A critical value extraction method, system, device and medium for diagnostic reports

Abstract

The application discloses a kind of critical value extraction method, system, equipment and medium for diagnostic report, is specifically related to critical value identification technical field, technical points are: image performance feature set is extracted from image performance text in report content text;Report content text and image performance feature set are respectively input into critical value extraction model, search whether there is critical value keyword or critical value synonym keyword synonym text in report content text, and judge whether there is negative word before and after critical value keyword or synonym text;Image performance feature is matched in image performance text set, and it is judged whether there is negative word before and after image performance feature;When there is critical value keyword or synonym text, and there is no negative word before and after critical value keyword or synonym text;Or when there is corresponding image performance feature, and there is no negative word before and after image performance feature, corresponding critical value is output.

Description

Technical Field

[0001] This invention relates to the field of critical value identification technology, specifically to a method, system, device, and medium for extracting critical values ​​for diagnostic reports. Background Technology

[0002] Reporting critical values ​​during imaging diagnosis is a mandatory requirement of the National Health Commission for diagnostic quality. Timely reporting of critical values ​​to clinicians is extremely important for accelerating the diagnosis and treatment of acute and critical illnesses, reducing medical delays, and even medical errors.

[0003] Currently, the method for reporting critical values ​​during the diagnostic process involves the diagnosing physician manually clicking the critical value reporting button and manually selecting the type of critical value discovered by imaging examinations. This manual selection method has several drawbacks. First, some physicians may not remember the types of critical values ​​and therefore neglect to report them proactively; they may also forget to report them due to various distractions while writing the report; secondly, manually clicking to report also reduces work efficiency.

[0004] In the era of NLP text analysis, there have been many attempts to automatically report critical values ​​by analyzing text. The advantage of this method is that it avoids omissions without reducing reporting efficiency; however, its disadvantages are that NLP analysis requires defining many synonyms for judgment. If the doctor's description is highly personalized and these synonyms cannot be accurately matched, the scenario in which the critical value occurs still cannot be identified. Furthermore, it cannot perform matching analysis on imaging manifestations, and cannot help doctors identify critical value risks when doctors have not written keywords.

[0005] Therefore, the present invention aims to provide a method, system, device and medium for extracting critical values ​​for diagnostic reports, in order to solve the aforementioned problems. Summary of the Invention

[0006] The technical problem this invention aims to solve is that existing technologies cannot perform matching analysis on image manifestations, and cannot help doctors identify critical value risks even when doctors have not written keywords. The purpose is to provide a method, system, device, and medium for extracting critical values ​​from diagnostic reports. By constructing a critical value keyword synonym mapping dataset and a critical value image manifestation feature dataset, a large-scale LLM model is trained. This enables the trained critical value extraction model to search for synonyms of critical value keywords and synonyms of critical value keywords within the report text, thus preventing inaccurate critical value extraction due to doctors' personalized descriptions. Simultaneously, the trained critical value extraction model can extract the image manifestation features corresponding to critical values ​​from the image manifestation feature set, thereby enabling matching analysis of image manifestation features to help doctors identify critical value risks even when doctors have not written keywords.

[0007] This invention is achieved through the following technical solution:

[0008] A method for extracting critical values ​​for diagnostic reports, the method comprising:

[0009] A critical value keyword synonym mapping dataset and a critical value image performance feature dataset were constructed. The LLM large model was trained using the constructed critical value keyword synonym mapping dataset and critical value image performance feature dataset to obtain the critical value extraction model.

[0010] The report content text of the target patient is obtained, and an image performance feature set is extracted from the image performance text within the report content text; wherein, the image performance feature set includes at least one image performance feature;

[0011] The report content text and the image performance feature set are respectively input into the critical value extraction model. The model searches for the existence of critical value keywords or synonyms of critical value keywords in the report content text, and determines whether there are negative words before or after the critical value keywords or synonyms. In the image performance text set, the corresponding image performance features are matched according to the pre-built matching rules, and it is determined whether there are negative words before or after the image performance features.

[0012] When there is a critical value keyword or a synonym of a critical value keyword, and there are no negative words before or after the critical value keyword or the synonym, the corresponding critical value will be output.

[0013] When a corresponding image performance feature exists, and there are no negative words before or after the image performance feature, the corresponding critical value will be output.

[0014] Furthermore, obtain the report content text, specifically:

[0015] During the report content text generation process, the current report content text is obtained in real time at preset intervals; or the report content text is obtained after it has been generated.

[0016] Furthermore, after determining that a critical value exists, the method also includes: using the critical value to generate a patient critical value report.

[0017] Furthermore, the pre-constructed critical value keyword synonym mapping dataset includes mapping relationships between multiple critical value keywords and their corresponding synonym texts. The critical value keywords include acute pulmonary embolism, acute aortic dissection (DeBakey I / II), acute aortic aneurysm rupture, cardiac tamponade, massive fluid / blood / pneumothorax, tracheal / bronchial foreign body, and acute cerebral infarction. The synonym texts for acute pulmonary embolism include PE, pulmonary embolus, and saddleembolus; the synonym texts for acute aortic dissection (DeBakey I / II) include dissection sign and true / false lumen; the synonym texts for acute aortic aneurysm rupture include hematoma extravasation and aortic rupture sign; the synonym texts for cardiac tamponade include massive pericardial effusion and cardiac compression; the synonym texts for massive fluid / blood / pneumothorax include lung collapse greater than 2 / 3 or mediastinal shift; the synonym texts for tracheal / bronchial foreign body include high-density / gas shadow obstructing the airway; and the synonym texts for acute cerebral infarction include DWI high signal less than 24 hours.

[0018] Furthermore, the matching rules include matching relationships between multiple image performance features of different quantities, intensities, or severity levels and severity values.

[0019] The present invention also provides a critical value extraction system for diagnostic reports, which is used in any of the above-described methods for critical value extraction in diagnostic reports, the system comprising:

[0020] The model training module is used to construct a critical value keyword synonym mapping dataset and a critical value image performance feature dataset, and to train the LLM large model using the constructed critical value keyword synonym mapping dataset and critical value image performance feature dataset to obtain the critical value extraction model.

[0021] A text search module is used to acquire the report content text of the target patient and extract an image performance feature set from the image performance text within the report content text; wherein, the image performance feature set includes at least one image performance feature; the report content text and the image performance feature set are respectively input into the critical value extraction model, which searches for the existence of critical value keywords or synonyms of critical value keywords in the report content text, and determines whether there are negative words before or after the critical value keywords or synonyms; the corresponding image performance features are matched in the image performance text set according to pre-built matching rules, and it is determined whether there are negative words before or after the image performance features;

[0022] The critical value determination module is used to output the corresponding critical value when there is a critical value keyword or a synonym of a critical value keyword, and there are no negative words before or after the critical value keyword or the synonym. It will also output the corresponding critical value when there is a corresponding image performance feature, and there are no negative words before or after the image performance feature.

[0023] Furthermore, obtain the report content text, specifically:

[0024] During the report content text generation process, the current report content text is obtained in real time at preset intervals; or the report content text is obtained after it has been generated.

[0025] The present invention also provides a computer device, including a system memory and a processor, wherein the system memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0026] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods described above.

[0027] The present invention also provides a computer program product containing instructions that, when executed by a cluster of computer devices, cause the cluster of computer devices to perform the method described in any of the preceding claims.

[0028] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0029] Specifically, in this embodiment, the LLM large model is trained using the constructed critical value keyword synonym mapping dataset and critical value image performance feature dataset. This enables the trained critical value extraction model to search for synonyms of critical value keywords and synonyms of critical value keywords in the report content text, thereby preventing inaccurate critical value extraction due to doctors' personalized descriptions. At the same time, the trained critical value extraction model can extract the image performance features corresponding to critical values ​​from the image performance feature set, thereby enabling matching analysis of image performance features. This helps doctors identify critical value risks even when doctors have not written keywords. Attached Figure Description

[0030] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0031] Figure 1 This is a flowchart illustrating a method for extracting critical values ​​for diagnostic reports in this embodiment;

[0032] Figure 2 This is a schematic diagram of the display interface for obtaining the report content text in this embodiment;

[0033] Figure 3 This is a schematic diagram of the display interface for obtaining image representation text in this embodiment;

[0034] Figure 4 This is a schematic diagram of the critical value determination result interface in this embodiment;

[0035] Figure 5 This is a schematic diagram of the module connections of a critical value extraction system for diagnostic reports in this embodiment;

[0036] Figure 6 This is a schematic diagram of the structure of a computer device in this embodiment. Detailed Implementation

[0037] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0038] In this disclosure, unless otherwise stated, the use of terms such as "first," "second," etc., to describe various elements is not intended to limit the positional, temporal, or importance relationships of these elements; such terms are merely used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of that element, while in other cases, based on the context, they may refer to different instances.

[0039] The terminology used in the description of the various examples in this disclosure is for the purpose of describing particular examples only and is not intended to be limiting. Unless the context explicitly indicates otherwise, an element may be one or more unless the number of elements is specifically limited. Furthermore, the term "and / or" as used in this disclosure covers any one of the listed items and all possible combinations thereof.

[0040] Example 1

[0041] See Figure 1 , Figure 1A flowchart illustrating a method for extracting critical values ​​for diagnostic reports is shown, wherein the method includes:

[0042] S1: Construct a critical value keyword synonym mapping dataset and a critical value image performance feature dataset, and use the constructed critical value keyword synonym mapping dataset and critical value image performance feature dataset to train the LLM large model to obtain the critical value extraction model.

[0043] It should be noted that in this embodiment, a critical value keyword synonym mapping dataset is first constructed, which includes the mapping relationship between multiple critical value keywords and their corresponding synonyms, as shown in the table below:

[0044] 1 Acute pulmonary embolism PE, pulmonary embolism, saddleembolus 2 Acute aortic dissection DeBakey I / II Atrial dissection, true / false cavity 3 Acute aortic aneurysm rupture Hematoma extravasation, aortic rupture signs 4 Cardiac tamponade Massive pericardial effusion and cardiac compression 5 Massive fluid / blood / pneumothorax Lung collapse >2 / 3 or mediastinal shift 6 Foreign body in trachea / bronchus High-density / gas shadows obstruct the airway 7 Acute cerebral infarction DWI high signal <24h 8 Acute cerebral hemorrhage Fresh and high-density substance 9 Acute epidural / subdural hemorrhage Biconvex / crescent-shaped acute hematoma 10 Acute subarachnoid hemorrhage High density in basal cisterns / sulci 11 Brain herniation Cingulate gyrus / cerebellar tonsil displacement, etc. 12 Gastrointestinal perforation Free air under the diaphragm, sickle-shaped air shadow 13 Rupture and bleeding of solid abdominal organs Active bleeding in the liver, spleen, kidneys, and pancreas 14 strangulated intestinal obstruction Closed loop sign, intestinal wall ischemia and necrosis sign 15 ... ...

[0045] Then, a critical value image manifestation feature dataset is constructed. The critical value image manifestation feature dataset includes multiple critical value keywords and the image manifestation features corresponding to the synonyms of critical value keywords. For example, taking the critical value keyword acute pulmonary embolism as an example, its synonyms are PE, pulmonary embolus, and saddleembolus. The corresponding image manifestation features in X-ray images are: pulmonary artery dilation, pulmonary artery segment protrusion, right ventricular enlargement, sparse vascular texture in some areas of the lung field, and wedge-shaped shadow. The corresponding image manifestation features in CT images are: pulmonary artery filling defect, pulmonary artery dilation, wedge-shaped consolidation, and right ventricular enlargement.

[0046] S2: Obtain the report content text of the target patient and extract the image performance feature set from the image performance text within the report content text; wherein, the image performance feature set includes at least one image performance feature; input the report content text and the image performance feature set into the critical value extraction model respectively, search for the existence of critical value keywords or synonyms of critical value keywords in the report content text, and determine whether there are negative words before or after the critical value keywords or synonyms; match the corresponding image performance features in the image performance text set according to the pre-built matching rules, and determine whether there are negative words before or after the image performance features;

[0047] It should be noted that in this embodiment, the report content text includes image performance text and diagnostic content text. The image performance text is the image feature content automatically generated based on the medical image after image acquisition, and the diagnostic content text is the content text written by medical staff based on the medical image. The following are the ways to obtain the report content text, i.e., trigger the extraction of critical values: (1) During the generation of the report content text, the current report content text is obtained in real time at a preset interval. In this way, the critical value result generated based on the currently obtained report content text will overwrite the critical value result generated by the previously obtained report content text, thereby realizing the real-time update of critical values. At the same time, the preset interval is 10s, which can be set to 5s or other intervals in other embodiments. No further restrictions are imposed here. (2) The report content text is obtained after the report content text is generated. See Figure 2 , Figure 2 A schematic diagram of the display interface for retrieving the report content text is shown; see also Figure 3 , Figure 3 A schematic diagram of the display interface for obtaining image representation text is shown.

[0048] Meanwhile, it should also be noted that, due to the need for real-time detection, which takes 5-10 seconds for LLM analysis, the LLM task priority for this task in the background is set to the highest level, the real-time processing level; for reports that are closed immediately after review, the last background detection still exists; at the same time, the general report doctor writes the report content text and obtains it using the above method (1), because the report template is likely to be used when writing the report, and triggering during the writing process will increase the recognition error rate. If method (2) is used, the doctor can view the report content after the report template content is basically completed; the general review doctor uses method (2) during the review of the report, and the review doctor usually modifies the report content, which has a high accuracy rate for real-time detection of the report content.

[0049] Additionally, it should be noted that after inputting the report content text and image performance feature set into the critical value extraction model, prompt words are also set. These prompt words are input according to the actual situation. For example, a prompt word template for matching the corresponding image performance features from the image performance text set is provided. Specifically, for different types of inspection items, if the above-mentioned image performance features appear in the image description, it is considered that there is a risk of the critical value. In other embodiments, other prompt words can also be set, which are not limited here.

[0050] It should be noted that, in this embodiment, synonymous text refers to a common name that has the same meaning as the professional name of the disease, which is determined according to the daily language habits of different regions and different doctors. At the same time, negative words refer to words that express negative meanings, such as words like "not", "not", "none", "no" etc. The "before and after" mentioned above is described according to the conventional word order from left to right, that is, the negative words are on the left or right side of the synonymous text, keywords or features.

[0051] Specifically, in this embodiment, the matching rules include the matching relationships between multiple imaging features of different quantities, different feature intensities, or different degrees of urgency and the urgency value. For example, in X-ray images, acute pulmonary embolism can only be determined when pulmonary artery dilation and pulmonary artery segment protrusion coexist. In CT images, acute pulmonary embolism can be determined when there is a pulmonary artery filling defect. The specific matching relationship depends on the actual situation and is not subject to many restrictions here.

[0052] S3: When there is a critical value keyword or a synonym of a critical value keyword, and there are no negative words before or after the critical value keyword or the synonym, the corresponding critical value will be output; when there is a corresponding image performance feature, and there are no negative words before or after the image performance feature, the corresponding critical value will be output.

[0053] Specifically, in this embodiment, see Figure 4 , Figure 4 The diagram shows the interface for determining critical values. After a critical value is determined, a critical value report is generated using that value. At the same time, the doctor's terminal displays a prompt or reminder, allowing the doctor to modify and confirm the critical value, or reject the critical value determination, or abandon it.

[0054] Specifically, in this embodiment, the LLM large model is trained using the constructed critical value keyword synonym mapping dataset and critical value image performance feature dataset. This enables the trained critical value extraction model to search for synonyms of critical value keywords and synonyms of critical value keywords in the report content text, thereby preventing inaccurate critical value extraction due to doctors' personalized descriptions. At the same time, the trained critical value extraction model can extract the image performance features corresponding to critical values ​​from the image performance feature set, thereby enabling matching analysis of image performance features. This helps doctors identify critical value risks even when doctors have not written keywords.

[0055] Example 2

[0056] See Figure 5The present invention also provides a critical value extraction system for diagnostic reports, which is used in any of the above-described methods for extracting critical values ​​for diagnostic reports, the system comprising:

[0057] The model training module 100 is used to construct a critical value keyword synonym mapping dataset and a critical value image performance feature dataset, and to train the LLM large model using the constructed critical value keyword synonym mapping dataset and critical value image performance feature dataset to obtain a critical value extraction model.

[0058] The text search module 200 is used to acquire the report content text of the target patient and extract an image performance feature set from the image performance text within the report content text; wherein, the image performance feature set includes at least one image performance feature; the report content text and the image performance feature set are respectively input into the critical value extraction model, which searches for the existence of critical value keywords or synonyms of critical value keywords in the report content text, and determines whether there are negative words before or after the critical value keywords or synonyms; the corresponding image performance features are matched in the image performance text set according to pre-built matching rules, and it is determined whether there are negative words before or after the image performance features;

[0059] The critical value determination module 300 is used to output the corresponding critical value when there is a critical value keyword or a synonym of a critical value keyword, and there are no negative words before or after the critical value keyword or the synonym. It will also output the corresponding critical value when there is a corresponding image performance feature, and there are no negative words before or after the image performance feature.

[0060] Furthermore, obtain the report content text, specifically:

[0061] During the report content text generation process, the current report content text is obtained in real time at preset intervals; or the report content text is obtained after it has been generated.

[0062] It should be noted that the modules in the system of Embodiment 2 correspond to the steps in the method of Embodiment 1. The steps in the method of Embodiment 1 have been described in detail in Embodiment 1, and the module content in the system will not be described in detail in this Embodiment 2.

[0063] Example 3

[0064] See Figure 6 This embodiment also provides a computer device, including a system memory 1005 and a processor 1001. The system memory 1005 stores a computer program, and the processor 1001 executes the computer program to implement the steps of any of the methods described above.

[0065] It should be noted that the processor 1001 is used to execute the steps in the above method embodiments according to the instructions in the program code. Alternatively, when the processor 1001 executes the computer program, it implements the functions of each module / unit in the above system / device embodiments.

[0066] Specifically, in this embodiment, the computer program can be divided into one or more modules / units. One or more modules / units are stored in the system memory 1005 and executed by the processor 1001 to complete this application. One or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the terminal device.

[0067] The terminal device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor 1001 and a system memory 1005. Those skilled in the art will understand that this does not constitute a limitation on the terminal device, which may include more or fewer components than shown, or a combination of certain components, or different components. For example, the terminal device may also include an input / output device 1003, a network access device 1002, a bus 1006, etc.

[0068] The processor 1001 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0069] System memory 1005 can be an internal storage unit of the terminal device, such as a hard drive or RAM. System memory 1005 can also be a storage device 1004 of the terminal device, such as an external hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, or FlashCard. Furthermore, system memory 1005 can include both internal storage units and storage device 1004. System memory 1005 is used to store computer programs and other programs and data required by the terminal device. System memory 1005 can also be used to temporarily store data that has been output or will be output.

[0070] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0071] Example 4

[0072] This embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0073] The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), registers, hard disks, optical fibers, compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof, or any other form of computer-readable storage medium in the art.

[0074] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can reside within an application-specific integrated circuit (ASIC). In embodiments of the invention, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device.

[0075] Example 5

[0076] This embodiment also provides a computer program product containing instructions that, when executed by a cluster of computer devices, cause the cluster of computer devices to perform the method described in Embodiment 1.

[0077] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for extracting critical values ​​in diagnostic reports, characterized in that, The methods include: A critical value keyword synonym mapping dataset and a critical value image performance feature dataset were constructed. The LLM large model was trained using the constructed critical value keyword synonym mapping dataset and critical value image performance feature dataset to obtain the critical value extraction model. The report content text of the target patient is obtained, and an image performance feature set is extracted from the image performance text within the report content text; wherein, the image performance feature set includes at least one image performance feature; The report content text and the image performance feature set are respectively input into the critical value extraction model. The model searches for the existence of critical value keywords or synonyms of critical value keywords in the report content text, and determines whether there are negative words before or after the critical value keywords or synonyms. In the image performance text set, the corresponding image performance features are matched according to the pre-built matching rules, and it is determined whether there are negative words before or after the image performance features. When there is a critical value keyword or a synonym of a critical value keyword, and there are no negative words before or after the critical value keyword or the synonym, the corresponding critical value will be output. When a corresponding image performance feature exists, and there are no negative words before or after the image performance feature, the corresponding critical value will be output.

2. The method for extracting critical values ​​for diagnostic reports according to claim 1, characterized in that, To obtain the report content text, specifically: During the report content text generation process, the current report content text is obtained in real time at preset intervals; or the report content text is obtained after it has been generated.

3. The method for extracting critical values ​​for diagnostic reports according to claim 1, characterized in that, After determining that a critical value exists, the method also includes: using the critical value to generate a patient critical value report.

4. The method for extracting critical values ​​for diagnostic reports according to claim 1, characterized in that, The pre-constructed critical value keyword synonym mapping dataset includes mapping relationships between multiple critical value keywords and their corresponding synonym texts. The critical value keywords include acute pulmonary embolism, acute aortic dissection (DeBakey I / II), acute aortic aneurysm rupture, cardiac tamponade, massive fluid / blood / pneumothorax, tracheal / bronchial foreign body, and acute cerebral infarction. Synonyms for acute pulmonary embolism include PE, pulmonary embolus, and saddleembolus; synonyms for acute aortic dissection (DeBakey I / II) include dissection sign and true / false lumen; synonyms for acute aortic aneurysm rupture include hematoma extravasation and aortic rupture sign; synonyms for cardiac tamponade include massive pericardial effusion and cardiac compression; synonyms for massive fluid / blood / pneumothorax include lung collapse greater than 2 / 3 or mediastinal shift; synonyms for tracheal / bronchial foreign body include high-density / gas shadow obstructing the airway; and synonyms for acute cerebral infarction include DWI high signal less than 24 hours.

5. The method for extracting critical values ​​for diagnostic reports according to claim 1, characterized in that, The matching rules include the matching relationships between multiple image performance features of different quantities, different feature intensities, or different degrees of urgency and the urgency value.

6. A critical value extraction system for diagnostic reports, characterized in that, This system is used in a critical value extraction method for diagnostic reports according to any one of claims 1-5, the system comprising: The model training module is used to construct a critical value keyword synonym mapping dataset and a critical value image performance feature dataset, and to train the LLM large model using the constructed critical value keyword synonym mapping dataset and critical value image performance feature dataset to obtain the critical value extraction model. A text search module is used to acquire the report content text of the target patient and extract an image performance feature set from the image performance text within the report content text; wherein, the image performance feature set includes at least one image performance feature; the report content text and the image performance feature set are respectively input into the critical value extraction model, which searches for the existence of critical value keywords or synonyms of critical value keywords in the report content text, and determines whether there are negative words before or after the critical value keywords or synonyms; the corresponding image performance features are matched in the image performance text set according to pre-built matching rules, and it is determined whether there are negative words before or after the image performance features; The critical value determination module is used to output the corresponding critical value when there is a critical value keyword or a synonym of a critical value keyword, and there are no negative words before or after the critical value keyword or the synonym. It will also output the corresponding critical value when there is a corresponding image performance feature, and there are no negative words before or after the image performance feature.

7. A critical value extraction system for diagnostic reports according to claim 1, characterized in that, To obtain the report content text, specifically: During the report content text generation process, the current report content text is obtained in real time at preset intervals; or the report content text is obtained after it has been generated.

8. A computer device comprising a system memory and a processor, wherein the system memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to 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, it implements the steps of the method described in any one of claims 1 to 6.

10. A computer program product containing instructions, characterized in that, When the instructions are executed by a cluster of computer devices, the cluster of computer devices causes the cluster of computer devices to perform the method as described in any one of claims 1 to 6.

Images