Method, device and storage medium for quantifying workload of image report

Abstract

The present application discloses a method, device and storage medium for quantifying the workload of an imaging report, including: obtaining a target imaging report corresponding to a patient, and extracting multiple first key features in the text content of the target imaging report using an NLP model; encoding each first key feature using a pre-set encoding rule, and determining multiple feature codes corresponding to each first key feature and multiple encoding values ​​corresponding to each feature code; generating a corresponding first data set based on multiple feature codes and multiple encoding values; and inputting the first data set into a pre-set workload prediction model, and determining the workload corresponding to the target imaging report. The technical effect of truly reflecting the workload of doctors when filling out diagnostic reports based on target imaging reports and enhancing the work enthusiasm of doctors is achieved.

Description

Technical Field

[0001] The present application relates to the field of medical diagnosis technology, and in particular to a method, device and storage medium for quantifying the diagnostic workload of an imaging report. Background Art

[0002] As medical imaging becomes an indispensable part of modern medical services, the workload of doctors filling out diagnostic reports based on imaging reports is also increasing. How to correctly and effectively evaluate the workload of doctors is related to the important issue of how to improve the enthusiasm of doctors. The traditional way to quantify the workload of doctors is to assign a score to each examination item. However, the time spent on the process of doctors filling out diagnostic reports based on imaging reports is not the same. That is, for more complex imaging reports, doctors spend more time; for simpler imaging reports, doctors spend less time. Therefore, if the traditional quantification method is used to quantify the workload of doctors filling out diagnostic reports based on imaging reports, it cannot truly reflect the workload of doctors.

[0003] For example, for patients with more complicated conditions, the imaging report contains more content describing their condition, and the doctor needs to spend a longer time filling out the diagnosis report based on the imaging report. Therefore, it would be biased to use traditional quantitative methods to evaluate the doctor's workload.

[0004] The publication number is CN117689242A, and the name is verification service work quantification method, device, electronic device and storage medium. The method includes: obtaining data sets such as current basic data, current accounting boundaries and monitoring data of the object to be verified, and the object to be verified includes at least one accounting subject; based on the pre-established mapping relationship between basic data, accounting boundaries and workload, determining the workload corresponding to the current basic data and current accounting boundaries of each accounting subject as the first workload of each accounting subject; based on the monitoring data of each accounting subject, through a preset accounting algorithm, determining the second workload of each accounting subject; according to the first workload and the second workload of each accounting subject, obtaining a quantified result.

[0005] The publication number is CN116227972A, and the name is a quantitative evaluation method for science and technology commissioners based on a cloud platform. The method includes the following steps: S1. Constructing an evaluation index system for science and technology commissioners, and based on the data of the cloud platform, standardizing the evaluation factors in the evaluation index system to obtain the standardized score values ​​of the evaluation factors; S2. Determining the weight coefficients of the evaluation factors through the hierarchical analysis method and the entropy weight method, and establishing a quantitative evaluation model for the work of science and technology commissioners.

[0006] With regard to the technical problem in the above-mentioned prior art that the traditional workload quantification method of filling out diagnostic reports based on imaging reports cannot truly reflect the workload of doctors, and thus cannot further enhance the work enthusiasm of doctors, no effective solution has been proposed yet. Summary of the invention

[0007] The embodiments of the present disclosure provide a method, device and storage medium for quantifying the workload of imaging reports, so as to at least solve the technical problem in the prior art that the traditional workload quantification method of filling out diagnostic reports based on imaging reports cannot truly reflect the workload of doctors, and thus cannot further enhance the work enthusiasm of doctors.

[0008] According to one aspect of an embodiment of the present disclosure, a method for quantifying the workload of an imaging report is provided, comprising: obtaining a target imaging report corresponding to a patient, and extracting a plurality of first key features in the text content of the target imaging report using an NLP model, wherein the plurality of first key features include examination items, the number of included parts, the number of positive parts, the number of lesions, the number of lesion measurements, the positive grade, whether it is an emergency, whether it is a trauma, and whether there is a historical examination; encoding each first key feature respectively using a preset coding rule, and determining a plurality of feature codes corresponding to each first key feature and a plurality of coding values ​​corresponding to each feature code, wherein the coding rule includes a key feature code and a coding value; generating a corresponding first data set based on the plurality of feature codes and the plurality of coding values; and inputting the first data set into a preset workload prediction model, and determining the workload corresponding to the target imaging report.

[0009] According to another aspect of an embodiment of the present disclosure, a storage medium is further provided, the storage medium including a stored program, wherein when the program is running, a processor executes any one of the methods described above.

[0010] According to another aspect of the embodiments of the present disclosure, there is also provided a workload quantification device for an imaging report, comprising: a first feature extraction module, for obtaining a target imaging report corresponding to a patient, and using an NLP model to extract a plurality of first key features in the text content of the target imaging report, wherein the plurality of first key features include examination items, the number of included parts, the number of positive parts, the number of lesions, the number of lesion measurements, the positive grade, whether it is an emergency, whether it is a trauma, and whether there is a historical examination; an encoding module, for encoding each first key feature respectively using a preset encoding rule, and determining a plurality of feature codes corresponding to each first key feature and a plurality of encoding values ​​corresponding to each feature code, wherein the encoding rule includes a key feature code and a encoding value; a first data set generation module, for generating a corresponding first data set based on the plurality of feature codes and the plurality of encoding values; and a workload determination module, for inputting the first data set into a preset workload prediction model, and determining the workload corresponding to the target imaging report.

[0011] According to another aspect of the embodiments of the present disclosure, there is also provided a workload quantification device for an imaging report, comprising: a processor; and a memory connected to the processor, for providing the processor with instructions for processing the following processing steps: obtaining a target imaging report corresponding to a patient, and extracting a plurality of first key features in the text content of the target imaging report using an NLP model, wherein the plurality of first key features include examination items, the number of included parts, the number of positive parts, the number of lesions, the number of lesion measurements, the positive grade, whether it is an emergency, whether it is a trauma, and whether there is a historical examination; encoding each first key feature respectively using a preset coding rule, and determining a plurality of feature codes corresponding to each first key feature and a plurality of coding values ​​corresponding to each feature code, wherein the coding rule includes a key feature code and a coding value; generating a corresponding first data set based on the plurality of feature codes and the plurality of coding values; and inputting the first data set into a preset workload prediction model, and determining the workload corresponding to the target imaging report.

[0012] The present application provides a method for quantifying the workload of an imaging report. First, the processor obtains a target imaging report corresponding to the patient, and uses an NLP model to extract multiple first key features in the text content of the target imaging report. Then, the processor encodes each first key feature respectively using a pre-set encoding rule, and determines multiple feature codes corresponding to each first key feature and multiple encoding values ​​corresponding to each feature code. Further, the processor generates a corresponding first data set based on the multiple feature codes and the multiple encoding values. Finally, the processor inputs the first data set into a pre-set workload prediction model, and determines the workload corresponding to the target imaging report.

[0013] Referring to the above content, it can be seen that since the present application first uses the NLP model to extract multiple first key features that may affect the doctor's workload from the text content of the target imaging report, and uses pre-set encoding rules to encode each first key feature, therefore, unlike the prior art, the present application quantifies the degree of impact of different key features on the doctor's workload, thereby significantly distinguishing the workload consumed by the doctor when filling out each key feature of the diagnostic report based on the target imaging report.

[0014] Furthermore, since the generated first data set contains feature codes and code values ​​corresponding to each first key feature, when the first data set is input into a pre-set workload prediction model, the workload prediction model can truly reflect the workload of the doctor in filling out the diagnostic report based on the target imaging report in a visual form.

[0015] Thus, the technical effect of truly reflecting the workload of doctors when filling out diagnostic reports based on target imaging reports and further improving the work enthusiasm of doctors is achieved. This solves the technical problem in the prior art that the traditional workload quantification method of filling out diagnostic reports based on imaging reports cannot truly reflect the workload of doctors, and thus cannot further improve the work enthusiasm of doctors. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present application. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

[0017] Figure 1 is a hardware structure block diagram of a computing device for implementing the method according to Embodiment 1 of the present application;

[0018] Figure 2 is a schematic diagram of a workload quantification system for image reporting according to Example 1 of the present application;

[0019] Figure 3 is a flow chart of the workload quantification method of the image report according to Example 1 of the present application;

[0020] Figure 4 is a schematic diagram of the anatomical structure according to Example 1 of the present application;

[0021] Figure 5 is a schematic diagram of a target image report corresponding to a patient and corresponding critical value keywords according to Example 1 of the present application;

[0022] Figure 6 is a schematic diagram of an interface for critical value management according to Example 1 of the present application;

[0023] Figure 7 is a schematic diagram of a workload quantification device for image reporting according to Embodiment 2 of the present application;

[0024] Figure 8 It is a schematic diagram of the workload quantification device of the image report described in Example 3 of the present application. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only embodiments of a part of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present disclosure.

[0026] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0027] Example 1

[0028] According to this embodiment, an embodiment of a method for quantifying the workload of image reporting is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings 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 can be executed in an order different from that shown here.

[0029] The method embodiment provided in this embodiment can be executed in a mobile terminal, a computer terminal, a server or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computing device for implementing a method for quantifying the workload of an image report. Figure 1As shown, the computing device may include one or more processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory for storing data, a transmission device for communication functions, and an input / output interface. The memory, the transmission device, and the input / output interface are connected to the processor via a bus. In addition, it may also include: a display, a keyboard, and a cursor control device connected to the input / output interface. A person skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0030] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computing device. As involved in the embodiments of the present disclosure, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0031] The memory can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the workload quantification method of image reporting in the embodiment of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, the workload quantification method of image reporting of the above-mentioned application program is realized. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computing device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0032] The transmission device is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computing device. In one example, the transmission device includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0033] The display may be, for example, a touch screen liquid crystal display (LCD) that may enable a user to interact with a user interface of the computing device.

[0034] It should be noted that, in some optional embodiments, the above Figure 1 The computing device shown may include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware elements and software elements. Figure 1 This is merely one example of a particular embodiment and is intended to illustrate the types of components that may be present in the computing devices described above.

[0035] Figure 2 is a schematic diagram of a workload quantification system for image reporting according to this embodiment. Figure 2 As shown, the system includes: a terminal device 100 and a processor 200. The doctor uploads the patient's target image report to the processor 200 through the terminal device 100. Therefore, when the processor 200 receives the target image report corresponding to the patient, it first uses the NLP model to extract multiple first key features in the text content of the target image report. Then, each first key feature is encoded using a preset encoding rule, and a corresponding first data set is generated. Finally, the first data set is input into a preset workload prediction model to determine the workload required for the doctor to fill in the diagnosis report based on the target image report.

[0036] Further, when the processor 200 determines the workload required for the doctor to fill in the diagnosis report based on the target image report, the processor 200 sends the determined workload to the terminal device 100. Thus, the terminal device 100 can display the determined workload to the doctor in a visual form.

[0037] It should be noted that the terminal device 100 and the processor 200 in the system can both be applicable to the hardware structure described above.

[0038] In the above operating environment, according to the first aspect of this embodiment, a method for quantifying the workload of image reporting is provided. The method comprises: Figure 2 The processor 200 shown in FIG. Figure 3 A schematic diagram showing the process of the method is shown in FIG. Figure 3 As shown, the method includes:

[0039] S302: Obtain a target imaging report corresponding to the patient, and use the NLP model to extract multiple first key features in the text content of the target imaging report, wherein the multiple first key features include examination items, the number of included parts, the number of positive parts, the number of lesions, the number of lesion measurements, the positive grade, whether it is an emergency, whether it is trauma, and whether there is a historical examination;

[0040] S304: Encode each first key feature respectively using a preset encoding rule, and determine a plurality of feature codes corresponding to each first key feature and a plurality of encoding values ​​corresponding to each feature code, wherein the encoding rule includes a key feature code and an encoding value;

[0041] S306: Generate a corresponding first data set based on the multiple feature codes and the multiple code values; and

[0042] S308: Input the first data set into a preset workload prediction model, and determine the workload corresponding to the target image report.

[0043] Specifically, first, the processor 200 obtains the target image report corresponding to the patient through the terminal device 100, and uses the NLP model to extract multiple first key features in the text content of the target image report (S302). In this embodiment, the target image report can be, for example, a medical image report, and the medical image includes two parts: text content and radiological images. The text content should meet the basic requirements of the medical image report and include at least the following aspects: the patient's gender, the patient's age, the site of examination, the type of equipment for examination, examination findings, diagnostic impressions, and information sent for examination, etc.

[0044] The text content of the target image report is exemplified as follows:

[0045] “Gender: Female, Age: 7 years old, Examination site: Chest, Examination equipment: CT

[0046] Examination findings:

[0047] The bilateral thorax is roughly symmetrical, the bronchial vessels of both lungs are clear, and strip-like high-density shadows can be seen in the middle lobe of the left lung, with a relatively uniform density. No enlarged lymph nodes were found in the bilateral hilum and mediastinum. The mediastinum did not widen, and the position was centered without displacement. No abnormalities were found in the heart and large blood vessels, and no pleural effusion was found on both sides. No obvious abnormalities were found in the bone and soft tissue structure of the thorax within the scanning range. The bilateral cerebral hemispheres were symmetrical, and spots and patchy low-density foci were found in the center of the bilateral semi-oval, the radiating crown, and the basal ganglia. The boundaries of some lesions were clear, some were blurred, and some were watery. There was no obvious expansion of the ventricular system, no widening of the cerebral sulci, cerebral columns, and cerebral cisterns, no displacement of the midline structure, and no abnormal density shadows were found in the brainstem and cerebellum. No abnormal changes were found in the skull bones.

[0048] Diagnostic Impression:

[0049] Atelectasis of the right middle lobe, the cause is unknown, and CT enhancement or bronchoscopy is recommended for further examination. Multiple ischemic lesions in the bilateral semiovale center, corona radiata, and base segment areas, and senile cerebral infarction. No abnormalities were found in the cavernous bodies.

[0050] Inspection information:

[0051] Patient source: outpatient department; Application department: internal medicine; Application form number: YY20240912; Whether it is an emergency: No; Whether it is a trauma: Yes; Whether it is a follow-up examination: Yes; Description of condition: chest tightness and nausea. "

[0052] Thus, when the processor 200 obtains the target image report corresponding to the patient, the target image report is input into the general NLP model, and the entity content of each sentence of the image report in the text content of the target image report is extracted using the NLP model. Among them, the entity content includes entities and the relationship between entities. Entities include, for example, lungs, hilum, trachea, kidneys, liver, etc. The relationship between entities includes, for example, positional relationships, sign descriptions, and qualitative relationships, etc. Among them, the positional relationship includes location and location. Sign descriptions include location and positive / negative signs. Qualitative relationships include qualitative words and signs.

[0053] Table 1 shows the entity types.

[0054] Table 1

[0055]

[0056] Table 1 shows entity types, English representations of entity types, descriptions of entity types, and corresponding examples of entity types. For example, the entity type is a part entity, and the English representation corresponding to the part entity is nsf. The part entity is used to indicate an anatomical part of the body, and the entity type may be, for example, the brain, thoracic vertebrae, lungs, and blood vessels.

[0057] Table 2 shows the entity relationships.

[0058] Table 2

[0059]

[0060] Among them, the above Table 2 shows the entity relationship, the antecedent entity corresponding to the entity relationship, the posterior entity corresponding to the entity relationship, and the corresponding examples. And the positive sign relationship and the negative sign relationship shown in Table 2 are descriptions of a part. For example, the entity relationship is a direction relationship, the antecedent entity is a direction entity (ntcb), the posterior entity is a part entity (nsf), and the entity relationship can be, for example, the left side, thorax, etc.

[0061] Further, the processor 200 may determine the masculine or feminine nature of the current description sentence according to the qualitative relationship and the positive sign relationship shown in Table 2. Table 3 shows a method for determining the masculine or feminine nature of a description sentence.

[0062] Table 3

[0063]

[0064] Referring to Table 3, the determination method of the masculine and feminine nature of the description sentence includes: whether there are masculine signs in the description sentence, whether there is a qualitative relationship in the description sentence, and the masculine and feminine nature of the description sentence. For example, if there are masculine signs in the description sentence but no qualitative relationship, the description sentence is masculine.

[0065] The processor 200 then determines the positive level (L0-L5) of the current lesion according to the configuration level of the positive sign, where L0 indicates negative, L1-L5 indicates positive, and the higher the level, the more serious the disease. For example, "pulmonary calcification" is L2, and "lung cancer" is L5.

[0066] Figure 4 is a schematic diagram of the anatomical structure according to an embodiment of the present application. Figure 4 As shown, the anatomical structure includes the head, neck, and chest. The chest includes the lungs. The lungs include pulmonary lobules, lung fields, lungs, and pulmonary valves. Thus, the processor 200 can confirm the inspection scope of the image report by combining the part entity with the anatomical structure. That is, several major parts are diagnosed. For example, "lung-chest", "liver-abdomen", etc.

[0067] The processor 200 can also determine how many positive parts there are in the report based on the positive sentences. In addition, the processor 200 can also determine the number of lesion measurements based on the description of the lesion. The processor 200 can also confirm whether it is an emergency, whether it is a trauma, and whether there is a historical examination based on the report inspection information.

[0068] Thus, through the above method, the processor 200 can determine multiple first key features using the NLP model. Among them, the selection of the first key feature should be able to reflect the text content characteristics of the target image report and the time invested by the doctor in filling out the diagnosis report based on the target image report. And among them, the first key feature includes: examination items, the number of parts included, the number of positive parts, the number of lesions, the number of lesion measurements, the positive grade, whether it is an emergency, whether it is traumatic, and whether there is a historical examination, etc. It is worth noting that the above is just an example of the type of the first key feature, and the actual situation is not limited to this.

[0069] Table 4 shows the first key features corresponding to the target image report.

[0070] Table 4

[0071]

[0072] For example, the text content of the target image report includes "A scan from the apex to the lower edge of the pubic symphysis of both lungs showed: increased transmittance of both lungs, high-density linear shadows and calcifications in the right upper lobe and lower lobes of both lungs, regular distribution of textures in both lungs, clear hilar structure, unobstructed trachea, no displacement of the mediastinum, and enlarged lymph nodes visible therein". After the above text content is input into the NLP model, the content shown in Table 4 can be output. Therefore, when the processor 200 determines the content of the above Table 4, it can further determine the first key feature corresponding to the text content of the target image report.

[0073] For example, when the processor 200 obtains the target image report in the above example, the target image report is input into the NLP model, and the processor 200 determines "Examination items: CT chest; Number of parts included: 8; Number of positive parts: 3; Number of lesions: 2; Number of lesion measurements: 1; Positive level: L4; Whether it is an emergency: 1 (yes); Whether it is a trauma: 0 (not a trauma); Whether there is a historical examination: 0 (no historical examination)" according to the output result of the NLP model.

[0074] When the processor 200 determines a plurality of first key features in the target image report, it encodes each first key feature using a preset encoding rule, and determines a plurality of feature codes corresponding to each first key feature and a plurality of encoding values ​​corresponding to each feature code (S304). The preset encoding rule may be, for example, a encoding table, and the encoding table includes a number, a key feature name, a key feature code, a encoding value, and a feature description.

[0075] Table 5 shows the encoding rules.

[0076] Table 5

[0077]

[0078] Referring to Table 5 above, the number of parts included is used to indicate which parts are diagnosed in the target imaging report. For example, when the value is 3, it means that 3 parts of the patient are diagnosed in the target imaging report.

[0079] The number of positive sites is used to indicate which of the diagnosed sites are diseased. For example, when the value is 3, it means that the patient has 3 diseased sites.

[0080] The number of lesions is used to indicate how many lesions are diagnosed in the target imaging report. For example, when the value is 1, it means that 1 lesion is diagnosed.

[0081] The lesion measurement number is used to indicate how many lesions are measured in the target image report. For example, when the value is 3, it means that 3 lesions are measured.

[0082] The positive grade is used to indicate the severity of the lesions, where 0 means no disease, 1 means mild, 2 means milder, 3 means moderate, 4 means heavier, and 5 means severe.

[0083] Whether it is an emergency report is used to indicate whether the target imaging report is an emergency report, where 0 indicates that it is not an emergency report and 1 indicates that it is an emergency report.

[0084] Whether trauma is used to indicate whether the target image report is a trauma report, where 0 indicates not a trauma report and 1 indicates a trauma report.

[0085] Whether there is a historical check is used to indicate whether there is a historical check that needs to be compared. 0 means there is no historical check that needs to be compared, and 1 means there is a historical check that needs to be compared.

[0086] For example, the processor 200 determines multiple coding features of the target image report in the above example and the coding values ​​corresponding to each coding feature according to the pre-set coding rules: "Examination items (jcxm): CT chest—0105; Number of parts included (bwzl): 8—8; Number of positive parts (yxbw): 3—3; Number of lesions (bzsl): 2—2; Number of lesion measurements (bzcl): 1—1; Positive level (yxdj): L4—4; Whether it is an emergency (jz): 1 (yes)—1; Whether it is a trauma (ws): 0 (not a trauma)—0; Whether there is a historical examination (lsjc): 0 (no historical examination)—0".

[0087] Then, the processor 200 generates a corresponding first data set based on the determined multiple coding features and the coding values ​​corresponding to each coding feature (S306). For example, the processor 200 generates a first data set X={jcxm:0105; bwzl:8; yxbw:3; bzsl:2; bzcl:1; yxdj:4; jz:1; ws:0; lsjc:0}.

[0088] Finally, the processor 200 inputs the first data set into a preset workload prediction model, and determines the workload corresponding to the target imaging report (S308). The workload prediction model may be, for example, a random forest model. And when the processor 200 inputs the first data set into the trained random forest model, the random forest model can predict the time (i.e., workload) required for the doctor to fill in the diagnostic report based on the target imaging report.

[0089] For example, the processor 200 inputs the first data set in the above example into the workload prediction model, so that the workload prediction model outputs Y = 18. Wherein, Y is used to indicate the time consumption of the doctor to diagnose the target imaging report.

[0090] As described in the background technology, the traditional way to quantify the workload of doctors is to assign a score to each examination item. However, the time spent by doctors in filling out diagnostic reports based on imaging reports is not the same. That is, for more complex imaging reports, doctors spend more time; for simpler imaging reports, doctors spend less time. Therefore, if the traditional quantification method is used to quantify the workload of doctors filling out diagnostic reports based on imaging reports, it cannot truly reflect the workload of doctors.

[0091] In view of this, the present application provides a method for quantifying the workload of imaging reports. Referring to the above-mentioned contents, it can be seen that since the present application first uses the NLP model to extract multiple first key features that may affect the workload of doctors from the text content of the target imaging report, and encodes each first key feature separately using the pre-set encoding rules, therefore, unlike the prior art, the present application quantifies the degree of influence of different key features on the workload of doctors, thereby being able to significantly distinguish the workload consumed by doctors when filling out each key feature of the diagnostic report based on the target imaging report.

[0092] Furthermore, since the generated first data set contains feature codes and code values ​​corresponding to each first key feature, when the first data set is input into a pre-set workload prediction model, the workload prediction model can truly reflect the workload of the doctor in filling out the diagnostic report based on the target imaging report in a visual form.

[0093] Thus, the technical effect of truly reflecting the workload of doctors when filling out diagnostic reports based on imaging reports and further improving the work enthusiasm of doctors is achieved. This solves the technical problem in the prior art that the traditional workload quantification method of filling out diagnostic reports based on imaging reports cannot truly reflect the workload of doctors, thereby failing to further improve the work enthusiasm of doctors.

[0094] Optionally, it also includes: constructing a workload prediction model and training the workload prediction model. Further optionally, the operations of constructing a workload prediction model and training the workload prediction model include: determining input samples for training the workload prediction model; determining output samples for training the workload prediction model; and using the input samples and the output samples to train the workload prediction model. Further optionally, the operations of using the input samples and the output samples to train the workload prediction model include: collecting multiple image report samples, and using the NLP model to extract multiple second key features of each image report sample; using pre-set encoding rules to encode each second key feature, and generating corresponding multiple second data sets; determining the workload corresponding to each image report sample; and using multiple second data sets as input samples, using the workload corresponding to each image report sample as output samples, and training the workload prediction model.

[0095] Specifically, before the processor 200 uses the workload prediction model to determine the workload required for the doctor to fill out the diagnosis report based on the target imaging report, it is also necessary to pre-build the workload prediction model and train the workload prediction model.

[0096] First, the processor 200 constructs a workload prediction model, wherein the processor 200 can, for example, construct a random forest model using the RandomForestRegressor class provided by the Scikit-learn library.

[0097] Then, the processor 200 randomly obtains a plurality of image report samples from the database. The database may include, for example, pre-stored historical image reports. Further, the processor 200 extracts each image report sample using the NLP model. For example, the processor 200 uses the NLP model to extract the image report sample Multiple second key features in , the processor 200 uses the NLP model to extract the image report sample Multiple second key features in , the processor 200 uses the NLP model to extract the image report sample Multiple second key features in , ..., the processor 200 uses the NLP model to extract the image report sample Multiple second key features in .

[0098] The processor 200 then uses the encoding rules shown in Table 5 to encode each image report sample. Encode multiple second key features in and generate corresponding multiple second data sets For example, the processor 200 reports a sample image. The second key features are encoded and a second data set is generated. ; Processor 200 reports sample images The second key features are encoded and a second data set is generated. ; Processor 200 reports sample images The second key features are encoded and a second data set is generated. ; ...; Processor 200 reports sample images The second key features are encoded and a second data set is generated. .

[0099] Then, the processor 200 determines the image report samples For example, the processor 200 determines that the doctor based on the image report sample The amount of work required to fill out the diagnostic report is , the processor 200 determines that the doctor based on the image report sample The amount of work required to fill out the diagnostic report is , the processor 200 determines that the doctor based on the image report sample The amount of work required to fill out the diagnostic report is , ..., the processor 200 determines that the doctor based on the image report sample The amount of work required to fill out the diagnostic report is .

[0100] Finally, the processor 200 converts the plurality of second data sets into as input samples and the workload corresponding to each image report sample As output samples, the workload prediction model is trained. and workload Corresponding to the second data set and workload Corresponding to the second data set and workload Corresponding to,..., the second data set and workload correspond.

[0101] Among them, for a given sample (i.e., the second data set ), the prediction results of the random forest model (i.e., workload ) is the average of all individual decision tree predictions:

[0102]

[0103] in, represents the prediction result of the random forest model, u represents the number of decision trees, and u=1~U, Represents the prediction result of the decision tree.

[0104] Random forest is an integrated machine learning method that improves prediction accuracy and controls overfitting by building multiple decision trees and taking the average of their prediction results. First, random forest extracts samples from the original data set through replacement sampling to create several training subsets, ensuring that the data input of each model is different, enhancing the diversity of the data, helping to improve the feature space resolution, and forming a more accurate and smooth decision boundary.

[0105] Later, when splitting the node, instead of considering all features, a subset of features is randomly selected for evaluation. The number of features selected is usually the square root or logarithm of the total number of features. This method not only speeds up the search, but also increases the nonlinearity of the model, making the decision boundary more complex and effective. During the training process, decision trees use specific evaluation criteria (such as information gain, Gini impurity, etc.) to determine the best split point. Different decision trees generate different splitting rules because they are based on different samples and feature subsets, thereby reducing the bias of a single model. Finally, random forests reduce the variance of a single tree by integrating the results of multiple decision trees, improve the model's prediction performance for unseen data, and achieve stronger representation capabilities and better generalization results than a single decision tree.

[0106] It is worth noting that the processor 200 can adjust the number of decision trees in the random forest, the method of selecting features, and how each tree grows at any time according to the changes in the loss and accuracy of the training random forest. The selection of these parameters will directly affect the behavior of the model, so careful evaluation, analysis and adjustment are required. The random forest model gradually establishes multiple decision trees by analyzing the samples in the training set and their corresponding target variable values, and performs feature selection and partitioning on each decision tree.

[0107] In this embodiment, for example, a test set for detecting the random forest model can also be set. Thus, after the training of the random forest model is completed, the processor 200 uses the samples in the test set to detect the random forest model. For the random forest model, it averages or weighted averages the prediction results of each decision tree to obtain the final prediction value. Then, the processor 200 compares these predicted results with the actual target variable value to evaluate the performance of the random forest model. In order to more accurately measure the performance of the model, accuracy evaluation indicators such as mean square error (MSE), mean absolute error (MAE) and / or determination coefficient R2 are calculated, and the optimal model result is selected based on the results.

[0108] Preferably, in this embodiment, the decision trees in the trained random forest model can be screened using the principal component analysis (PCA) method, and then the optimal model result can be further determined from the screened decision trees using accuracy evaluation indicators such as mean square error (MSE), mean absolute error (MAE) and / or determination coefficient R2.

[0109] Specifically, in this embodiment, it is assumed that the trained decision trees in the random forest model have n Define the vector X =[ x 1 , x 2 , ..., x n ] T .in, x j For the same sample, j The number of hours predicted by a decision tree.

[0110] Thus, for the test set m samples, and input them into n A decision tree can be obtained m The first decision tree vector { X 1 , X 2 , ..., X m}.in, X i =[ x i,1 , x i,2 , ..., x i,n ] T . And among them x i,j Indicates that the i Sample inputj The predicted working hours by the decision tree. X 1 , X 2 , ..., X m ] constitutes the following first decision tree matrix:

[0111]

[0112] In generating the above first decision tree matrix [ X 1 , X 2 , ..., X m ], the processor 200 processes the first decision tree matrix [ X 1 , X 2 , ..., X m ] is normalized to generate a second decision tree matrix. The specific calculation formula for normalizing the first decision tree matrix is ​​as follows:

[0113]

[0114] in, i =1~m. j =1~n. And among them, Represents the element value in the second decision tree matrix after normalization. Represents the element value in the first decision tree matrix without normalization. represents the average of the first decision tree vectors in the first decision tree matrix, where the first decision tree vector is X i , and i=1~m. Represents the variance of each first decision tree vector.

[0115] And the specific calculation formula of the average value of each first decision tree vector is as follows:

[0116]

[0117] Among them, j=1~n.

[0118] The specific calculation formula of the variance of each first decision tree vector is as follows:

[0119]

[0120] Among them, j=1~n.

[0121] Thus, the processor 200 calculates the value of the element in the second decision tree matrix after the normalization process. , the second decision tree matrix can be determined .in, .

[0122] After that, when the processor 200 determines the second decision tree matrix, it can calculate the corresponding sample correlation matrix based on the second decision tree matrix. The specific calculation formula is as follows:

[0123]

[0124] Among them, R represents the sample correlation matrix, and Y represents the second decision tree matrix.

[0125] Further, the processor 200 performs eigendecomposition on the sample correlation matrix R and determines a plurality of first eigenvalues. The specific calculation formula is as follows:

[0126]

[0127] Where R represents the sample correlation matrix, I represents the identity matrix, and λ represents the first eigenvalue.

[0128] Thus, the processor 200 can calculate the n first eigenvalues ​​λ of the sample correlation matrix R based on the above formula: 1 ,λ 2 , ..., λ n . And among them, λ 1 ≥λ 2 ≥...≥λ n .

[0129] Then, when the processor 200 determines the n first eigenvalues ​​of the sample correlation matrix R, the number of first eigenvalues ​​whose variance contribution rate reaches a preset threshold is further determined, and a plurality of second eigenvalues ​​k are determined. The specific calculation formula is as follows:

[0130]

[0131] in, represents the variance contribution rate of the kth first eigenvalue among multiple first eigenvalues, represents the sum of the n first eigenvalues, Represents the variance contribution rate.

[0132]

[0133] in, represents the sum of the variance contributions of the k first eigenvalues, represents the sum of the k first eigenvalues, represents the sum of the n first eigenvalues.

[0134] Further, the processor 200 determines the number of variance contribution rates of the plurality of first eigenvalues ​​that reaches a preset threshold, and determines a plurality of second eigenvalues ​​based on the determined number. It is worth noting that the preset threshold is a threshold set by the operator based on actual conditions, usually 70% to 80%. In this embodiment, the preset threshold is 70%. Thus, since λ has been mentioned in the above content 1 ≥λ 2 ≥...≥λ n Therefore, when the sum of the variance contribution rates of the k first eigenvalues ​​is greater than or equal to the preset value, the value of k can be determined, and multiple second eigenvalues ​​λ can be determined. 1 ,λ 2 , ..., λ k .

[0135] That is, the processor 200 can determine that k decision trees are selected from n decision trees, and the k decision trees are decision trees after redundancy removal. And when the processor 200 determines k decision trees, the precision evaluation indexes such as mean square error (MSE), mean absolute error (MAE) and / or determination coefficient R2 are used to select decision trees with higher precision from the decision trees after redundancy removal, thereby achieving the technical effect of reducing redundant calculations and improving calculation accuracy.

[0136] Therefore, according to the first aspect of this embodiment, a technical effect is achieved that can truly reflect the workload of doctors when filling out diagnosis reports based on imaging reports and further enhance the work enthusiasm of doctors.

[0137] In addition, critical value keywords are also set in the target image report, and the processor 200 can prompt the doctor to focus on the critical value keywords through the terminal device 100 according to the critical value keywords. Among them, the doctor can set the critical value keywords that need to be prompted in the application platform. For example, Figure 5 Schematic diagram of a target image report corresponding to a patient and corresponding critical value keywords according to an embodiment of the present application. Figure 5 As shown, the display interface shows the examination findings, report conclusions and critical values. Among them, the examination findings include: bilateral thoracic symmetry, tracheomediastinum in the middle. Flake-like translucent shadows can be seen in the bilateral chest cavity, which can compress the edge of the lung, the right lung compression is about 75%, and the left lung compression is less than 10%. Bilateral pleural thickening. The translucency of both lungs is reduced.

[0138] The report conclusions include: 1. Bilateral pneumothorax signs, right lung compression is about 75%, left lung compression is less than 10%, and the translucency of both lungs is reduced. Please combine clinical analysis. 2. Bilateral pleural thickening.

[0139] Critical values ​​include: Compression. Among them, the examination findings include 2 "compressions" and the report conclusion includes 1 "compression".

[0140] Further, Figure 6 Schematic diagram of the interface of the critical value management according to the embodiment of the present application. Figure 6 As shown in the figure, the main menu of the data platform includes: error report log, BPR management, part conflict whitelist, NER hot update, REF hot update, critical value management and positive level management. The interface of critical value management includes: critical value sequence number, critical value abbreviation, critical value description, part word in short sentence, short sentence symptom word, key words in report and equipment type.

[0141] For example, the critical value corresponding to the sequence number "3" is "brain herniation", and the description corresponding to "brain herniation" is "brain herniation Midline structure displacement>1.0cm", there is no location word in the short sentence, the symptom word in the short sentence is "brain hernia", there are no keywords in the report, and there is no device type.

[0142] In addition, reference Figure 6 As shown, doctors can click "Add" on this interface to customize the critical value keywords that need to be reported, thereby ensuring that doctors can receive prompts in a timely manner when filling out the diagnosis report based on the target imaging report.

[0143] In addition, reference Figure 1 As shown, according to the second aspect of this embodiment, a storage medium is provided, wherein the storage medium includes a stored program, wherein when the program is run, a processor executes any one of the above methods.

[0144] Therefore, according to this embodiment, a technical effect is achieved that can truly reflect the workload of doctors when filling out diagnosis reports based on imaging reports and further enhance the work enthusiasm of doctors.

[0145] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0146] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0147] Example 2

[0148] Figure 7 FIG. 7 shows a device 700 for quantifying workload of an image report according to this embodiment, and the device 700 corresponds to the method according to embodiment 1. Figure 7 As shown, the device 700 includes: a first feature extraction module 710, which is used to obtain a target imaging report corresponding to the patient, and use an NLP model to extract multiple first key features in the text content of the target imaging report, wherein the multiple first key features include examination items, the number of parts included, the number of positive parts, the number of lesions, the number of lesion measurements, the positive grade, whether it is an emergency, whether it is a trauma, and whether there is a historical examination; an encoding module 720, which is used to encode each first key feature respectively using a preset encoding rule, and determine multiple feature codes corresponding to each first key feature and multiple coding values ​​corresponding to each feature code, wherein the encoding rule includes a key feature code and a coding value; a first data set generation module 730, which is used to generate a corresponding first data set based on multiple feature codes and multiple coding values; and a workload determination module 740, which is used to input the first data set into a preset workload prediction model and determine the workload corresponding to the target imaging report.

[0149] Optionally, the device 700 further includes: a model training module, used to construct a workload prediction model and train the workload prediction model.

[0150] Optionally, the model training module includes: an input sample determination module, used to determine the input samples for training the workload prediction model; an output sample determination module, used to determine the output samples for training the workload prediction model; and a model training submodule, used to train the workload prediction model using the input samples and output samples.

[0151] Optionally, the model training submodule includes: a second key feature extraction module, which is used to collect multiple imaging report samples and use the NLP model to extract multiple second key features of each imaging report sample; a second data set generation module, which is used to encode each second key feature respectively using a preset encoding rule and generate corresponding multiple second data sets; a workload determination module, which is used to determine the workload of diagnosing each imaging report sample; and a model training submodule, which is used to take the multiple second data sets as input samples, take the workload corresponding to each imaging report sample as output sample, and train the workload prediction model.

[0152] Therefore, according to this embodiment, a technical effect is achieved that can truly reflect the workload of doctors when filling out diagnosis reports based on imaging reports and further enhance the work enthusiasm of doctors.

[0153] Example 3

[0154] Figure 8 FIG. 8 shows a device 800 for quantifying workload of image reporting according to this embodiment, and the device 800 corresponds to the method according to embodiment 1. Figure 8 As shown, the device 800 includes: a processor 810; and a memory 820, which is connected to the processor 810 and is used to provide the processor 810 with instructions for processing the following processing steps: obtaining a target imaging report corresponding to the patient, and using the NLP model to extract multiple first key features in the text content of the target imaging report, wherein the multiple first key features include examination items, the number of parts included, the number of positive parts, the number of lesions, the number of lesion measurements, the positive grade, whether it is an emergency, whether it is trauma, and whether there is a historical examination; using a preset coding rule to encode each first key feature respectively, and determine multiple feature codes corresponding to each first key feature and multiple coding values ​​corresponding to each feature code, wherein the coding rule includes a key feature code and a coding value; based on the multiple feature codes and the multiple coding values, generate a corresponding first data set; and input the first data set into a preset workload prediction model, and determine the workload corresponding to the target imaging report.

[0155] Optionally, the device 800 further includes: constructing a workload prediction model, and training the workload prediction model.

[0156] Optionally, the operations of constructing a workload prediction model and training the workload prediction model include: determining input samples for training the workload prediction model; determining output samples for training the workload prediction model; and training the workload prediction model using the input samples and the output samples.

[0157] Optionally, the operation of training the workload prediction model using input samples and output samples includes: collecting multiple imaging report samples, and extracting multiple second key features of each imaging report sample using an NLP model; encoding each second key feature respectively using a pre-set encoding rule, and generating corresponding multiple second data sets; determining the workload of diagnosing each imaging report sample; and using the multiple second data sets as input samples, using the workload corresponding to each imaging report sample as output samples, and training the workload prediction model.

[0158] Therefore, according to this embodiment, a technical effect is achieved that can truly reflect the workload of doctors when filling out diagnosis reports based on imaging reports and further enhance the work enthusiasm of doctors.

[0159] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0160] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0161] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0162] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0163] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0164] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0165] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for quantifying workload of image reporting, characterized in that: include: Obtain a target imaging report corresponding to the patient, and extract a plurality of first key features in the text content of the target imaging report using an NLP model, wherein the plurality of first key features include examination items, the number of included parts, the number of positive parts, the number of lesions, the number of lesion measurements, the positive grade, whether it is an emergency, whether it is trauma, and whether there is a historical examination; Encode each first key feature respectively using a preset encoding rule, and determine a plurality of feature codes corresponding to each first key feature and a plurality of encoding values ​​corresponding to each feature code, wherein the encoding rule includes a key feature code and an encoding value; Based on the plurality of feature codes and the plurality of code values, generating a corresponding first data set; as well as The first data set is input into a preset workload prediction model, and the workload corresponding to the target image report is determined, wherein the workload prediction model is a random forest model, and The method further includes: screening the plurality of first decision trees in the random forest model using a principal component analysis method, and determining a plurality of second decision trees; and In the case of determining the plurality of second decision trees, a plurality of third decision trees are determined using mean square error, mean absolute error and / or determination coefficient, wherein the precision of the plurality of third decision trees is greater than the precision of the plurality of second decision trees, and wherein the plurality of first decision trees in the random forest model are screened using principal component analysis, and an operation of determining the plurality of second decision trees comprises: Inputting a plurality of predetermined samples into a plurality of first decision trees respectively, to obtain first decision tree vectors corresponding to the respective samples, wherein the dimension of the first decision tree vector corresponds to the number of the first decision trees, and each element in the first decision tree vector indicates the man-hours predicted by inputting the corresponding sample into each first decision tree; Based on the multiple first decision tree vectors, a first decision tree matrix is ​​constructed, and the first decision tree matrix is ​​normalized to generate a second decision tree matrix, wherein the calculation formula for normalizing the first decision tree matrix is ​​as follows: ； in, i =1~m, j =1~n, represents the element value in the second decision tree matrix after normalization. represents the element value in the first decision tree matrix without normalization, represents the average value of each first decision tree vector in the first decision tree matrix, where the first decision tree vector is X i , and i=1~m, Represents the variance of each of the first decision tree vectors; And the calculation formula of the average value of each first decision tree vector is as follows: ； Among them, j=1~n, represents the element value in the first decision tree matrix without normalization; And the calculation formula of the variance of each first decision tree vector is as follows: ； Among them, j=1~n, represents the element value in the first decision tree matrix without normalization, represents the average value of each first decision tree vector in the first decision tree matrix; Based on the second decision tree matrix, the corresponding sample correlation matrix is ​​calculated, wherein the calculation formula of the sample correlation matrix is ​​as follows: ； Where R represents the sample correlation matrix, and Y represents the second decision tree matrix; Perform eigendecomposition on the sample correlation matrix to determine a plurality of first eigenvalues, wherein a calculation formula for determining the plurality of first eigenvalues ​​is as follows: ； Where R represents the sample correlation matrix, I represents the identity matrix, and λ represents the first eigenvalue; When the plurality of first eigenvalues ​​are determined, the number of first eigenvalues ​​whose variance contribution rate reaches a preset threshold is calculated, and a plurality of second eigenvalues ​​are determined, wherein the calculation formula of the variance contribution rate is as follows: ； in, represents the variance contribution rate of the kth first eigenvalue among multiple first eigenvalues, represents the sum of the n first eigenvalues, represents the variance contribution rate; ； in, represents the sum of the variance contributions of the k first eigenvalues, represents the sum of the k first eigenvalues, represents the sum of the n first eigenvalues; and In the case of determining the plurality of second eigenvalues, the plurality of second decision trees are determined based on the number corresponding to the plurality of second eigenvalues.

2. The method according to claim 1, characterized in that Also includes: Constructing the workload prediction model and training the workload prediction model.

3. The method according to claim 2, characterized in that The operations of constructing the workload prediction model and training the workload prediction model include: Determining input samples for training the workload prediction model; Determining output samples for training the workload prediction model; and The workload prediction model is trained using the input samples and the output samples.

4. The method according to claim 3, characterized in that The operation of training the workload prediction model using the input sample and the output sample includes: Acquire multiple imaging report samples, and extract multiple second key features from each imaging report sample using the NLP model; Encode each second key feature using a preset encoding rule, and generate a corresponding plurality of second data sets; Determining the workload corresponding to each of the image report samples; and The multiple second data sets are used as the input samples, the workload corresponding to each image report sample is used as the output sample, and the workload prediction model is trained.

5. A storage medium, characterized in that: The storage medium includes a stored program, wherein when the program is run, the processor executes the method according to any one of claims 1 to 4.

6. A device for quantifying workload of image reporting, characterized in that: include: Processor and memory; The memory is used to store computer programs; The processor is used to execute the computer program and implement the method according to any one of claims 1 to 4 when executing the computer program.

Images