, and the frequency of use is 4 to 6 times per surgery. Stress-strain testing was performed at 500,000 cycles, and the instrument fatigue strength was measured to be 800 MPa and the plastic deformation was 0.1 mm;

[0240] The long short-term memory neural network uses a three-layer structure to process monitoring data. The input layer contains three sets of parameters: force data, motion trajectory, and wear degree. The number of hidden layer nodes is 64. The output layer reflects the degree of equipment wear.

[0241] The number of model training samples was 5,000, the verification accuracy reached 93%, and the wear degree score was 0.72, indicating that the device was in a moderate wear state;

[0242] The life decay curve is nonlinear. The wear rate is 0.0001 mm per use in the first 100,000 times, and accelerates to 0.0002 mm per use in the 100,000 to 300,000 times stage. After 300,000 times, it enters a rapid decay period.

[0243] The length of the sliding time window is set to 1000 usage records, and the cumulative calculation shows that the current device has completed 250,000 operations, and the remaining life is about 100,000 times;

[0244] The deep learning model is built based on a convolutional neural network, which includes 5 convolutional layers and 3 fully connected layers. The input features include the remaining life value, the average number of daily uses, and the single use duration parameters. The output predicts that the remaining number of uses is 85,000 times.

[0245] Taking into account the uncertainties in actual use, the replenishment threshold is set at 100,000 times, and the current device is close to the threshold. According to the surgical scheduling data, the device is used 6 times a day on average, about 180 times a month. According to the current loss rate, it is expected to reach the scrap standard in 3 months;

[0246] The minimum stock quantity is calculated to be 2 pieces. Considering the procurement cycle and inventory management requirements, a replenishment requisition is finally generated with a replenishment quantity of 3 pieces.

[0247] Actual application data show that the prediction method has an accuracy rate of 90% in estimating the life of the device, effectively reducing the sudden failure rate of the device.

[0248] In the description of the present invention, it is necessary to understand that the directions or positional relationships indicated by directional words such as "front, back, up, down, left, right", "lateral, vertical, horizontal" and "top, bottom" are usually based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description. Unless otherwise specified, these directional words do not indicate or imply that the device or element referred to must have a specific direction or be constructed and operated in a specific direction. Therefore, they cannot be understood as limiting the scope of protection of the present invention.

[0249] For those skilled in the art, various other corresponding changes and deformations can be made according to the technical solutions and concepts described above, and all of these changes and deformations should fall within the protection scope of the claims of the present invention.

Claims

1. A method for intelligent management of neurosurgical instruments, characterized in that: The following steps are involved: S1. Obtain the specification parameters, performance indicators and material information of neurosurgical instruments from different manufacturers, establish an instrument attribute feature library, extract the movement precision and instrument material indicators of the instruments, form a numerical feature vector and use it as an instrument attribute feature vector, summarize the instrument attribute feature vectors, and build an instrument attribute feature library; S2. Based on the instrument attribute feature vector, cluster analysis is performed on similar instruments from different manufacturers, and the similarity between different instruments is determined according to the motion precision of the surgical instruments and the clustering results, and the instrument interchangeability mapping relationship is obtained, and an instrument interchangeability mapping table is constructed; S3. If it is necessary to temporarily replace the instrument, the instrument attribute feature vector of the spare instrument is obtained from the instrument attribute feature library, and the instrument with the highest similarity to the spare instrument is searched in the instrument interchangeability mapping relationship. The functional matching degree of the instrument to be replaced is obtained by combining the precision of the surgical operation and the frequency of surgical perspective conversion; S4. If the functional matching degree of the device to be replaced is lower than the preset risk threshold, a device interchange risk reminder is issued to the doctor, and the brand, weight and handle of the device to be replaced are obtained according to the doctor's personal preference; S5. After the device replacement is confirmed, the specification parameters of the device to be replaced are obtained and compared with the specification parameters of the surgical operation specification. If the device to be replaced meets the requirements of the surgical operation specification, the device to be replaced is determined to be the target device; S6. During the operation, the usage status of the target instrument is monitored in real time, and the remaining usage times of the target instrument are predicted. When the predicted value is lower than the preset replenishment threshold, the instrument spare parts replenishment process is triggered.

2. The method for intelligent management of neurosurgical instruments according to claim 1, characterized in that: The step S1 specifically includes: An optical scanner is used to obtain the three-dimensional morphological data of neurosurgical instruments, and the stress threshold and material hardness value of the instruments are obtained through a mechanical testing platform to obtain the original data set of the instruments; According to the original data set of the device, a Gaussian filter is used to process the surface morphology of the device, and a characteristic spectrum of the device contour curve is obtained by Fourier transform to obtain a device surface feature data set; For the device surface feature data set, a mapping relationship between device feature parameters and performance indicators is established through an artificial neural network, a comprehensive parameter matrix is ​​input, and a device performance prediction value is output; According to the predicted value of the instrument performance, the random forest algorithm is used to process the instrument motion trajectory data, and the instrument attribute feature vector is obtained through a numerical calculation method to obtain an instrument attribute feature library.

3. The method for intelligent management of neurosurgical instruments according to claim 1, characterized in that: The step S2 specifically includes: Acquire device attribute feature vectors from the device attribute feature library, and group the device attribute feature vectors using a hierarchical clustering algorithm to obtain an initial device classification data set; A principal component decomposition method is used to perform a dimensionality reduction operation on the initial classification data set of the device, and a similarity quantification calculation is performed on the result of the dimensionality reduction operation by a multidimensional scaling method to obtain a device similarity data set; The tensile strength of the device similarity dataset is measured using a tensile testing machine, the elastic modulus is obtained through a bending test, and the device mechanical property dataset is obtained based on the fatigue test data; An interchangeability evaluation matrix is ​​constructed according to the instrument similarity data set and the instrument mechanical property data set, and an interchangeability matching relationship is extracted from the interchangeability evaluation matrix by a numerical calculation method to obtain an instrument interchangeability mapping table.

4. The method for intelligent management of neurosurgical instruments according to claim 1, characterized in that: The step S3 specifically includes: The cosine similarity calculation method is used to calculate the feature vectors of the backup device and the original device, and the calculation of the feature vector is based on the feature data in the device attribute feature library; Acquire a device similarity data set according to the feature vector, wherein the device similarity data set is generated by a feature matching algorithm between the backup device and the original device; Recording the motion trajectory of the instrument in the instrument similarity data set by a motion capture device, wherein the motion trajectory of the instrument includes displacement data and three-axis posture angle data; A convolutional neural network is used to process the instrument motion trajectory to obtain an instrument operation precision index, and a support vector machine prediction model is established based on the instrument operation precision index and the instrument similarity data set. The support vector machine prediction model outputs a function matching score.

5. The method for intelligent management of neurosurgical instruments according to claim 1, characterized in that: The step S4 specifically includes: Compare the functional matching value of the device to be replaced with the preset risk threshold, obtain the basic risk value of the device through a numerical calculation method, and generate a risk level identification according to the preset risk classification rules; Obtain the physician's historical selection data on device brand, weight range, and handle type from the physician habit record database, establish a physician habit prediction model through support vector machine, and obtain the physician's device adaptability score; Matching the basic parameters of the device to be replaced with the physician's device adaptability score, determining the adaptability of the device to the physician's operating habits through a quantitative calculation method, and generating device adaptability score data; A comprehensive risk assessment model is constructed using a neural network algorithm, the risk level identifier, the device adaptability score data and the storage adaptability data are input, the comprehensive risk value of device replacement is determined, and graded risk warning information is generated according to a preset risk threshold.

6. The method for intelligent management of neurosurgical instruments according to claim 1, characterized in that: The step S5 specifically includes: Acquire the size data of the device to be replaced by an optical measuring instrument, acquire the material parameters of the device to be replaced by a mechanical sensor, and generate a device specification parameter set according to the size data and the material parameters; Extracting the surgical type classification standard from the surgical specification database according to the instrument specification parameter set, the surgical type classification standard including the surgical site, the degree of trauma and the duration of the operation, and generating a specification requirement parameter set; A support vector machine is used to construct a specification parameter evaluation model, and the specification standard compliance is calculated according to the instrument specification parameter set and the specification requirement parameter set, wherein the specification standard compliance indicates the degree of matching between the instrument specification and the surgical requirement; A parameter deviation data set is obtained by performing Gaussian filtering on the degree of conformity to the specification standard, and a comprehensive deviation score is calculated based on the parameter deviation data set. If the comprehensive deviation score is less than a preset score threshold, the device to be replaced is determined to be the target device.

7. The method for intelligent management of neurosurgical instruments according to claim 1, characterized in that: The step S6 specifically includes: A sensor array is used to obtain the force parameters and motion trajectory data of the target device, wherein the sensor array includes a mechanical sensor and a displacement sensor, and a photoelectric sensor is used to measure the surface wear value of the target device to obtain a device status monitoring data set; Inputting the equipment status monitoring data set into a long short-term memory neural network for feature extraction to obtain a numerical value of the degree of wear of the target equipment; Perform fatigue performance testing on the target device by using a stress strain tester, and establish a life decay curve based on the target device history database records; The degree of wear and tear is cumulatively calculated according to the life decay curve, and the remaining number of uses of the target device is predicted through a deep learning algorithm. If the remaining number of uses is lower than a preset replenishment threshold, an equipment replenishment application form is generated.

Images