Bone hammer knocking force early warning system for hip replacement surgery

Abstract

The invention relates to the technical field of intelligent early warning, and particularly discloses a bone hammer knocking force early warning system for hip replacement surgery, comprising: an acquisition module for acquiring influence information, marking a surgery score for the influence information based on manual work, and generating a training sample; the training module is used for acquiring a training set and a verification set, establishing an early warning model, and training and verifying the early warning model; the optimization module is used for optimizing hyper-parameters of the early warning model until the performance of the model reaches the expectation to obtain a standard model; and the early warning module is used for acquiring an undetermined score according to the real-time information, and sending early warning information to a user for prompting when the undetermined score is smaller than or equal to a preset undetermined score threshold value. According to the invention, the accuracy and safety of an operation can be improved, and intraoperative and postoperative complications are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent early warning, and particularly relates to a bone hammer knocking force early warning system for hip replacement surgery. Background Art

[0002] A medical bone hammer is a commonly used surgical tool, usually made of stainless steel or titanium alloy to ensure its durability and safety. In terms of appearance, one end of it is a flat or slightly curved hammer head for knocking bones or tools, and the other end may be smooth or slightly protruding for different operation requirements, such as assisting bone reduction, implant fixation, and osteotomy surgery, etc.

[0003] During the actual use process, the magnitude of the bone hammer knocking force has a direct impact on the patient. The strength of the knocking force needs to be precisely controlled according to the specific requirements of the surgical operation. If the knocking force is not properly controlled, it may have an adverse impact on the patient. For example, when the knocking force is too large, the following impacts may occur: (1) Bone damage: Excessive knocking force may cause bone damage during fracture reduction or implant fixation, and even result in bone cracks or comminuted fractures; (2) Soft tissue damage: Excessive force may affect the surrounding soft tissues (such as muscles, ligaments, or nerves), leading to further damage; (3) Affect the stability of the implant: It may cause the implant to be overly embedded in the bone, affecting the stability of the prosthesis or fixation device, and may require additional repair operations. (4) It may cause bone damage during fracture reduction or implant fixation, and even result in bone cracks or comminuted fractures, and may also affect the surrounding soft tissues (such as muscles, ligaments, or nerves), leading to further damage. If the knocking force is too small, the following impacts may occur: (1) Unable to achieve the implant effect: Insufficient force may cause the prosthesis to be unable to be accurately implanted into the corresponding position and unable to achieve a good press-fit fixation effect, affecting the quality of postoperative bone integration; (2) Unstable implant fixation: If the knocking force is insufficient, the implant may not be firmly fixed to the bone, thereby increasing the risk of postoperative loosening or dislocation; (3) Prolong the operation time: Insufficient force may lead to repeated knocking, thereby prolonging the operation time and increasing the patient's anesthesia and infection risks, and may also cause the fracture site to be unable to be accurately reduced, affecting the quality of postoperative healing. Therefore, how to ensure the accuracy of the knocking force has become an urgent problem to be solved. Summary of the Invention

[0004] The purpose of the present invention is to provide a bone hammer knocking force early warning system for hip replacement surgery to solve the following technical problems:

[0005] Improper control of the knocking force may have an adverse impact on the patient. For example, when the knocking force is too large, it may cause bone damage during fracture reduction or implant fixation, and even result in bone cracks or comminuted fractures; if the knocking force is too small, it may cause the prosthesis to be unable to achieve accurate press-fit fixation, affecting the quality of postoperative prosthesis and bone integration.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A bone hammer impact force warning system for hip replacement surgery, comprising:

[0008] Acquisition module: collecting impact information, the impact information includes patient information, surgery information and real-time tapping force, the patient information includes age, bone density and bone health status score, and the surgery information includes surgery type and surgery progress, manually annotating the impact information with surgery scores, generating training samples, a single training sample includes an impact score and a corresponding surgery score, and different training samples have at least one different impact information;

[0009] Training module: randomly grouping all training samples to obtain a training set and a validation set, wherein the ratio of the number of training samples in the training set to the number of training samples in the validation set is a preset value, establishing an early warning model based on a deep learning model, training the early warning model based on the training set, and training the early warning model based on the validation set;

[0010] Wherein, the early warning model is trained by using a back propagation algorithm, and the early warning model is verified by using a five-fold cross validation;

[0011] Optimization module: obtaining the performance parameters of the early warning model after training and verification, the performance parameters including accuracy, precision and F1 score, calculating the performance score of the early warning model based on the superior and inferior solution distance method, and when the performance score is less than a preset performance score threshold, updating the hyperparameters of the early warning model, training and verifying the early warning model after the updated hyperparameters again, repeating the above steps until the performance score of the early warning model is greater than or equal to the performance score threshold after a certain hyperparameter optimization, and taking the early warning model at this time as the standard model;

[0012] Early warning module: periodically collects influencing information during the current operation and records it as real-time information, inputs the real-time information into the standard model, outputs the operation score, and marks it as a pending score. When the pending score is less than or equal to the preset pending score threshold, an early warning message is sent to the user for prompting.

[0013] As a further solution of the present invention: in the acquisition module, the real-time striking force is obtained by a bone hammer with a force detection device or directly by the force detection device.

[0014] As a further solution of the present invention: the bone hammer with a force detection device includes: a hammer handle and a hammer body, the hammer handle and the hammer body are connected, and a force sensor is arranged in the hammer body.

[0015] As a further solution of the present invention: the hammer body includes an upper hammer body and a lower hammer body, both the upper hammer body and the lower hammer body are provided with through holes for passing bolts, the force sensor is fixed by bolts between the upper hammer body and the lower hammer body, and the bolts are all detachable.

[0016] As a further solution of the present invention: in the warning module, the following steps are further included:

[0017] Obtain the knocking force in the real-time information, record it as the real-time force, and draw a curve of the real-time force changing with time, and visually display the curve;

[0018] Among them, based on the transmitter, the analog quantity of the real-time force obtained by the force detection device is converted into a digital quantity, and the digital quantity is transmitted to the computer device through the transmitter. The computer device receives the force through the upper computer, and the communication methods between the computer device and the transmitter include Ethernet communication, CAN communication, and serial communication.

[0019] As a further solution of the present invention: the process of obtaining the real-time force specifically includes:

[0020] Obtain the real-time forces Fi-1, Fi, and Fi at three adjacent timestamps +1 , Fi +1 The corresponding timestamp of Fi is greater than the corresponding timestamp of Fi, and the corresponding timestamp of Fi is greater than the corresponding timestamp of Fi-1;

[0021] If Fi-1 < Fi > Fi +1 , then take Fi as the real-time force within the period corresponding to the three timestamps.

[0022] As a further solution of the present invention: in the warning module, before sending a warning message to the user for prompting, the following steps are further included:

[0023] Set the knocking force range. If Fi belongs to the knocking force range, no alarm is given;

[0024] If Fi does not belong to the knocking force range, an alarm is given;

[0025] Among them, the alarm methods include voice alarm and image alarm.

[0026] As a further solution of the present invention: in the acquisition module, the number of training samples is greater than or equal to the preset sample number threshold.

[0027] As a further solution of the present invention: in the training module, the deep learning model includes random forest and support vector regression.

[0028] As a further solution of the present invention: in the optimization module, the method for updating the hyperparameters of the warning model includes Bayesian optimization, random search, and grid search.

[0029] As a further solution of the present invention: in the warning module, the following steps are further included:

[0030] Plot the curve f(t) of the to-be-determined score changing with time, where t represents time;

[0031] Obtain the monotonicity of the curve f(t);

[0032] If the curve f(t) is monotonically increasing, no warning information is sent;

[0033] If the curve f(t) is monotonically decreasing, send warning information to indicate that the overall effect of the surgery is poor;

[0034] If the curve f(t) has no monotonicity, execute the following steps:

[0035] Calculate the judgment value If the judgment value is greater than or equal to the preset judgment value threshold, it is determined that the overall effect of the surgery is good, Pys represents the to-be-determined score threshold, and [t1, t2] represents the domain of the curve f(t).

[0036] The beneficial effects of the present invention: compared with the prior art, the present invention can:

[0037] 1) Realize real-time monitoring of the knocking force

[0038] The force detection device can monitor the force of each knock in real time and feed the data back to the doctor, enabling the doctor to accurately control the knocking force and avoid the problems of excessive or insufficient force; through force monitoring, the doctor can better adapt to the bone strength of different patients (for example, the bones of osteoporosis patients are more fragile) and surgical requirements.

[0039] 2) Improve the surgical accuracy

[0040] During the fracture reduction or implant fixation process, excessive or insufficient knocking force may cause bone fracture or implant instability. The force detection device can help the doctor find the optimal knocking force range, thus ensuring accurate operation; the force detection device can assist the doctor in adjusting the operation rhythm, optimizing the number of knocks and the force, so that the bone block, prosthesis or implant can fit more closely;

[0041] 3) Reduce the intraoperative and postoperative risks of patients

[0042] The force detection device can prevent additional fractures, bone cracks or bone injuries caused by excessive force; during implant fixation or osteotome operation, it can avoid damage to surrounding tissues (such as ligaments, blood vessels and nerves) caused by excessive percussion; accurate percussion force can improve the fracture healing effect and reduce the risk of postoperative prosthesis loosening, implant displacement or dysfunction;

[0043] 4) Improve the visualization and standardization of operations

[0044] The force detection device can display the real-time force value on the screen and even prompt the doctor whether the current percussion force is appropriate through sound or vibration; the visualization and recording of force data help the doctor to conduct postoperative analysis, summarize operation experience and provide data support for the standardization of operations;

[0045] 5) Assist doctors in training and experience improvement

[0046] The force detection device is particularly important for novice doctors, which can help them understand the correct percussion force through real-time feedback and quickly accumulate experience; for senior doctors, the force detection device can be used as an auxiliary tool to further improve the operation precision;

[0047] 6) Data recording and postoperative analysis

[0048] The force detection device can record the force, frequency and time of each percussion, generating a complete operation data log. These data can be used for postoperative analysis and improvement of surgical strategies; the data can also be used for teaching, research and product development to promote the continuous optimization of surgical tools and methods;

[0049] 7) Promote the development of intelligent surgical tools

[0050] The application of the force detection device is an important step towards intelligent surgical tools, which can work in cooperation with other intelligent devices (such as surgical robots or navigation systems) to improve the automation and precision of operations. Brief Description of the Drawings

[0051] The present invention will be further described below with reference to the accompanying drawings.

[0052] Figure 1 It is a schematic structural diagram of a bone hammer percussion force warning system for hip replacement surgery according to the present invention;

[0053] Figure 2 is a bone hammer with a force detection device;

[0054] Figure 3 is Figure 2 mechanical explosion diagram of;

[0055] Markings in the figure: 1 - hammer handle, 2 - upper hammer body, 3 - lower hammer body, 4 - force sensor, 5 - through hole, 6 - bolt, 7 - adjusting nut, 8 - boss. Specific implementation

[0056] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0057] Please refer to Figure 1 As shown, the present invention is a bone hammer knocking force warning system for hip replacement surgery, including:

[0058] Collection module: Collect influence information. The influence information includes patient information, surgical information, and real-time knocking force. The patient information includes age, bone density, and bone health status score, and the surgical information includes surgical type and surgical progress. Based on manual annotation of the influence information with a surgical score, training samples are generated. Each training sample includes an influence score and a corresponding surgical score, and at least one piece of influence information is different in different training samples;

[0059] Training module: Randomly group all the training samples to obtain a training set and a validation set. The ratio of the number of training samples in the training set to the number of training samples in the validation set is a preset value. An early warning model is established based on a deep learning model, the early warning model is trained based on the training set, and the early warning model is trained based on the validation set;

[0060] Among them, the backpropagation algorithm is used to train the early warning model, and five-fold cross-validation is used to validate the early warning model;

[0061] Optimization module: Obtain the performance parameters of the early warning model after training and validation. The performance parameters include accuracy, precision, and F1 score. Calculate the performance score of the early warning model based on the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS). When the performance score is less than a preset performance score threshold, update the hyperparameters of the early warning model, and retrain and validate the early warning model with the updated hyperparameters. Repeat the above steps until after a certain optimization of the hyperparameters, the performance score of the early warning model is greater than or equal to the performance score threshold. Take the early warning model at this time as the standard model;

[0062] Early warning module: Periodically collect the impact information during the current surgical procedure, record it as real-time information, input the real-time information into the standard model, output the surgical score, and mark it as the pending score. When the pending score is less than or equal to the preset pending score threshold, an early warning message will be sent to the user for prompt.

[0063] It should be noted that the acquisition module is responsible for collecting the key information that affects the surgical score, such as patient information, surgical information, and real-time tapping force. By obtaining data such as the patient's age, bone density, and skeletal health status score, as well as the surgical type and progress, it can provide multi-dimensional input for the model to ensure the comprehensiveness of model training; in addition, the correlation between the manually marked surgical score and the impact score provides an accurate target value for subsequent training. This step provides rich and diverse training data for the model to help the model learn accurate scoring patterns; then, the training module will randomly group all training samples to generate a training set and a validation set. Setting the ratio of the training set and the validation set can ensure that the training data and validation data of the model have good representativeness and avoid overfitting at the same time. During the training process, the introduction of a deep learning model can automatically extract features and rules from the data, and the backpropagation algorithm continuously optimizes the prediction ability of the model by adjusting the weights. Adopting the five-fold cross-validation method can effectively verify the performance of the model on different data subsets, reduce bias, and improve the generalization ability of the model. The advantage of this step is that by training and validating the model, it helps to ensure that the early warning model has strong accuracy and robustness; then, the optimization module calculates the performance score of the early warning model through the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) based on the performance parameters (such as accuracy, precision, and F1 score) after training and validation. This method ensures that we can quantitatively evaluate the quality of the model and make necessary adjustments according to the score. When the performance score is lower than the set threshold, the update of hyperparameters and retraining can improve the performance of the model, ensuring that each optimization can improve the prediction accuracy of the model until the standard performance level is reached. Through repeated optimization, this process can finally obtain an early warning model with the best performance, providing reliable technical support for actual use; finally, the early warning module regularly obtains real-time surgical impact information, inputs it into the standard model for scoring and outputs the pending score. If the pending score is lower than the preset threshold, the system will send an early warning message in time to remind the user that there may be risks. The key role of this module in the solution is to combine the trained model with the actual surgical process, be able to monitor in real time and send out early warnings in time, thereby improving the safety of the surgical process. This step ensures that the system can not only provide accurate surgical scores based on historical data, but also provide instant risk alerts during actual surgeries to ensure the safety of patients.

[0064] It should be noted that in this solution, the surgical effect is preferentially judged. If the surgical effect is good (when the to-be-determined score is greater than the preset to-be-determined score threshold), no warning information is sent. In the case of a good surgical effect, there is no need for warning. If the surgical effect is not good, then the problem of the tapping force will be judged, avoiding the inconvenience caused by frequent adjustment of the tapping force. It can be understood that here it refers to whether the effect within a cycle is good;

[0065] In another preferred embodiment of the present invention, in the acquisition module, the real-time tapping force is obtained through a mallet with a force detection device or directly through the force detection device.

[0066] The mallet with a force detection device includes: a hammer handle and a hammer body. The hammer handle and the hammer body are connected, and a force sensor is provided inside the hammer body. The hammer body includes an upper hammer body and a lower hammer body. Both the upper hammer body and the lower hammer body are provided with through holes for passing bolts. The upper hammer body and the lower hammer body fix the force sensor through bolts, and the bolts are all detachable. Specifically, the hammer handle and the hammer body can be detachably connected or fixedly connected. In this specific implementation, the hammer handle and the hammer body are detachably connected, and an adjusting nut is provided between them. A boss is provided on the side of the upper hammer body, and an internal thread is provided inside the boss. One end of the hammer handle is provided with an external thread. The upper hammer body and the hammer handle are connected by threads, and an adjusting nut is also provided at the connection between the two. The boss can also be provided on the lower hammer body.

[0067] The upper hammer body and the lower hammer body can wrap the side of the force sensor or not wrap the side of the force sensor. The data collected by the force sensor is processed by the data processing module, and then the tapping force data is sent by the data sending module. The data processing module and the data sending module can be installed inside the hammer body. If installed inside the hammer body, the hammer body is in a groove shape, or they can be installed inside the hammer handle, and the hammer handle is designed to be hollow. The data processing module can use the ADS1120 module to amplify the signal and filter it through a filter circuit, and needs to be equipped with a chip with wireless functions such as ESP32. The data sending module can use low-power Bluetooth, wifi or LoRa. The power management of the above modules is powered by a lithium battery, and the lithium battery is installed inside the hammer handle.

[0068] In another preferred embodiment of the present invention, in the warning module, the following steps are further included:

[0069] Obtain the tapping force in the real-time information, record it as the real-time force, and draw a curve of the real-time force changing with time, and visually display the curve;

[0070] Among them, the transmitter converts the analog quantity of the real-time force obtained by the force detection device into a digital quantity, and transmits the digital quantity to the computer device through the transmitter. The computer device receives the force through the host computer. The communication methods between the computer device and the transmitter include Ethernet communication, CAN communication, and serial communication.

[0071] In another preferred embodiment of the present invention, the process of obtaining the real-time force specifically includes:

[0072] Obtain the real-time forces Fi-1, Fi, and Fi at three adjacent timestamps +1 , Fi +1 The corresponding timestamp of Fi is greater than the timestamp of Fi, and the timestamp of Fi is greater than the timestamp of Fi-1;

[0073] If Fi-1 < Fi > Fi +1 , then Fi is used as the real-time force within the period corresponding to the three timestamps.

[0074] In another preferred embodiment of the present invention, in the warning module, before sending a warning message to the user for prompting, the following steps are further included:

[0075] Set the knocking force range. If Fi belongs to the knocking force range, no alarm is made;

[0076] If Fi does not belong to the knocking force range, an alarm is made;

[0077] Among them, the alarm methods include voice alarm and image alarm.

[0078] In another preferred embodiment of the present invention, in the acquisition module, the number of training samples is greater than or equal to a preset sample number threshold.

[0079] It can be understood that in the design of the acquisition module, it is required that the number of training samples is greater than or equal to a preset sample number threshold. This requirement helps to ensure the sufficiency and representativeness of the training data. The increase in the number of samples allows the model to come into contact with more actual cases, thereby improving the learning ability of the model. Through a large number of training samples, the model can more comprehensively understand the relationship between surgical scores and influencing information in different situations, learn more general laws, and avoid bias problems caused by over-reliance on a small number of samples. In addition, more samples also help to improve the generalization ability of the model, reduce the risk of overfitting, and thus improve the performance of the model in actual applications.

[0080] In another preferred embodiment of the present invention, in the training module, the deep learning model includes random forest and support vector regression.

[0081] It is worth noting that Random Forest is an ensemble learning method that improves the prediction accuracy by constructing multiple decision trees and combining the prediction results of these trees. Random Forest effectively avoids the overfitting problem that is prone to occur in a single decision tree by randomly selecting features from the data multiple times and constructing decision trees. Each tree is trained on a subset of the data, and different random selections can be made for the construction of each tree, enhancing the diversity and robustness of the model. By combining the prediction results of multiple trees, Random Forest can provide more accurate and stable predictions. In the surgical risk early warning system, Random Forest can help the model extract effective patterns from complex medical data and improve the adaptability to different patient conditions. Its advantages are that it can handle high-dimensional data and non-linear relationships, and is relatively insensitive to noise in the data;

[0082] Support Vector Regression (SVR) is the regression application of Support Vector Machine (SVM). Support Vector Machine classifies by finding the optimal hyperplane, while Support Vector Regression makes numerical predictions by finding the optimal regression boundary. In SVR, the goal of the model is to find a hyperplane that minimizes the deviation of data points from this plane, so as to obtain accurate prediction values in the regression task. Support Vector Regression is particularly suitable for handling high-dimensional data and can find potential non-linear relationships in the training data, which makes it particularly useful in the surgical score prediction task. The advantages of SVR are that it can handle complex data structures by using kernel functions and has strong generalization ability, and can provide accurate predictions even with limited training samples.

[0083] In another preferred embodiment of the present invention, in the optimization module, the method for updating the hyperparameters of the early warning model includes Bayesian optimization, random search, and grid search.

[0084] It is worth noting that Bayesian optimization is a global optimization method based on Bayesian statistics, especially suitable for optimizing high-dimensional and computationally expensive black-box functions. In hyperparameter optimization, Bayesian optimization uses a probability model to describe the hyperparameter space and gradually selects the hyperparameter combination most likely to improve the model performance. The hyperparameter space of deep learning models is usually very large, and using Bayesian optimization can effectively reduce the waste of computing resources and accelerate the optimization process;

[0085] Random search is an optimization method that searches by randomly selecting hyperparameter values. It does not rely on gradient information, but randomly selects several combinations from all possible hyperparameter spaces for evaluation. Different from grid search, random search does not exhaust all possible combinations. Instead, it explores the hyperparameter space through random sampling and can avoid wasting computing resources in some irrelevant hyperparameter ranges;

[0086] Grid search is an optimization method that searches by exhausting all possible hyperparameter combinations. It sets a range of candidate values for each hyperparameter and then tries each combination to ensure that the entire hyperparameter space is traversed, and it can guarantee to find the optimal combination in the hyperparameter space.

[0087] In another preferred embodiment of the present invention, in the warning module, the following steps are further included:

[0088] Plot the curve f(t) of the to-be-determined score changing with time, where t represents time;

[0089] Obtain the monotonicity of the curve f(t);

[0090] If the curve f(t) is monotonically increasing, no warning message is sent;

[0091] If the curve f(t) is monotonically decreasing, a warning message is sent to indicate that the overall effect of the surgery is poor;

[0092] If the curve f(t) has no monotonicity, the following steps are executed:

[0093] Calculate the judgment value If the judgment value is greater than or equal to the preset judgment value threshold, it is judged that the overall effect of the surgery is good, Pys represents the to-be-determined score threshold, and [t1, t2] represents the domain of the curve f(t).

[0094] It should be noted that if the overall duration of the surgery is less than the preset duration threshold, the above steps of obtaining monotonicity and subsequent steps are not executed.

[0095] The above has described a detailed description of an embodiment of the present invention, but the above content is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the present invention application should still fall within the scope covered by the patent of the present invention.

Claims

1. An osteotome hammer percussion force warning system for hip replacement surgery, characterized in that, Including: Collection module: Collect influencing information, where the influencing information includes patient information, surgical information, and real-time tapping force. Based on manual annotation of the surgical score for the influencing information, training samples are generated. Each training sample includes an influencing score and the corresponding surgical score, and there is at least one difference in the influencing information among different training samples. Training module: Randomly group all training samples to obtain a training set and a validation set, and the ratio of the number of training samples in the training set to the number of training samples in the validation set is a preset value. Establish an early warning model based on a deep learning model, train the early warning model based on the training set, and train the early warning model based on the validation set. Optimization module: Obtain the performance parameters of the early warning model after training and validation, calculate the performance score of the early warning model. When the performance score is less than the preset performance score threshold, update the hyperparameters of the early warning model, retrain and validate the early warning model with the updated hyperparameters, and repeat the above steps until, after a certain hyperparameter optimization, the performance score of the early warning model is greater than or equal to the performance score threshold, and use the early warning model at this time as the standard model. Early warning module: Periodically collect the influencing information during the current surgical process, denoted as real-time information, input the real-time information into the standard model, output the surgical score, and mark it as a pending score. When the pending score is less than or equal to the preset pending score threshold, send a warning message to the user for prompt.

2. The bone hammer percussion force warning system for hip replacement surgery according to claim 1, characterized in that, In the collection module, the patient information includes age, bone density, and bone health status score, and the surgical information includes surgical type and surgical progress; in the optimization module, the performance parameters include accuracy, precision, and F1 score, and calculate the performance score of the early warning model based on the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS).

3. The bone hammer impact force warning system for hip replacement surgery according to claim 1, characterized in that, In the collection module, the real-time tapping force is obtained by a bone hammer with a force detection device or directly by a force detection device.

4. The bone hammer percussion force warning system for hip replacement surgery according to claim 3, wherein, The bone hammer with a force detection device includes: a hammer handle and a hammer body, which are connected between the hammer handle and the hammer body, and a force sensor is provided in the hammer body.

5. The bone hammer impact force warning system for hip replacement surgery according to claim 4, characterized in that, The hammer body includes an upper hammer body and a lower hammer body, and through holes are provided in both the upper hammer body and the lower hammer body for passing bolts. The upper hammer body and the lower hammer body fix the force sensor through bolts, and the bolts are all detachable.

6. The bone hammer impact force warning system for hip replacement surgery according to claim 3, wherein, In the early warning module, the following steps are also included: Obtain the tapping force in the real-time information, denoted as real-time force, and draw a curve of the real-time force changing with time, and visually display the curve. Among them, based on a transmitter, convert the analog quantity of the real-time force obtained by the force detection device into a digital quantity, and transmit the digital quantity to a computer device through the transmitter. The computer device receives the force through a host computer, and the communication methods between the computer device and the transmitter include Ethernet communication, CAN communication, and serial communication.

7. The bone hammer percussion force warning system for hip replacement surgery according to claim 6, wherein The process of obtaining the real-time force specifically includes: Obtain the real-time force F of three adjacent timestamps i-1 , F i and F i+1 , F i+1 The corresponding timestamp of F is greater than that of F i , the corresponding timestamp of F i The corresponding timestamp of F is greater than that of F i-1 the corresponding timestamp; If F i-1 <F i >F i+1 , then F i is used as the real-time intensity within the periods corresponding to the three timestamps.

8. A bone hammer impact force warning system for hip replacement surgery according to claim 7, characterized in that, In the early warning module, before sending a warning message to the user for prompt, the following steps are also included: Set the range of tapping force. If F i falls within the range of tapping force, no alarm will be given; If F i is not within the range of tapping force, an alarm is given; Among them, the alarm methods include voice alarm and image alarm.

9. The bone hammer impact force warning system for hip replacement surgery according to claim 1, wherein In the described training module, the deep learning model includes a random forest and support vector regression; in the described optimization module, the methods for updating the hyperparameters of the early warning model include Bayesian optimization, random search, and grid search.

10. The bone hammer impact force warning system for hip replacement surgery according to claim 1, wherein In the described early warning module, the following steps are further included: Plot the curve f(t) of the to-be-evaluated score varying with time, where t represents time; Obtain the monotonicity of the curve f(t); If the curve f(t) is monotonically increasing, no early warning information is sent; If the curve f(t) is monotonically decreasing, early warning information is sent to indicate that the overall effect of the surgery is poor; If the curve f(t) has no monotonicity, perform the following steps: Calculate the judgment value If the calculated judgment value is greater than or equal to a preset judgment value threshold, it is determined that the overall effect of the surgery is good. Pys represents the to-be-determined scoring threshold, and [t1, t2] represents the domain of the curve f(t).

Images