A method, apparatus, and computer device for detecting errors in medical reports.

By training an entity recognition model to automatically detect anomalies in medical reports, and combining this with doctor review, the problem of low efficiency and accuracy in existing technologies has been solved, achieving efficient and accurate error correction.

CN123120176APending Publication Date: 2026-11-17SHANGHAI TENTH PEOPLES HOSPITAL
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410473570.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-04-19
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing methods for correcting errors in medical reports are inefficient and inaccurate, making it difficult to handle complex and ever-changing medical data. Furthermore, relying on manual review is inefficient and prone to misjudgment or omission.

Method used

The entity recognition model is trained using unsupervised and supervised learning methods. By deconstructing medical reports into description and conclusion sections, key entities are extracted and compared with preset standard values ​​to automatically detect abnormal entities and output abnormal or error correction reports. The model parameters are adjusted in conjunction with doctor review.

Benefits of technology

It improved the accuracy and processing efficiency of medical reports, reduced human intervention, lowered the error rate, and ensured consistency and accuracy of report content.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123120176A_ABST
    Figure CN123120176A_ABST
Patent Text Reader

Abstract

This invention relates to a method, apparatus, and computer device for detecting errors in medical reports. The method includes: acquiring a medical report to be tested; deconstructing the medical report into a description section and a conclusion section; extracting key entities from the description section; comparing the entity values ​​of the key entities with preset standard values ​​to detect whether the key entities are abnormal; if the key entity is detected as an abnormal entity, outputting an error report; determining whether the conclusion section needs error correction based on the detection results of the key entities; and if the conclusion section needs error correction, outputting an error correction report. This application solves the problems of low efficiency and accuracy in related medical report error correction methods, thereby improving the accuracy of medical reports.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, and computer equipment for detecting errors in medical reports. Background Technology

[0002] In the field of medical data analysis, particularly in the analysis of ultrasound reports, traditional assessment methods rely heavily on physicians' personal experience and clinical expertise. This expert-based approach faces several challenges: First, it is highly subjective; even experienced physicians may interpret data differently, leading to discrepancies in diagnoses and affecting the consistency of treatment. Second, with the surge in medical data volume, the efficiency of manually analyzing reports has become increasingly apparent, and the probability of human error also increases in high-pressure, fast-paced work environments. While attempts have been made to apply some basic text processing and rule-matching algorithms to assist diagnosis to address these issues, these rudimentary techniques are often insufficient to handle the complex data patterns and variable clinical situations in medical reports, thus limiting their practical value.

[0003] The error correction process in related technologies for medical reports mainly relies on batch importing reports and using rule-matching algorithms to detect potential errors. While this approach is relatively efficient for simple data scenarios, its performance exhibits significant limitations when faced with the unique complexity and diverse variations of data in the medical field. First, rule-matching algorithms require explicit rule settings in their design, meaning they can only identify error types already covered by pre-defined rules. They are often ineffective against newly emerging or atypical errors. Second, due to the complexity of medical data, the types and manifestations of errors are constantly changing, making it virtually impossible to cover all potential errors using pre-defined rules, leading to the omission of a large number of errors. Third, even if rule-matching algorithms can identify potential errors, their confirmation still relies on manual review. This not only significantly reduces processing efficiency but also introduces subjective human judgment, potentially leading to misjudgments or omissions. Finally, with the ever-increasing volume of medical data, the manual review model is particularly inadequate, being time-consuming, labor-intensive, and inefficient, failing to meet the demands of the rapidly evolving healthcare field. Therefore, while this process may be feasible in some simple scenarios, more advanced technological methods must be explored to improve accuracy and efficiency in the processing of complex and ever-changing medical data.

[0004] Currently, there are issues with the efficiency and accuracy of medical report error correction methods in related technologies, and no effective solution has yet been proposed. Summary of the Invention

[0005] The purpose of this application is to address the shortcomings of existing technologies by providing a method, apparatus, and computer device for detecting errors in medical reports, thereby at least resolving the problems of low efficiency and accuracy in related medical report error correction methods.

[0006] To achieve the above objectives, the technical solution adopted in this application is as follows:

[0007] In a first aspect, embodiments of this application provide a method for detecting errors in medical reports, the method comprising:

[0008] Obtain the medical report to be tested;

[0009] The medical report to be tested is deconstructed into a descriptive section and a conclusion section;

[0010] Extract key entities from the description section;

[0011] The entity value of the key entity is compared with a preset standard value to detect whether the key entity is abnormal;

[0012] If the critical entity is detected as an abnormal entity, an anomaly report will be output.

[0013] Determine whether the conclusion section needs error correction based on the detection results of the key entities;

[0014] If it is determined that the conclusion section requires error correction, an error correction report will be output.

[0015] In some embodiments, extracting key entities from the description portion includes:

[0016] Obtain multiple unlabeled medical reports as a pre-training dataset;

[0017] Using the pre-training dataset, an unsupervised learning method is employed to pre-train the model, resulting in a pre-trained model.

[0018] Obtain labeled medical reports as the retraining dataset;

[0019] Using the retraining dataset, the pre-trained model is trained using a supervised learning method to obtain a trained entity recognition model;

[0020] The description is input into the trained entity recognition model to obtain the key entity.

[0021] In some embodiments, the step of outputting the error correction report further includes:

[0022] Obtain the doctor's review and confirmation information on the error correction report;

[0023] The parameters of the trained entity recognition model are adjusted using the verification and confirmation information.

[0024] In some embodiments, comparing the entity value of the key entity with a preset standard value to detect whether the key entity is abnormal includes:

[0025] Obtain the context of the key entity and identify the entity value of the key entity from the context;

[0026] Detect whether the entity value of the key entity is within a preset standard range;

[0027] If the entity value of the key entity is detected to be within the preset standard range, then the key entity is determined to be a normal entity.

[0028] If the entity value of the critical entity is detected to be outside the preset standard range, then the critical entity is determined to be an abnormal entity.

[0029] Secondly, embodiments of this application provide a medical report error detection device, comprising:

[0030] The first acquisition unit is used to acquire the medical report to be tested.

[0031] A deconstruction unit is used to deconstruct the medical report to be tested into a description section and a conclusion section;

[0032] An extraction unit is used to extract key entities from the description portion;

[0033] The comparison unit is used to compare the entity value of the key entity with a preset standard value to detect whether the key entity is abnormal.

[0034] The first output unit is used to output an anomaly report if the key entity is detected as an abnormal entity.

[0035] The judgment unit is used to determine whether the conclusion part needs to be corrected based on the detection results of the key entity.

[0036] The second output unit is used to output an error correction report if it is determined that the conclusion part needs to be corrected.

[0037] In some embodiments, the extraction unit includes:

[0038] The first acquisition module is used to acquire multiple unlabeled medical reports as a pre-training dataset;

[0039] The first training module is used to pre-train the model using the pre-training dataset and an unsupervised learning method to obtain a pre-trained model.

[0040] The second acquisition module is used to acquire labeled medical reports as a retraining dataset;

[0041] The second training module is used to train the pre-trained model using the retraining dataset and a supervised learning method to obtain a trained entity recognition model.

[0042] The input module is used to input the description part into the trained entity recognition model to obtain the key entity.

[0043] In some embodiments, the extraction unit further includes:

[0044] The second acquisition unit is used to acquire the doctor's review and confirmation information on the error correction report after the error correction report is output;

[0045] The adjustment unit is used to adjust the parameters of the trained entity recognition model using the verification and confirmation information.

[0046] In some embodiments, the comparison unit includes:

[0047] The third acquisition module is used to acquire the context of the key entity and identify the entity value of the key entity from the context;

[0048] The detection module is used to detect whether the entity value of the key entity is within a preset standard range;

[0049] The first determining module is used to determine that the key entity is a normal entity if the entity value of the key entity is detected to be within the preset standard range.

[0050] The second determining module is used to determine that the key entity is an abnormal entity if the entity value of the key entity is detected to be outside the preset standard range.

[0051] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the medical report error detection method as described in the first aspect above.

[0052] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the medical report error detection method as described in the first aspect above.

[0053] This application adopts the above technical solution. Compared with the prior art, the embodiments of this application provide a method for detecting errors in medical reports. The method includes: acquiring a medical report to be tested; deconstructing the medical report to be tested into a description part and a conclusion part; extracting key entities from the description part; comparing the entity value of the key entity with a preset standard value to detect whether the key entity is abnormal; if the key entity is detected as an abnormal entity, an abnormal report is output; determining whether the conclusion part needs to be corrected based on the detection result of the key entity; if it is determined that the conclusion part needs to be corrected, an error correction report is output. This method solves the problems of low efficiency and accuracy in related medical report error correction methods, and achieves the effect of improving the accuracy of medical reports.

[0054] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0055] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0056] Figure 1 This is a structural block diagram of a mobile terminal according to an embodiment of this application;

[0057] Figure 2 This is a flowchart of a medical report error detection method according to an embodiment of this application;

[0058] Figure 3 This is a flowchart illustrating a preferred embodiment of a medical report error detection method according to this application.

[0059] Figure 4 This is a schematic diagram of a pre-trained fine-tuning model according to a preferred embodiment of this application;

[0060] Figure 5 This is a schematic diagram illustrating the use of a deep learning model for anomaly analysis according to a preferred embodiment of this application;

[0061] Figure 6 This is a structural block diagram of a medical report error detection device according to an embodiment of this application;

[0062] Figure 7 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of this application. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0064] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0065] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0066] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0067] This embodiment provides a mobile terminal. Figure 1 This is a structural block diagram of a mobile terminal according to an embodiment of this application. For example... Figure 1 As shown, the mobile terminal includes: a radio frequency (RF) circuit 110, a memory 120, an input unit 130, a display unit 140, a sensor 150, an audio circuit 160, a wireless fidelity (WiFi) module 170, a processor 180, and a power supply 190, among other components. Those skilled in the art will understand that... Figure 1 The mobile terminal structure shown does not constitute a limitation on the mobile terminal and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0068] The following is combined Figure 1 A detailed introduction to each component of a mobile terminal:

[0069] RF circuit 110 can be used for receiving and transmitting signals during information transmission or calls. Specifically, it receives downlink information from the base station and processes it with processor 180; additionally, it transmits uplink data to the base station. Typically, RF circuitry includes, but is not limited to, antennas, at least one amplifier, transceiver, coupler, low-noise amplifier (LNA), duplexer, etc. Furthermore, RF circuit 110 can also communicate wirelessly with networks and other devices. The aforementioned wireless communication can use any communication standard or protocol, including but not limited to Global System for Mobile Communications (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), etc.

[0070] The memory 120 can be used to store software programs and modules. The processor 180 executes various functional applications and data processing of the mobile terminal by running the software programs and modules stored in the memory 120. The memory 120 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile terminal (such as audio data, phone book, etc.). In addition, the memory 120 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0071] The input unit 130 can be used to receive input numerical or character information, and to generate key signal inputs related to user settings and function control of the mobile terminal. Specifically, the input unit 130 may include a touch panel 131 and other input devices 132. The touch panel 131, also known as a touch screen, can collect touch operations performed by the user on or near it (such as operations performed by the user using a finger, stylus, or any suitable object or accessory on or near the touch panel 131), and drive the corresponding connection devices according to a pre-set program. Optionally, the touch panel 131 may include two parts: a touch detection device and a touch controller. The touch detection device detects the user's touch position and the signal generated by the touch operation, and transmits the signal to the touch controller; the touch controller receives touch information from the touch detection device, converts it into touch point coordinates, sends it to the processor 180, and can receive and execute commands sent by the processor 180. In addition, the touch panel 131 can be implemented using various types such as resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch panel 131, the input unit 130 may also include other input devices 132. Specifically, other input devices 132 may include, but are not limited to, one or more of the following: physical keyboard, function keys (such as volume control buttons, power buttons, etc.), trackball, mouse, joystick, etc.

[0072] Display unit 140 can be used to display information input by the user or information provided to the user, as well as various menus of the mobile terminal. Display unit 140 may include a display panel 141, optionally configured as a Liquid Crystal Display (LCD), Organic Light-Emitting Diode (OLED), or similar display panel 141. Further, touch panel 131 may cover display panel 141. When touch panel 131 detects a touch operation on or near it, it transmits the information to processor 180 to determine the type of touch event. Subsequently, processor 180 provides corresponding visual output on display panel 141 based on the type of touch event. Although in Figure 1 In this embodiment, the touch panel 131 and the display panel 141 are two separate components to realize the input and output functions of the mobile terminal. However, in some embodiments, the touch panel 131 and the display panel 141 can be integrated to realize the input and output functions of the mobile terminal.

[0073] The mobile terminal may also include at least one sensor 150, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor. The ambient light sensor can adjust the brightness of the display panel 141 according to the ambient light level, and the proximity sensor can turn off the display panel 141 and / or backlight when the mobile terminal is moved to the ear. As a type of motion sensor, an accelerometer sensor can detect the magnitude of acceleration in various directions (generally three axes). When stationary, it can detect the magnitude and direction of gravity and can be used for applications that identify the mobile terminal's posture (such as landscape / portrait switching, related games, magnetometer posture calibration), vibration recognition-related functions (such as pedometers, taps), etc. Other sensors that may be configured in the mobile terminal, such as gyroscopes, barometers, hygrometers, thermometers, and infrared sensors, will not be described in detail here.

[0074] The speaker 161 and microphone 162 in the audio circuit 160 provide an audio interface between the user and the mobile terminal. The audio circuit 160 can convert the received audio data into electrical signals and transmit them to the speaker 161, where the speaker 161 converts them into sound signals for output. On the other hand, the microphone 162 converts the collected sound signals into electrical signals, which are received by the audio circuit 160, converted into audio data, and then output to the processor 180 for processing. After processing, the audio data is transmitted via the RF circuit 110 to, for example, another mobile terminal, or the audio data is output to the memory 120 for further processing.

[0075] WiFi is a short-range wireless transmission technology. Mobile terminals using a WiFi module 170 can help users send and receive emails, browse web pages, and access streaming media, providing users with wireless broadband internet access. Although Figure 1 WiFi module 170 is shown, but it is understood that it is not a necessary component of a mobile terminal and can be omitted or replaced with other short-range wireless transmission modules, such as Zigbee modules or WAPI modules, as needed without changing the nature of the invention.

[0076] The processor 180 is the control center of the mobile terminal. It connects various parts of the mobile terminal via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 120, and by calling data stored in the memory 120, it performs various functions and processes data of the mobile terminal, thereby providing overall monitoring of the mobile terminal. Optionally, the processor 180 may include one or more processing units; preferably, the processor 180 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 180.

[0077] The mobile terminal also includes a power supply 190 (such as a battery) that supplies power to various components. Preferably, the power supply can be logically connected to the processor 180 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system.

[0078] Although not shown, mobile terminals may also include cameras, Bluetooth modules, etc., which will not be elaborated here.

[0079] In this embodiment, the processor 180 is configured as follows:

[0080] Obtain the medical report to be tested;

[0081] The medical report to be tested is deconstructed into a descriptive section and a conclusion section;

[0082] Extract key entities from the description section;

[0083] The entity value of the key entity is compared with a preset standard value to detect whether the key entity is abnormal;

[0084] If the critical entity is detected as an abnormal entity, an anomaly report will be output.

[0085] Determine whether the conclusion section needs error correction based on the detection results of the key entities;

[0086] If it is determined that the conclusion section requires error correction, an error correction report will be output.

[0087] In some embodiments, the processor 180 is further configured to:

[0088] Obtain multiple unlabeled medical reports as a pre-training dataset;

[0089] Using the pre-training dataset, an unsupervised learning method is employed to pre-train the model, resulting in a pre-trained model.

[0090] Obtain labeled medical reports as the retraining dataset;

[0091] Using the retraining dataset, the pre-trained model is trained using a supervised learning method to obtain a trained entity recognition model;

[0092] The description is input into the trained entity recognition model to obtain the key entity.

[0093] In some embodiments, the processor 180 is further configured to:

[0094] Obtain the doctor's review and confirmation information on the error correction report;

[0095] The parameters of the trained entity recognition model are adjusted using the verification and confirmation information.

[0096] In some embodiments, the processor 180 is further configured to:

[0097] Obtain the context of the key entity and identify the entity value of the key entity from the context;

[0098] Detect whether the entity value of the key entity is within a preset standard range;

[0099] If the entity value of the key entity is detected to be within the preset standard range, then the key entity is determined to be a normal entity.

[0100] If the entity value of the critical entity is detected to be outside the preset standard range, then the critical entity is determined to be an abnormal entity.

[0101] This embodiment provides a method for detecting errors in medical reports. Figure 2 This is a flowchart of a medical report error detection method according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps:

[0102] Step S201: Obtain the medical report to be tested;

[0103] Step S202: Deconstruct the medical report to be tested into a description section and a conclusion section;

[0104] Step S203: Extract key entities from the description portion;

[0105] Step S204: Compare the entity value of the key entity with a preset standard value to detect whether the key entity is abnormal;

[0106] Step S205: If the key entity is detected as an abnormal entity, an abnormal report is output.

[0107] Step S206: Determine whether the conclusion section needs error correction based on the detection results of the key entity;

[0108] Step S207: If it is determined that the conclusion part needs to be corrected, then an error correction report is output.

[0109] The type of medical report to be tested in this embodiment is not specifically limited, and can be an ultrasound report, blood test report, etc. After obtaining the medical report to be tested, this embodiment structuring it into a description part and a conclusion part to facilitate targeted analysis of each part. For the description part, this application can apply natural language processing to parse, identify, and extract key entities, such as gestational age and humerus length. Then, the entity values ​​of the extracted key entities are compared with preset standard values ​​to detect whether the key entities are abnormal. If abnormal entities are detected, an anomaly report is output, which describes the abnormality of the key entities. For the conclusion part, this embodiment compares the analysis results of the description part with the conclusion part to verify the accuracy of the conclusion part. If it is determined that the conclusion part is incorrect and requires correction, an error correction report is output, which describes the specific content that needs to be corrected.

[0110] In some embodiments, step S203 extracts key entities from the description portion, including:

[0111] Obtain multiple unlabeled medical reports as a pre-training dataset;

[0112] Using the pre-training dataset, an unsupervised learning method is employed to pre-train the model, resulting in a pre-trained model.

[0113] Obtain labeled medical reports as the retraining dataset;

[0114] Using the retraining dataset, the pre-trained model is trained using a supervised learning method to obtain a trained entity recognition model;

[0115] The description is input into the trained entity recognition model to obtain the key entity.

[0116] The key entities described in the embodiments of this application may include, but are not limited to, fetal biparietal diameter, humerus length, and femur length.

[0117] In some embodiments, following the output error correction report, the following is also included:

[0118] Obtain the doctor's review and confirmation information on the error correction report;

[0119] The parameters of the trained entity recognition model are adjusted using the verification and confirmation information.

[0120] In some embodiments, step S204 compares the entity value of the key entity with a preset standard value to detect whether the key entity is abnormal, including:

[0121] Obtain the context of the key entity and identify the entity value of the key entity from the context;

[0122] Detect whether the entity value of the key entity is within a preset standard range;

[0123] If the entity value of the key entity is detected to be within the preset standard range, then the key entity is determined to be a normal entity.

[0124] If the entity value of the critical entity is detected to be outside the preset standard range, then the critical entity is determined to be an abnormal entity.

[0125] Through the above steps, this application solves the problem of low efficiency and accuracy in medical report error correction methods in related technologies, and achieves the effect of improving the accuracy of medical reports.

[0126] The embodiments of this application will be described and illustrated below through preferred embodiments.

[0127] Figure 3 This is a flowchart illustrating a preferred embodiment of a medical report error detection method according to this application, as shown below. Figure 3 As shown, the execution process specifically includes:

[0128] (1) Report input stage:

[0129] This stage involves receiving medical reports for analysis. These reports can be presented in forms or other formats to prepare for further processing.

[0130] Function: As the starting point of the process, it receives medical reports that require analysis and error correction.

[0131] Job duties: The system takes the fetal ultrasound report as input, confirms receipt of the report, and prepares for the next step of processing.

[0132] (2) Report Analysis Phase:

[0133] The error correction model deconstructs the imported medical report, dividing it into a descriptive section and a conclusion section. This separation facilitates the specialized processing of different parts of the report.

[0134] Purpose: To divide the report into two parts, description and conclusion, for specialized analysis.

[0135] Job Responsibilities: Automatically analyze report content, identify the descriptive and concluding sections, and prepare for subsequent steps.

[0136] (3) Entity extraction stage:

[0137] In the description section of the report, natural language processing technology is applied to parse the text in order to identify and extract key entities such as gestational age, humerus length, and growth.

[0138] Function: To identify and extract key medical entities from report descriptions.

[0139] Job Responsibilities: Utilize named entity recognition technology in natural language processing to analyze report descriptions and accurately identify key entities such as gestational age and biological parameters.

[0140] (4) Automatic verification stage:

[0141] The extracted entities will be compared with pre-set medical standards and expert knowledge bases, automatically checking whether each entity is within the normal range and identifying any outliers.

[0142] Function: Compare entity values ​​with standard values ​​to identify any anomalies.

[0143] Job Description: Compare the extracted entity values ​​with preset standards and automatically identify values ​​outside the normal range.

[0144] (5) Exception handling phase:

[0145] When an abnormal entity is detected, the system generates an abnormal report based on a predefined abnormal report template, recording all the abnormal situations found in detail.

[0146] Function: Generate detailed anomaly reports for identified anomalies.

[0147] Job Responsibilities: Based on abnormal situations, create exception reports using predefined templates, and record the abnormal entities and their related information.

[0148] (6) Conclusion Verification Stage:

[0149] The conclusions section of the analysis report is compared with the analysis results in the description section to verify the accuracy of the conclusions and determine whether there are any inconsistencies or errors.

[0150] Purpose: To ensure the accuracy of the report's conclusions and the consistency of the descriptive sections.

[0151] Job Responsibilities: Analyze the conclusions section, compare them with the descriptive analysis results, and verify the accuracy and relevance of the conclusions.

[0152] (7) Error correction decision-making stage:

[0153] Based on the results of the conclusion verification, the system determines whether the report needs correction. This step ensures the reliability and consistency of the report's conclusions.

[0154] Function: To determine whether the report needs to be revised and to indicate the specific content that needs to be corrected.

[0155] Job Responsibilities: Based on the results of the conclusion verification, determine whether the report is accurate, and if there are errors, determine the specific direction and method for correction.

[0156] (8) Error correction report output stage:

[0157] After all necessary steps are completed, the final verification and correction report will be generated. This complete process not only ensures the consistency and accuracy of the report content but also significantly enhances the efficiency of diagnostic work. This application employs advanced automation technology, significantly reducing human intervention, minimizing the possibility of errors in medical reports, and improving the accuracy and efficiency of medical text processing, thus possessing significant practical value.

[0158] Function: Outputs the final report after error correction.

[0159] Job Responsibilities: Based on the error correction decisions, revise the report and output the final verified and corrected medical report.

[0160] The specific implementation of this preferred embodiment is as follows:

[0161] (1) Preliminary data processing stage:

[0162] Standardize the input ultrasound report text, for example, by removing extra spaces and line breaks.

[0163] Mark key information in the text, such as gestational age and fetal measurements (biparietal diameter, head circumference, abdominal circumference, femur length, etc.).

[0164] (2) Key data extraction and verification stage:

[0165] Natural language processing techniques, particularly entity recognition (NER) algorithms, are used to identify medical terms and measurements in the reports.

[0166] Use a pre-set gestational age-fetal development index reference table to verify whether the measurements in the report are within the normal range.

[0167] (3) Anomaly detection and report generation stage:

[0168] For each key metric, a deep learning model is used to analyze the degree of deviation from the normal value.

[0169] If an anomaly is detected, the system will automatically generate a correction report, clearly indicating which indicators are outside the normal range and providing reference thresholds.

[0170] (4) Doctor's review and feedback stage:

[0171] The automatically generated error correction report is submitted to the doctor for final review. The feedback data from the doctor's review will be used to optimize and adjust the model to improve future diagnostic accuracy.

[0172] (5) Report Output Stage:

[0173] After the doctor has reviewed and confirmed that everything is correct, a final report is provided to medical staff and the patient, which may include further clinical advice or instructions for follow-up examinations if necessary.

[0174] This application's embodiments employ unsupervised pre-training and targeted fine-tuning to improve the accuracy and efficiency of key entity recognition; and use deep learning models for anomaly analysis of indicators.

[0175] Specifically, the pre-training fine-tuning algorithm is implemented in the following steps:

[0176] Step 1: Pre-training phase (unsupervised learning)

[0177] (1) Data preparation: 4511 unlabeled ultrasound reports were collected as a pre-training dataset. These data cover a wide range of medical terms and expressions, providing the model with a rich linguistic environment and knowledge. The data started from 4511 unlabeled fetal ultrasound reports. These reports were not labeled in any way, that is, they did not contain any predefined labels for entities or abnormal indicators.

[0178] (2) Text preprocessing: This includes steps such as cleaning the data (e.g., removing irrelevant symbols and correcting typos), word segmentation (breaking the text into meaningful units such as words or phrases), and possible word embedding representations (converting words into vectors).

[0179] (3) Mask Prediction Task: An unsupervised learning method is adopted, using a mask language model (such as BERT). Specifically, 15% of the words or characters are randomly selected for masking (i.e., temporarily replaced with special [MASK] tags), and then the model attempts to predict the original content at these mask locations. This process is unsupervised because it does not rely on labeled data, but rather allows the model to learn the inherent rules of language from the context. In this way, the model can understand and grasp the deep semantic structure of medical texts, laying the foundation for entity recognition.

[0180] Step 2: Fine-tuning phase (supervised learning)

[0181] (1) Data preparation: 200 cases each from early, mid and late pregnancy were selected, totaling 600 fetal ultrasound reports with rich physical information. These reports were carefully selected by expert doctors to ensure that they contained physical features specific to each stage of pregnancy and that the relevant physical features (such as organ names, abnormal indicators, etc.) were manually marked.

[0182] (2) Fine-tuning task: Using the pre-trained model as a foundation, further fine-tuning is performed on these 600 labeled data examples. In this stage, the model learns to recognize and classify specific entities in the report.

[0183] During fine-tuning, additional layers, such as CRF layers, are typically added to the pre-trained model to adapt it to specific downstream tasks of named entity recognition. A schematic diagram of the pre-trained fine-tuning model is shown below. Figure 4 As shown.

[0184] This invention improves the accuracy of entity recognition in medical reports by combining unsupervised pre-training with targeted fine-tuning, while also enhancing the model's efficiency and generalization ability, providing an effective technical means for the automatic processing and analysis of medical reports.

[0185] Specifically, the steps for using deep learning models to perform anomaly analysis on metrics are as follows:

[0186] A diagram illustrating the use of deep learning models for indicator anomaly analysis is shown below. Figure 5 As shown, this application introduces an intelligent error correction system based on the NER (Named Entity Recognition) model. This system first uses the NER model to automatically identify key entities in the ultrasound report, such as the fetal biparietal diameter, humerus length, and femur length. This application analyzes the data surrounding these entities using attribute value extraction technology and matches it with expert standard values ​​to automatically identify abnormal parts in the report and verify the accuracy of the report's conclusions. The process involves data input, entity recognition, attribute value extraction, anomaly detection, and final error correction output. The specific steps are as follows:

[0187] Step 1: Enter the report.

[0188] Input: A fetal ultrasound report. This report contains important information about the fetus's health, which may include, but is not limited to, fetal weight, head circumference, abdominal circumference, and other indicators.

[0189] Step 2: Preprocessing report.

[0190] Preprocessing: The report text is preprocessed, including text cleaning (removing irrelevant characters, correcting punctuation, etc.) and text segmentation, to facilitate subsequent model processing and analysis.

[0191] Step 3: The entity recognition model identifies entities.

[0192] Model: A pre-trained Named Entity Recognition (NER) model was used. This model was pre-trained and fine-tuned on a dataset of fetal ultrasound reports and is able to identify key entities in the reports.

[0193] Operation: The model processes the cleaned report text sentence by sentence or paragraph by paragraph, identifies and labels predefined entity categories, such as "head circumference" and "biparietal diameter".

[0194] Attribute value extraction process: For each entity identified by the NER model, the surrounding context or specific patterns (such as numerical values) are further analyzed to extract attribute values ​​related to that entity. For example, if the entity "head circumference" is identified, the specific numerical value will be extracted from the surrounding text.

[0195] Step 4: Compare with the predetermined standard range.

[0196] Comparison criteria: Based on a predefined normal value range (which may vary depending on the stage of pregnancy), determine whether the extracted attribute values ​​are within the normal range.

[0197] Operation: For each extracted attribute value, compare it with a predefined standard range according to its corresponding entity type. If the attribute value is within the predetermined range, the indicator is considered normal; otherwise, it is marked as abnormal.

[0198] Step 5: Analysis of Comparison Results

[0199] If the attribute values ​​are normal, output that the entity is normal: For the identified entity and its attribute values ​​within the normal range, a confirmation message can be output to indicate that this part of the check is normal.

[0200] If the attribute value is abnormal, output the abnormal entity and the normal value range: For the identified abnormal indicators, output specific error correction information, clearly indicating which entity's attribute value is not within the normal range, so that doctors or examiners can conduct further analysis and processing.

[0201] Step 6: Output the error correction report.

[0202] This application utilizes advanced deep learning algorithms to simulate the diagnostic thinking of a professional physician in processing and understanding the text content of medical image reports. It automatically identifies key medical terms and clinical indicators and analyzes their relevance to the patient's condition. Combined with natural language processing technology, it can extract valuable information from both structured and unstructured text data, thereby providing more accurate diagnostic evidence. Furthermore, the technology employed in this application ensures information privacy and security while meeting medical regulatory requirements, all under the premise of data protection.

[0203] This preferred embodiment can solve the following problems:

[0204] 1. Reduce subjectivity in the diagnostic process: By introducing advanced artificial intelligence technology, the influence of human judgment in the diagnostic process can be reduced, thereby improving the accuracy and consistency of diagnosis.

[0205] 2. Handling complex data relationships: Develop algorithms that can effectively handle complex data relationships and variability in ultrasound reports to improve diagnostic accuracy.

[0206] 3. Automatic error correction: A system that can automatically identify and correct errors in reports to improve diagnostic efficiency.

[0207] This preferred embodiment is based on a medical report error detection method that integrates various cutting-edge technologies, including but not limited to data processing techniques, artificial intelligence and machine learning algorithms, natural language processing, and deep learning. Supported by these technologies, this application can not only be applied to the intelligent analysis of fetal ultrasound reports but also extended to the intelligent error correction of various medical reports, such as X-ray, CT scans, MRI, and PET scans. Through intelligent analysis and real-time error correction, the accuracy and efficiency of diagnosis are effectively improved, while significantly reducing the workload of doctors. By comprehensively applying these technologies to analyze medical imaging reports, the quality of medical services can be improved, and highly efficient and accurate diagnoses can be achieved.

[0208] It should be noted that the steps shown in the above process or in the flowchart of the accompanying figures can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0209] This embodiment provides a medical report error detection device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as described previously. As used below, the terms "module," "unit," "subunit," etc., can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0210] Figure 6 This is a structural block diagram of a medical report error detection device according to an embodiment of this application, such as... Figure 6 As shown, the device includes:

[0211] The first acquisition unit 61 is used to acquire the medical report to be tested;

[0212] Deconstruction unit 62 is used to deconstruct the medical report to be tested into a description part and a conclusion part;

[0213] Extraction unit 63 is used to extract key entities from the description portion;

[0214] The comparison unit 64 is used to compare the entity value of the key entity with a preset standard value to detect whether the key entity is abnormal.

[0215] The first output unit 65 is used to output an anomaly report if the key entity is detected as an abnormal entity.

[0216] The judgment unit 66 is used to determine whether the conclusion part needs to be corrected based on the detection result of the key entity.

[0217] The second output unit 67 is used to output an error correction report if it is determined that the conclusion part needs to be corrected.

[0218] In some embodiments, the extraction unit 63 includes:

[0219] The first acquisition module is used to acquire multiple unlabeled medical reports as a pre-training dataset;

[0220] The first training module is used to pre-train the model using the pre-training dataset and an unsupervised learning method to obtain a pre-trained model.

[0221] The second acquisition module is used to acquire labeled medical reports as a retraining dataset;

[0222] The second training module is used to train the pre-trained model using the retraining dataset and a supervised learning method to obtain a trained entity recognition model.

[0223] The input module is used to input the description part into the trained entity recognition model to obtain the key entity.

[0224] In some embodiments, the extraction unit 63 further includes:

[0225] The second acquisition unit is used to acquire the doctor's review and confirmation information on the error correction report after the error correction report is output;

[0226] The adjustment unit is used to adjust the parameters of the trained entity recognition model using the verification and confirmation information.

[0227] In some embodiments, the comparison unit 64 includes:

[0228] The third acquisition module is used to acquire the context of the key entity and identify the entity value of the key entity from the context;

[0229] The detection module is used to detect whether the entity value of the key entity is within a preset standard range;

[0230] The first determining module is used to determine that the key entity is a normal entity if the entity value of the key entity is detected to be within the preset standard range.

[0231] The second determining module is used to determine that the key entity is an abnormal entity if the entity value of the key entity is detected to be outside the preset standard range.

[0232] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0233] An embodiment provides a computer device. The medical report error detection method described in this application embodiment can be implemented using a computer device. Figure 7 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of this application.

[0234] The computer device may include a processor 71 and a memory 72 storing computer program instructions.

[0235] Specifically, the processor 71 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0236] The memory 72 may include a mass storage device for data or instructions. For example, and not limitingly, the memory 72 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 72 may include removable or non-removable (or fixed) media. Where appropriate, the memory 72 may be internal or external to a data processing device. In a particular embodiment, the memory 72 is non-volatile memory. In a particular embodiment, the memory 72 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), Extended Data Out Dynamic Random-Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0237] The memory 72 can be used to store or cache various data files that need to be processed and / or communicated, as well as possible computer program instructions executed by the processor 71.

[0238] The processor 71 implements any of the medical report error detection methods described in the above embodiments by reading and executing computer program instructions stored in the memory 72.

[0239] In some embodiments, the computer device may further include a communication interface 73 and a bus 70. For example, Figure 7 As shown, the processor 71, memory 72, and communication interface 73 are connected through bus 70 and complete communication with each other.

[0240] The communication interface 73 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of this application. The communication interface 73 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0241] Bus 70 includes hardware, software, or both, that couples components of a computer device together. Bus 70 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 70 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 70 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0242] Furthermore, in conjunction with the medical report error detection methods in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the medical report error detection methods in the above embodiments.

[0243] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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.

[0244] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for detecting errors in medical reports, characterized in that, include: Obtain the medical report to be tested; The medical report to be tested is deconstructed into a descriptive section and a conclusion section; Extract key entities from the description section; The entity value of the key entity is compared with a preset standard value to detect whether the key entity is abnormal; If the critical entity is detected as an abnormal entity, an anomaly report will be output. Determine whether the conclusion section needs error correction based on the detection results of the key entities; If it is determined that the conclusion section requires error correction, an error correction report will be output.

2. The method according to claim 1, characterized in that, Extracting key entities from the description portion includes: Obtain multiple unlabeled medical reports as a pre-training dataset; Using the pre-training dataset, an unsupervised learning method is employed to pre-train the model, resulting in a pre-trained model. Obtain labeled medical reports as the retraining dataset; Using the retraining dataset, the pre-trained model is trained using a supervised learning method to obtain a trained entity recognition model; The description is input into the trained entity recognition model to obtain the key entity.

3. The method according to claim 2, characterized in that, Following the output error correction report, the following is also included: Obtain the doctor's review and confirmation information on the error correction report; The parameters of the trained entity recognition model are adjusted using the verification and confirmation information.

4. The method according to any one of claims 1 to 3, characterized in that, The step of comparing the entity value of the key entity with a preset standard value to detect whether the key entity is abnormal includes: Obtain the context of the key entity and identify the entity value of the key entity from the context; Detect whether the entity value of the key entity is within a preset standard range; If the entity value of the key entity is detected to be within the preset standard range, then the key entity is determined to be a normal entity. If the entity value of the critical entity is detected to be outside the preset standard range, then the critical entity is determined to be an abnormal entity.

5. A medical report error detection device, characterized in that, include: The first acquisition unit is used to acquire the medical report to be tested. A deconstruction unit is used to deconstruct the medical report to be tested into a description section and a conclusion section; An extraction unit is used to extract key entities from the description portion; The comparison unit is used to compare the entity value of the key entity with a preset standard value to detect whether the key entity is abnormal. The first output unit is used to output an anomaly report if the key entity is detected as an abnormal entity. The judgment unit is used to determine whether the conclusion part needs to be corrected based on the detection results of the key entity. The second output unit is used to output an error correction report if it is determined that the conclusion part needs to be corrected.

6. The apparatus according to claim 5, characterized in that, The extraction unit includes: The first acquisition module is used to acquire multiple unlabeled medical reports as a pre-training dataset; The first training module is used to pre-train the model using the pre-training dataset and an unsupervised learning method to obtain a pre-trained model. The second acquisition module is used to acquire labeled medical reports as a retraining dataset; The second training module is used to train the pre-trained model using the retraining dataset and a supervised learning method to obtain a trained entity recognition model. The input module is used to input the description part into the trained entity recognition model to obtain the key entity.

7. The apparatus according to claim 6, characterized in that, Also includes: The second acquisition unit is used to acquire the doctor's review and confirmation information on the error correction report after the error correction report is output; The adjustment unit is used to adjust the parameters of the trained entity recognition model using the verification and confirmation information.

8. The apparatus according to any one of claims 5 to 7, characterized in that, The comparison unit includes: The third acquisition module is used to acquire the context of the key entity and identify the entity value of the key entity from the context; The detection module is used to detect whether the entity value of the key entity is within a preset standard range; The first determining module is used to determine that the key entity is a normal entity if the entity value of the key entity is detected to be within the preset standard range. The second determining module is used to determine that the key entity is an abnormal entity if the entity value of the key entity is detected to be outside the preset standard range.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 4.