Machine learning-based intelligent examination sheet auditing method and device

Abstract

The invention discloses an intelligent examination sheet auditing method and device based on machine learning, and relates to the field of artificial intelligence, and the method comprises the steps: obtaining to-be-audited examination sheet information; the test sheet information comprises patient name, gender, age and information of a plurality of biochemical detection items; and inputting the to-be-audited inspection sheet information into at least one pre-trained machine learning model to obtain an audit result output by the pre-trained machine learning model, and determining a final audit result according to at least one audit result. In the application, the machine learning algorithm is applied to perform intelligent auditing on the inspection sheet information, so that the auditing efficiency and accuracy are improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a method and device for intelligently reviewing inspection orders based on machine learning. Background Art

[0002] With the rapid development of modern medicine, clinical laboratory testing plays a critical role in disease diagnosis, treatment monitoring, and health assessment. With the continuous advancement of society and medical technology, the volume of test data generated daily by clinical laboratories is exploding. Traditional manual review methods are not only inefficient and unable to meet the growing demand for testing, but are also susceptible to human factors such as fatigue and experience differences, making it difficult to ensure the accuracy and consistency of review results. While automated laboratory validation offers many advantages, it inevitably has some drawbacks. From an accuracy perspective, automated validation systems rely on pre-set rules and algorithms, making them prone to misjudgment when faced with complex, variable, and unique samples. For example, in clinical testing, test data from patients with rare diseases or special health conditions may deviate significantly from the typical range. Automated validation systems may not be able to accurately identify the true significance of these abnormal data, resulting in erroneous judgments. Regarding adaptability, with the continuous emergence of new testing technologies and test indicators, automated validation systems struggle to quickly update and adapt to these changes. If the system is not updated promptly, it will be unable to effectively validate new test items, leading to bottlenecks in the testing process and inaccurate results. Therefore, the existing audit method has the problems of low audit efficiency and low accuracy. Summary of the Invention

[0003] The purpose of this application is to provide an intelligent inspection form review method and device based on machine learning, which can improve the review efficiency and accuracy.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a method for intelligently reviewing inspection orders based on machine learning, comprising:

[0006] Obtain the test order information to be reviewed; the test order information includes the patient's name, gender, age and several biochemical test items;

[0007] The inspection order information to be reviewed is input into at least one pre-trained machine learning model to obtain the review result output by the pre-trained machine learning model, and the final review result is determined based on the at least one review result.

[0008] In a second aspect, the present application provides an intelligent inspection form review device based on machine learning, comprising:

[0009] The test information acquisition module is used to obtain the test order information to be reviewed; the test order information includes the patient's name, gender, age and several biochemical test items;

[0010] The intelligent audit module is used to input the inspection order information to be audited into at least one pre-trained machine learning model, obtain the audit result output by the pre-trained machine learning model, and determine the final audit result based on at least one audit result.

[0011] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned intelligent inspection form review method based on machine learning.

[0012] In a fourth aspect, the present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements the above-mentioned machine learning-based intelligent inspection form review method.

[0013] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned machine learning-based intelligent inspection form review method.

[0014] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0015] This application provides a machine learning-based intelligent review method and device for test order information. The method obtains test order information to be reviewed; the test order information includes the patient's name, gender, age, and information on several biochemical test items; the test order information to be reviewed is input into at least one pre-trained machine learning model, and the review result output by the pre-trained machine learning model is obtained. The final review result is determined based on the at least one review result. In this application, a machine learning algorithm is applied to perform intelligent review of test order information, improving the efficiency and accuracy of the review. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 This is a diagram of an application environment of an intelligent inspection form review method based on machine learning in one embodiment of the present application;

[0018] Figure 2 A flowchart of an intelligent inspection form review method based on machine learning provided in one embodiment of the present application;

[0019] Figure 3 A schematic diagram of the functional modules of an intelligent inspection form review device based on machine learning provided in one embodiment of the present application;

[0020] Figure 4 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0022] The successful application of artificial intelligence technology in various fields has brought new opportunities for the review of clinical laboratory test orders. With its powerful data processing and analysis capabilities, artificial intelligence can quickly and accurately screen, evaluate, and judge massive amounts of test data, effectively improving review efficiency and quality and reducing human error.

[0023] In recent years, the application of machine learning in medicine has steadily increased, with significant progress particularly in medical imaging, disease prediction, and diagnosis. Machine learning algorithms can automatically learn patterns and regularities from large amounts of data, enabling efficient processing and analysis of complex data. In clinical biochemistry laboratories, machine learning can be used to build intelligent review systems to improve the efficiency and accuracy of test result review. However, surprisingly, there have been relatively few attempts to apply machine learning algorithms to the review of test orders.

[0024] Therefore, this application proposes a machine learning-based intelligent test order review method and device, which aims to improve the review efficiency and accuracy, and is specifically used for the review of biochemical test orders.

[0025] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0026] The intelligent inspection form review method based on machine learning provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal communicates with the server through the network. The data storage system can store the data that the server needs to process. The data storage system can be set up separately, integrated on the server, or placed on the cloud or other servers. The terminal can send the inspection order information to be reviewed to the server. After the server receives the inspection order information to be reviewed, the server inputs the inspection order information to be reviewed into at least one pre-trained machine learning model, obtains the review result output by the pre-trained machine learning model, and determines the final review result based on at least one review result. The server can feed back the final review result obtained to the terminal. In addition, in some embodiments, the intelligent review method of the inspection order based on machine learning can also be implemented separately by the server or the terminal. For example, the terminal can directly perform an intelligent review of the inspection order based on machine learning on the inspection order information to be reviewed, or the server can obtain the inspection order information to be reviewed from the data storage system and perform an intelligent review of the inspection order based on machine learning.

[0027] Terminals include, but are not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices include smart speakers, smart TVs, smart air conditioners, and smart car devices. Portable wearable devices include smart watches, smart bracelets, and head-mounted devices. Servers can be implemented as standalone servers or server clusters consisting of multiple servers, or even cloud servers.

[0028] In an exemplary embodiment, Figure 2 As shown, a method for intelligent inspection form review based on machine learning is provided. The method is executed by a computer device, specifically a computer device such as a terminal or a server, or a terminal and a server. In the embodiment of the present application, the method is applied to Figure 1 The server in is used as an example for description, including the following steps 101 to 102.

[0029] Step 101, obtain the test order information to be reviewed; the test order information includes the patient's name, gender, age and several biochemical test items, as well as the department where the test was sent, etc.

[0030] Step 102: Input the inspection order information to be reviewed into at least one pre-trained machine learning model, obtain the review result output by the pre-trained machine learning model, and determine the final review result based on at least one review result.

[0031] Implement steps 101 and 102 above to obtain the inspection order information to be reviewed; input the inspection order information to be reviewed into at least one pre-trained machine learning model to obtain the review results output by the pre-trained machine learning model, and determine the final review result based on the at least one review result. In this application, the application of machine learning algorithms to intelligently review the inspection order information improves the efficiency and accuracy of the review.

[0032] In another exemplary embodiment of the present application, in step 102, inputting the inspection order information to be reviewed into at least one pre-trained machine learning model specifically includes:

[0033] (1) Obtaining a historical audit data set; the historical data set includes a number of historical inspection order information and corresponding historical audit results; when the historical audit result is failure, the historical audit result also includes the reason for failure.

[0034] As an example, a total of 6,382 test reports were collected, covering the patient's name, gender, age, and several biochemical test items. In routine practice, the number of passed reports far exceeds the number of failed reports. To ensure data diversity and representativeness, the test reports were collected in three batches: the first batch randomly selected 2,939 test reports, the second batch selected 1,968 test reports reviewed during daily work and annotated, and the third batch attempted to annotate 1,475 reports using extreme value selection.

[0035] The collected inspection forms were independently and manually labeled by multiple experienced staff members, with each labeling them as either pass (negative) or fail (positive), along with the reasons for failure. There was no time limit. When staff disagreed, a majority-decisions strategy was adopted as the final true value for modeling. This strategy effectively reduced human error and improved the reliability of the labeled data.

[0036] Differences in staff experience can have a certain impact on report annotation, potentially introducing human error. To improve model performance, further optimization of the data annotation process is needed during model training. For example, using senior professionals with at least five years of experience to perform annotations, and providing comprehensive and detailed reasons for failed annotations, can improve annotation consistency.

[0037] (2) Based on the historical audit dataset, the SMOTE algorithm is used for oversampling to construct a training dataset.

[0038] The collected test report dataset requires preprocessing before applying the machine learning model. First, missing values are handled by imputing the median results of the corresponding test items. Ideally, each test report is trained on 38 test items. For test reports that do not include the full set of test items, additional items can be imputed, with the corresponding values imputed using the median results of those items. Secondly, oversampling is performed using the Synthetic Minority Over-sampling Technique (SMOTE) algorithm to balance the ratio of pass and fail reports.

[0039] (3) The initial machine learning model is trained using the training data set. When the model loss converges or reaches the maximum number of iterations, the pre-trained machine learning model is obtained.

[0040] As an example, a training dataset was fed into three machine learning algorithms for model building: logistic regression, random forest, and Xgboost. After training, the models were applied to the validation test set reports, and the audit predictions on the test set were compared with the manually labeled results. These three models were used in parallel, and the final pass / fail decision was made based on the results of the three models, using the majority rule. Model performance was evaluated using metrics such as precision, recall, F1 score, area under the receiver operating characteristic curve (AUC), and mean average precision on the test set. All programs were written in Python 3.0. The random forest and logistic regression models were implemented using Scikit-learn 1.3.0, and the Xgboost model was implemented using Xgboost 1.7.6.

[0041] (4) Inputting the inspection order information to be reviewed into at least one pre-trained machine learning model to obtain the review result output by the pre-trained machine learning model.

[0042] In another exemplary embodiment of the present application, in step (2), the SMOTE algorithm is a method for solving the data imbalance problem by generating synthetic samples. The main steps are as follows:

[0043] Step 1): Identify the minority class. Goal: Identify the minority class in the dataset (i.e., the class that needs to be oversampled). Procedure: Go through all the samples, count the number of samples in each category, and select the class with the fewest samples as the minority class.

[0044] Step 2): Calculate the k nearest neighbors of the minority class samples. Goal: For each minority class sample, find its k nearest neighbors in the feature space (usually using Euclidean distance). Operation: For each minority class sample x i, calculate its distance to all other minority class samples and select the k samples closest to it as its nearest neighbors (usually k=5).

[0045] Step 3): Randomly select a neighboring sample. Goal: Randomly select a neighboring sample for each minority class sample to generate a synthetic sample. Operation: For each minority class sample x i , randomly select a sample x from its k nearest neighbors j .

[0046] Step 4): Generate synthetic samples. Goal: In the current minority class sample x i and its selected neighbor sample x j Generate synthetic samples between. Operation: Calculate sample x i and x j Difference in each feature: diff = x j -x i Generate a random number α∈[0,1]. Each eigenvalue of the synthetic sample is: new =x i +α·diff.

[0047] Repeat the above process multiple times (the specific number is determined by the oversampling rate) until the number of minority class samples reaches the expected number.

[0048] Step 5): Merge the original data with the synthesized samples. Goal: Add the generated synthesized samples to the original dataset to form a balanced new dataset. Action: Merge all original samples (including both majority and minority classes) with the newly generated synthesized samples. This significantly increases the number of minority class samples in the final dataset, alleviating class imbalance.

[0049] Therefore, in step (2), based on the historical audit dataset, the SMOTE algorithm is used for oversampling to construct a training dataset, which specifically includes:

[0050] (2-1) Determine the minority class samples in the historical audit data set; the minority class samples refer to samples whose audit result categories are smaller than the preset number of samples.

[0051] (2-2) For each minority class sample, calculate the distance between the minority class sample and all other minority class samples of the same audit result category.

[0052] (2-3) Select the k samples closest to each other as neighbor samples.

[0053] (2-4) Generate synthetic samples corresponding to the current minority class samples based on the neighboring samples.

[0054] Specifically, the difference between the current minority class sample and the neighboring sample on each feature is calculated; a random number is generated; and the corresponding synthetic sample is generated based on the difference, the random number and the current minority class sample.

[0055] (2-5) Return to step “Determine the minority class samples in the historical audit data set” until the number of minority class samples reaches the preset number of samples.

[0056] (2-6) All synthetic samples are combined with the historical audit dataset to generate a training dataset.

[0057] In another exemplary embodiment of the present application, after executing step (3) of "training the initial machine learning model using the training data set, and obtaining the pre-trained machine learning model when the model loss converges or reaches the maximum number of iterations", the machine learning-based intelligent inspection form review method further includes:

[0058] The cross-validation method is used to evaluate the model performance of the pre-trained machine learning model.

[0059] To evaluate model performance, we used cross-validation. The initial sample was split into five subsamples, one of which was retained as validation data, while the other four were used for training. Cross-validation was repeated five times, with each subsample validated once. The average of the results was then used as the model performance metric.

[0060] In another exemplary embodiment of the present application, in order to evaluate the effectiveness of the intelligent audit method of the present application, the present application conducted a double-blind test, in which a total of 100,048 reports were double-blind tested by laboratory personnel and the AI audit system (the intelligent audit method of the present application), and the same report was manually reviewed and independently evaluated by the AI audit system. The results of the model judgment were based on the minority obeys majority rule, and the results manually marked by the laboratory staff were regarded as the gold standard. Finally, the pass rate and false negative rate FNR of the AI audit system were calculated. Reports marked as passed by AI but marked as failed by laboratory staff were regarded as false negatives. The effectiveness of the intelligent audit method of the present application was evaluated based on the pass rate and false negative rate FNR. In addition, the present application counted the TAT (turnaround time) of a total of 27,473 reports that were reviewed simultaneously by laboratory personnel and the AI audit system. Comparing the TAT of the two can effectively highlight the advantages of the intelligent audit method of the present application.

[0061] In another exemplary embodiment of the present application, in step 102, the inspection order information to be reviewed is input into at least one pre-trained machine learning model, a review result output by the pre-trained machine learning model is obtained, and a final review result is determined based on the at least one review result, specifically including:

[0062] (1) The inspection order information to be reviewed is input into the logistic regression model, random forest model and Xgboost model to obtain the first review result, second review result and third review result respectively.

[0063] (2) Based on the results of the first, second and third audits, the final audit result shall be determined in accordance with the rule that the minority shall obey the majority.

[0064] The audit model established in this application will be embedded in the clinical laboratory medical computer system, allowing the model to audit the test orders. Those that pass the audit can be directly released to patients, while those that fail the audit will be manually reviewed and then released to patients.

[0065] In this application, a retrospective cohort analysis of 27,473 reports was conducted to compare the turnaround time of manual review and artificial intelligence review (turnaround time refers to the total time from the receipt of the sample to the completion of the test order review and release by the clinical laboratory. It is a key indicator for measuring laboratory efficiency and service quality, and directly affects the timeliness of patient diagnosis and treatment). Compared with the median turnaround time of 89 minutes for manual review, the median turnaround time for artificial intelligence review was 68 minutes, a reduction of 23.6%. The 90th percentile turnaround time was shortened from 154 minutes to 132 minutes, a reduction of 14.3%, which reflects the improved efficiency in processing complex reports.

[0066] At the same time, the number of reports completed by manual review and AI review was recorded every 30 minutes. The results showed that AI review was able to issue reports faster than manual review, and this difference was particularly significant within 150 minutes. For example, at 60 minutes, the number of reports reviewed by AI increased from 10.47% for manual review to 37.07%. This means that patients can receive test results faster, reducing diagnostic delays, especially in emergency departments.

[0067] The reports reviewed in this review were reviewed by at least three staff members during the same time period, a common practice in most laboratories. Therefore, AI-powered review is not only efficient in terms of time, but also saves manpower by reducing the workload of repetitive data verification, freeing up manpower for reviewing complex reports.

[0068] In summary, this application uses machine learning technology to build an artificial intelligence audit system for clinical biochemistry laboratories. This audit method performs well in terms of area under the receiver operating curve, precision, recall, F1 score, and average precision. It can effectively identify abnormal test results and significantly improve laboratory efficiency, providing a new solution for intelligent auditing of clinical biochemistry laboratories.

[0069] Regarding the audit results of some special patients, for example, the activity of creatine kinase isoenzyme CK-MB is theoretically less than that of total creatine kinase CK, but in clinical practice, a certain proportion will have the opposite situation. At this time, manual review can accurately judge the audit results of the report based on the patient's clinical diagnosis, literature study, and multi-center and multi-method comparative analysis, and can ensure that the results truly reflect the patient's pathological state. This application uses a machine learning model for intelligent review. The machine learning model is trained under the training data set of manual review, so the machine learning model can more accurately identify the patient's special circumstances. Automatic verification can only judge the audit results based on pre-established rules and cannot accurately identify the patient's special circumstances, while machine learning can be more flexible. In addition, in terms of adaptability, new detection technologies and detection indicators are constantly emerging, and the automatic verification system cannot be updated in time. The machine learning model can use new detection indicator data to train the model. Compared with the system update of the automatic verification system, the training is faster, so the update is more timely.

[0070] The present application also provides an application scenario, which applies the above-mentioned intelligent review method for test orders based on machine learning. Specifically: the intelligent review method for test orders based on machine learning provided in this embodiment can be applied in the biochemical test order review scenario. The scenario includes a data collection link and an intelligent review link; the data collection link is used to collect the patient's biochemical test order information to be reviewed; the intelligent review link is used to apply a machine learning algorithm to perform an intelligent review based on the collected biochemical test order information. The intelligent review method for test orders based on machine learning provided in this embodiment belongs to the intelligent review link.

[0071] Based on the same inventive concept, the embodiments of the present application also provide a machine learning-based intelligent inspection form audit device for implementing the machine learning-based intelligent inspection form audit method mentioned above. The implementation solution provided by this device is similar to the implementation solution described in the above method. Therefore, the specific limitations of one or more machine learning-based intelligent inspection form audit device embodiments provided below can be found in the limitations of the machine learning-based intelligent inspection form audit method above, and will not be repeated here.

[0072] In an exemplary embodiment, Figure 3 As shown, a machine learning-based intelligent inspection form review device is provided, including:

[0073] The test information acquisition module M1 is used to obtain the test order information to be reviewed; the test order information includes the patient's name, gender, age and several biochemical test items.

[0074] The intelligent audit module M2 is used to input the inspection order information to be audited into at least one pre-trained machine learning model, obtain the audit result output by the pre-trained machine learning model, and determine the final audit result based on at least one audit result.

[0075] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 4 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store intelligent audit data of inspection orders based on machine learning. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for intelligent auditing of inspection orders based on machine learning is implemented.

[0076] Those skilled in the art will understand that Figure 4 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application and does not constitute a limitation on the computer device to which the solution of the present application is applied. A specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of the above-mentioned method embodiments when executing the computer program.

[0077] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0078] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0079] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0080] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0081] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0082] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0083] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A machine learning-based intelligent inspection form review method, characterized in that: The machine learning-based intelligent inspection form review method includes: Obtain the test order information to be reviewed; the test order information includes the patient's name, gender, age and several biochemical test items; The inspection order information to be reviewed is input into at least one pre-trained machine learning model to obtain the review result output by the pre-trained machine learning model, and the final review result is determined based on the at least one review result.

2. The intelligent inspection form review method based on machine learning according to claim 1, characterized in that: Input the inspection order information to be reviewed into at least one pre-trained machine learning model, including: Obtain a historical audit data set; the historical data set includes a number of historical inspection order information and corresponding historical audit results; when the historical audit result is failure, the historical audit result also includes the reason for failure; Based on the historical audit data set, the SMOTE algorithm is used for oversampling to construct a training data set; The initial machine learning model is trained using the training data set. When the model loss converges or the maximum number of iterations is reached, a pre-trained machine learning model is obtained. The inspection order information to be reviewed is input into at least one pre-trained machine learning model to obtain the review result output by the pre-trained machine learning model.

3. The intelligent inspection form review method based on machine learning according to claim 2, characterized in that: Based on the historical audit dataset, the SMOTE algorithm is used for oversampling to construct a training dataset, which includes: Determine minority samples in the historical audit data set; the minority samples refer to samples whose audit result categories are smaller than the preset number of samples; For each minority class sample, calculate the distance between the minority class sample and all other minority class samples of the same audit result category; Select the k samples closest to each other as neighbor samples; Generate synthetic samples corresponding to the current minority class samples based on the neighboring samples; Return to step "Determine the minority class samples in the historical audit dataset" until the number of minority class samples reaches the preset number of samples; All synthetic samples are merged with the historical audit dataset to generate the training dataset.

4. The intelligent inspection form review method based on machine learning according to claim 3, characterized in that: Generate synthetic samples corresponding to the current minority class samples based on neighboring samples, specifically including: Calculate the difference between the current minority class sample and the neighboring sample on each feature; Generate a random number; Generate corresponding synthetic samples based on the difference, random number and current minority class samples.

5. The intelligent inspection form review method based on machine learning according to claim 2, characterized in that: After executing the step of "training the initial machine learning model using the training data set, and obtaining a pre-trained machine learning model when the model loss converges or reaches the maximum number of iterations", the machine learning-based intelligent inspection order review method further includes: The cross-validation method is used to evaluate the model performance of the pre-trained machine learning model.

6. The intelligent inspection form review method based on machine learning according to claim 1, characterized in that: Input the inspection order information to be reviewed into at least one pre-trained machine learning model, obtain the review result output by the pre-trained machine learning model, and determine the final review result based on the at least one review result, specifically including: Input the inspection order information to be reviewed into the logistic regression model, random forest model, and Xgboost model to obtain the first review result, second review result, and third review result respectively; Based on the results of the first, second and third reviews, the final review results will be determined in accordance with the rule of minority obeys majority.

7. An intelligent inspection form review device based on machine learning, characterized in that: The machine learning-based intelligent inspection form review device includes: The test information acquisition module is used to obtain the test order information to be reviewed; the test order information includes the patient's name, gender, age and several biochemical test items; The intelligent audit module is used to input the inspection order information to be audited into at least one pre-trained machine learning model, obtain the audit result output by the pre-trained machine learning model, and determine the final audit result based on at least one audit result.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the intelligent inspection form review method based on machine learning as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for intelligently reviewing inspection orders based on machine learning as described in any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for intelligently reviewing inspection orders based on machine learning as described in any one of claims 1 to 6 is implemented.

Images