CN116050691A

CN116050691A - Medical image report evaluation method, device, electronic equipment and storage medium

Info

Publication number: CN116050691A
Application number: CN202211421959.3A
Authority: CN
Inventors: 许祥军; 刘鸣谦; 付星月; 周炜; 陈旭; 孙嘉明; 宋晓霞; 高玉杰; 赵大平; 黄智勇
Original assignee: Winning Health Technology Group Co Ltd
Current assignee: Winning Health Technology Group Co Ltd
Priority date: 2022-11-14
Filing date: 2022-11-14
Publication date: 2023-05-02

Abstract

The application provides a medical image report evaluation method, a medical image report evaluation device, electronic equipment and a storage medium, and relates to the technical field of medical data processing. The method comprises the following steps: after acquiring text content in a medical image report, carrying out integrity assessment by adopting a pre-trained machine reading understanding model according to the attribute text and the image diagnosis text to obtain an integrity assessment result of the image diagnosis text; carrying out consistency evaluation on the medical image report according to the image diagnosis text and the pathological diagnosis report corresponding to the medical image report to obtain a diagnosis consistency evaluation result of the medical image report; and finally, obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text and the diagnosis consistency evaluation result. Therefore, the method and the device realize the evaluation of the integrity, the diagnostic consistency and the like of the image diagnostic text.

Description

Medical image report evaluation method, device, electronic equipment and storage medium

Technical Field

The invention relates to the technical field of medical data processing, in particular to a medical image report evaluation method, a device, electronic equipment and a storage medium.

Background

To monitor the quality of medical text, the medical text is typically intrinsically controlled by pre-specified rules.

However, the current medical text object for quality control of the medical text is single in form and can only be controlled according to all listed rule descriptions, and once the medical text is subjected to non-listed items, the medical text cannot be controlled, so that the efficiency is low and the accuracy is low.

Disclosure of Invention

The present invention aims to solve the above-mentioned drawbacks of the prior art and provide a medical image report evaluation method, device, electronic equipment and storage medium, so as to perform multi-angle and comprehensive evaluation on a medical image report.

In order to achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

in a first aspect, an embodiment of the present application provides a medical image report evaluation method, including:

acquiring text content in a medical image report, wherein the text content comprises: attribute text and image diagnosis text; the attribute text is used for representing the attribute of the medical image report;

and carrying out integrity assessment by adopting a pre-trained machine reading understanding model according to the attribute text and the image diagnosis text to obtain an integrity assessment result of the image diagnosis text, wherein the integrity assessment result comprises the following steps: the integrity evaluation value is used for representing the integrity of a plurality of medical attention items corresponding to the medical image report in the image diagnosis text;

Carrying out consistency evaluation on the medical image report according to the image diagnosis text and the pathological diagnosis report corresponding to the medical image report to obtain a diagnosis consistency evaluation result of the medical image report;

and obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text and the diagnosis consistency evaluation result.

Optionally, the performing integrity assessment by using a pre-trained machine reading understanding model according to the attribute text and the image diagnosis text to obtain an integrity assessment result of the image diagnosis text includes:

determining the plurality of medical attention items according to the attribute text;

and according to the medical attention items and the image diagnosis text, evaluating by adopting the machine reading understanding model to obtain the integrity evaluation result.

Optionally, the integrity evaluation result further includes: position information of a text corresponding to a target medical attention item in the medical attention items; the method further comprises the steps of:

according to the position information, a pre-trained sign abnormality evaluation model is adopted to obtain an abnormality evaluation conclusion of the sign corresponding to the target attention item;

The obtaining the evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text and the diagnosis consistency evaluation result comprises the following steps:

and obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text, the diagnosis consistency evaluation result and the abnormal evaluation conclusion.

Optionally, the performing consistency evaluation on the medical image report according to the image diagnosis text and the pathological diagnosis report corresponding to the medical image report to obtain a diagnosis consistency evaluation result of the medical image report includes:

according to the image diagnosis text and the pathological diagnosis report, determining standard terms of a plurality of first candidate diseases corresponding to the image diagnosis text and standard terms of a plurality of second candidate diseases corresponding to the pathological diagnosis report from a preset disease knowledge base respectively;

determining a first target disease corresponding to the image diagnosis text from the plurality of first candidate diseases according to the image diagnosis text and standard terms of the plurality of first candidate diseases;

determining a second target disease corresponding to the pathological diagnosis text from the plurality of second candidate diseases according to the pathological diagnosis text and standard terms of the plurality of second candidate diseases;

And carrying out consistency assessment on the medical image report according to the first target disease and the second target disease to obtain a diagnosis consistency assessment result of the medical image report.

Optionally, the method further comprises:

adopting a pre-trained diagnosis conclusion identification model to identify the diagnosis conclusion of the image diagnosis text, and obtaining a diagnosis conclusion identification result corresponding to the image diagnosis text;

and obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text, the diagnosis consistency evaluation result and the diagnosis conclusion identification result.

Optionally, the method further comprises:

determining an abnormal part corresponding to the medical image report according to the image diagnosis text;

obtaining a checking part matching result of the medical image report according to the abnormal part and the diagnosis conclusion in the image diagnosis text;

obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text and the diagnosis consistency evaluation result, wherein the evaluation result comprises the following steps:

And obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text, the diagnosis consistency evaluation result and the examination part matching result.

Optionally, the method further comprises:

respectively carrying out integrity assessment of patient information on the attribute texts, carrying out integrity assessment of the image diagnosis texts and checking information, and obtaining an integrity conclusion of the patient information and an integrity conclusion of the checking information;

and obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text, the diagnosis consistency evaluation result, the patient information integrity conclusion and the examination information integrity conclusion.

In a second aspect, embodiments of the present application further provide a medical image report evaluation apparatus, including:

the acquisition module is used for acquiring text contents in the medical image report, wherein the text contents comprise: attribute text and image diagnosis text; the attribute text is used for representing the attribute of the medical image report;

the integrity evaluation module is used for carrying out integrity evaluation by adopting a pre-trained machine reading understanding model according to the attribute text and the image diagnosis text to obtain an integrity evaluation result of the image diagnosis text, and the integrity evaluation result comprises: the integrity evaluation value is used for representing the integrity of a plurality of medical attention items corresponding to the medical image report in the image diagnosis text;

The consistency evaluation module is used for carrying out consistency evaluation on the medical image report according to the image diagnosis text and the pathological diagnosis report corresponding to the medical image report to obtain a diagnosis consistency evaluation result of the medical image report;

and the evaluation result generation module is used for obtaining the evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text and the diagnosis consistency evaluation result.

In a third aspect, an embodiment of the present application further provides an electronic device, including: the medical image report evaluation method according to any one of the first aspect, wherein the program instructions executable by the processor are stored in the storage medium, and when the electronic device is running, the processor communicates with the storage medium through the bus, and the processor executes the program instructions to perform the steps of the medical image report evaluation method according to any one of the first aspect.

In a fourth aspect, embodiments of the present application further provide a computer-readable storage medium, on which a computer program is stored, which when executed by a processor performs the steps of the medical image report evaluation method according to any one of the first aspects.

The beneficial effects of this application are: the embodiment of the application provides a medical image report evaluation method, which comprises the steps of acquiring text content in a medical image report, and performing integrity evaluation by adopting a pre-trained machine reading understanding model according to attribute text and image diagnosis text to obtain an integrity evaluation result of the image diagnosis text; carrying out consistency evaluation on the medical image report according to the image diagnosis text and the pathological diagnosis report corresponding to the medical image report to obtain a diagnosis consistency evaluation result of the medical image report; and finally, obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text and the diagnosis consistency evaluation result. Therefore, the method and the device realize the evaluation of the integrity, the consistency and the like of the image diagnosis text, and prompt the improvement direction of the medical image report by obtaining the evaluation result of the medical image report. In addition, the evaluation result of the medical image report can also guide related staff to improve the writing quality of the image diagnosis text.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flowchart of a medical image report evaluation method according to an embodiment of the present application;

FIG. 2 is a flowchart of a medical image report evaluation method according to another embodiment of the present disclosure;

FIG. 3 is a flowchart of a medical image report evaluation method according to another embodiment of the present disclosure;

FIG. 4 is a flowchart of a medical image report evaluation method according to another embodiment of the present disclosure;

FIG. 5 is a flowchart of a medical image report evaluation method according to another embodiment of the present disclosure;

FIG. 6 is a flowchart of a medical image report evaluation method according to a further embodiment of the present application;

FIG. 7 is a flowchart of a medical image report evaluation method according to a fourth embodiment of the present disclosure;

FIG. 8 is a schematic diagram of a medical image report evaluation device according to an embodiment of the present application;

fig. 9 is a schematic diagram of an electronic device according to an embodiment of the present application.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention.

In this application, the terms "first," "second," and "second" are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated unless otherwise explicitly specified and defined. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include at least one feature. In the description of the present invention, the meaning of "plurality" means at least two, for example, two, three, unless explicitly specified otherwise. The terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises an element.

At present, technicians try to realize the processing of medical data by using various algorithms, and the quality control of medical records is mainly realized at present, for example, information extraction is usually carried out based on natural language processing (Natural Language Processing, NLP) technology in practice, so that the extraction work of texts is realized; and matching the extracted text with a knowledge base, and finally controlling the quality of the medical record document according to a pre-established quality control rule.

On the one hand, the method is only aimed at medical records, and the evaluation angle and the evaluation mode of the method are obviously different from those of the medical image report. In addition, the quality control of the medical text is carried out by pre-establishing quality control rules, the quality control by using the quality control rules lacks the capability of carrying out three-way and independent learning of related knowledge logic, the quality control can only be carried out according to all listed rule descriptions, and once the medical text encounters an item which is not listed, the quality control can not be carried out. The medical science essence control is carried out by the doctors in the hospitals at present based on the listed rules, and the efficiency and the accuracy are low.

On the other hand, the intelligent quality control can be performed on the medical text by using a natural language processing technology during the quality control, but the existing product usually only considers the category of the current medical quality control at the beginning of design, and the function and the realization technology of the existing product lack mobility on the quality control of other medical texts.

On the other hand, the accuracy of natural language processing is greatly affected by the labeling corpus data and quality. A large amount of marked data is usually needed to improve the effect of the quality control model, but the marking cost is increased along with the increase of the data amount.

In view of the problems currently existing, the embodiments of the present application provide a number of possible implementations for multi-angle, comprehensive assessment of medical image reports. The following is explained by way of several examples in connection with the accompanying drawings. Fig. 1 is a flowchart of a medical image report evaluation method according to an embodiment of the present application, where the method may be implemented by an electronic device running the medical image report evaluation method, and the electronic device may be, for example, a terminal device or a server. As shown in fig. 1, the method includes:

step 101: acquiring text content in a medical image report, wherein the text content comprises: attribute text and image diagnosis text; the attribute text is used to characterize the attributes of the medical image report.

The text content of the medical image report refers to information related to the medical image, wherein the information can comprise attribute text and image diagnosis text; the attribute text is a text representing an attribute of the medical image report, for example, a medical image acquisition device corresponding to the medical image report, an acquisition part corresponding to the medical image report, and the like, which is not limited in the present application. The image diagnosis text is diagnosis-related information that can be obtained from a medical image, such as a diagnosis conclusion, sign information observed from the image, and the like, and is not limited in this application. The text content included in the image diagnosis text may be different according to the medical image acquisition device, the device type, the corresponding portion, the generation mode of the image diagnosis text, and the like, which is not limited in this application.

Step 102: and carrying out integrity assessment by adopting a pre-trained machine reading understanding model according to the attribute text and the image diagnosis text to obtain an integrity assessment result of the image diagnosis text, wherein the integrity assessment result comprises the following steps: the integrity evaluation value is used for representing the integrity of a plurality of medical attention items corresponding to the medical image report in the image diagnosis text.

It should be noted that the integrity assessment refers to assessing the integrity of the image diagnosis text including the medical attention. Specifically, all medical attention items which should be included in the current medical image can be determined according to the attribute text, and then an integrity evaluation value is obtained according to the medical attention items and all medical attention items included in the image diagnosis text.

In one possible implementation, the integrity assessment value may be a percentage, that is, each obtained by calculating a percentage of the medical attention item and all the medical attention items included in the image diagnosis text;

in another possible implementation, the integrity assessment value may also be a score, such as by determining a score for medical attention items in all medical attention items, and calculating an integrated score for medical attention items included in all of the visual diagnostic texts.

In a specific implementation manner, a plurality of medical attention items corresponding to the medical image report can be determined according to the attribute text, and then a pre-trained machine reading understanding model is adopted to determine fields related to the medical attention items in the image diagnosis text so as to obtain an integrity evaluation value.

The foregoing is merely an example, and other implementations are possible in actual implementation, which is not limited in this application.

Step 103: and carrying out consistency evaluation on the medical image report according to the image diagnosis text and the pathological diagnosis report corresponding to the medical image report to obtain a diagnosis consistency evaluation result of the medical image report.

It should be noted that, the pathological diagnosis report refers to a diagnosis conclusion determined by a consultant (such as a doctor) according to the medical image; consistency refers to consistency of diagnostic conclusions determined by the visual diagnostic text with diagnostic conclusions in the pathological diagnostic report.

The diagnostic consistency evaluation result of the medical image report obtained after the consistency evaluation may be, for example, a conclusion about whether there is a consistency, or a consistency score corresponding to whether there is a consistency, which is not limited in the present application.

Step 104: and obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text and the diagnosis consistency evaluation result.

And finally, according to the integrity evaluation result of the image diagnosis text and the diagnosis consistency evaluation result, obtaining an evaluation result of the medical image report, and obtaining the evaluation result of the medical image report. The evaluation result of the medical image report may be presented in a manner of comprehensive score or integrity and consistency score, which is not limited in this application.

In summary, the embodiment of the application provides a medical image report evaluation method, after acquiring text content in a medical image report, performing integrity evaluation by adopting a pre-trained machine reading understanding model according to attribute text and image diagnosis text to obtain an integrity evaluation result of the image diagnosis text; carrying out consistency evaluation on the medical image report according to the image diagnosis text and the pathological diagnosis report corresponding to the medical image report to obtain a diagnosis consistency evaluation result of the medical image report; and finally, obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text and the diagnosis consistency evaluation result. Therefore, the method and the device realize the evaluation of the integrity, the consistency and the like of the image diagnosis text, and prompt the improvement direction of the medical image report by obtaining the evaluation result of the medical image report. In addition, the medical image report can guide related staff to improve the writing quality of the image diagnosis text.

Optionally, on the basis of fig. 1, the present application further provides a possible implementation manner of a medical image report evaluation method, and fig. 2 is a flowchart of a medical image report evaluation method provided in another embodiment of the present application; as shown in fig. 2, according to the attribute text and the image diagnosis text, performing integrity assessment by using a pre-trained machine reading understanding model to obtain an integrity assessment result of the image diagnosis text, including:

step 201: determining a plurality of medical attention items according to the attribute text;

it should be noted that, in the specific use, since the type of the medical image report (for example, CT, MRI, PET, etc.) and the disease, the location, etc. for which the medical image report is aimed may be different, the medical attention item to be focused may be different, so that it is first necessary to determine the medical attention item corresponding to the currently evaluated medical image report according to the attribute text. The attribute text is a text including information such as a medical image report type and a medical image target position, and at least one medical attention item can be uniquely determined according to the attribute text.

The determining the plurality of medical attention items may be implemented by a machine-readable understanding model, or may be implemented in other manners, which is not limited in this application.

It should be noted that the machine-readable understanding model may be a model generated based on a pre-trained language model, which may determine a plurality of medical attention items for a currently evaluated medical image report based on the attribute text. The pre-training language model may be, for example, a pre-training language token BERT (Bidirectional Encoder Representation from Transformers) model, which is not limited in this application.

Step 202: and according to the plurality of medical attention items and the image diagnosis text, evaluating by adopting a machine reading understanding model to obtain an integrity evaluation result.

In one possible implementation, the machine-readable understanding model may take as input a plurality of medical attention items, as well as visual diagnostic text, and output an integrity assessment result. Each medical attention item can further comprise dictionary information corresponding to the attention item, so that the machine reading understanding model can determine information corresponding to each medical attention item in the image diagnosis text according to the dictionary information.

In a specific implementation, the machine-readable understanding model may obtain the integrity assessment result by:

Firstly, the machine reading understanding model can realize the retrieval of the information corresponding to each medical attention item in the image diagnosis text, namely the information corresponding to each medical attention item can be determined in the image diagnosis text according to dictionary information.

And secondly, the machine reading understanding model can also realize reasoning of the information corresponding to each medical attention item in the image diagnosis text, namely when the information corresponding to one or more medical attention items does not exist in the determining process, the machine reading understanding model can conduct reasoning according to the image diagnosis text to obtain the information corresponding to the one or more medical attention items in a reasoning mode, and further determine whether the information corresponding to the medical attention item exists in the image diagnosis text. If there is still no information corresponding to one or more medical care items, it may be marked or the result of the medical care item may be made empty.

And finally, integrating the information corresponding to all the medical attention items to obtain an integrity evaluation result.

For example, the machine-readable understanding model is used to determine the visual diagnostic text "data of input Al model: after pancreatic cancer operation, drainage tube retention shadow can be seen in the abdominal cavity, pancreas head deficiency is that pancreas tail shape, size are normal, density is normal, and abnormal strengthening stove is not seen after strengthening; the liver is normal in size and proportion, the intrahepatic bile duct is not expanded, the left and right liver leaves are provided with a plurality of low-density foci with different sizes, the maximum is 37 mm or 47mm, the post-enhancement arterial phase is edge-strengthened, the portal pulse phase and the delay phase are centrally filled, and the lower section of the right front liver leaf is not strengthened. The gallbladder is not explicitly shown, spleen morphology: no abnormality was seen in size and density. Soft tissue shadows are visible around the common hepatic artery. After pancreatic cancer surgery, a soft tissue shadow surrounding the common artery was visible. General mark? Is lymph node metastasis? Note that in the review of hepatic multiple hemangioma hepatic right leaf cyst "there is information corresponding to the following items of interest:

Biliary tract evaluation: the bile duct is not expanded;

liver lesions: the left and right leaves of the liver see a plurality of low-density stoves with different sizes;

common hepatic artery: soft tissue shadow wrapping is visible around the common hepatic artery;

and calculating the score of the integrity evaluation result according to the set score of each item according to the medical attention item in which the corresponding information does not exist in the medical attention items.

For example, preset: undescribed items: "pancreatic duct assessment" (point 5), mesenteric vein "(point 6), portal vein (point 7), suspicious lymph node (point 5), intravenous tumor plug (point 6), collateral venous circulation (point 6), peritoneal or omentum nodule (point 5), ascites (point 5), invasion of peripancreatic gland: structure "(button 5 minutes). The final score was 50 points.

The foregoing is merely illustrative, and other implementations are possible in actual implementation, which is not limited in this application.

The method for evaluating the integrity of the image diagnosis text integrates dictionary information into the model, effectively utilizes external knowledge and improves the accuracy of the model. Meanwhile, the method and the device can be applied to medical image report evaluation of various different types, and have mobility. Based on machine reading understanding the model, converting the terms into questions, and inputting the questions into the model, the terms with a portion of a priori knowledge can improve model accuracy compared to other NLP methods. In addition, each medical attention item in the machine reading understanding model can also comprise dictionary information corresponding to the attention item, a large amount of priori knowledge is provided for the model, the labeling data amount is effectively reduced, the labeling cost is reduced, and the accuracy of the model can be improved for the model by improving the external knowledge.

Optionally, on the basis of fig. 1, the present application further provides a possible implementation manner of a medical image report evaluation method, where an integrity evaluation result further includes: position information of a text corresponding to a target medical attention item in the medical attention items; FIG. 3 is a flowchart of a medical image report evaluation method according to another embodiment of the present disclosure; as shown in fig. 3, the method further includes:

step 301: and according to the position information, adopting a pre-trained sign abnormality evaluation model to obtain an abnormality evaluation conclusion of the sign corresponding to the target attention item.

The integrity assessment results also include: the position information of the text corresponding to the target medical attention item in the medical attention items, namely the first position information of the text corresponding to the target medical attention item in the image diagnosis text.

It should be noted that, the evaluation of the sign abnormality evaluation model aims at determining whether the target medical attention item corresponds to the text belonging to the negative result or the positive result for the target medical attention item. For example, if the target medical attention item is a liver lesion, if the text corresponding to the target medical attention item is a text of 'a plurality of low-density ranges with different sizes of left and right leaves of the liver', the liver is indicated to have a lesion, namely the text corresponding to the target medical attention item is a positive result; if the text corresponding to the target medical attention item is 'liver is not abnormal', the fact that no lesion exists in the liver is indicated, and the text corresponding to the target medical attention item is a negative result.

In one possible implementation, the machine-readable understanding model further includes, prior to performing step 202: and carrying out coding processing on the image diagnosis text, thereby obtaining coding vector information corresponding to the image diagnosis text. On the basis, the pre-trained sign abnormality evaluation model can acquire an abnormality evaluation conclusion of the sign corresponding to the target medical attention item by extracting the corresponding coding vector information of the text corresponding to the target medical attention item.

Specifically, the pre-trained sign anomaly evaluation model can be used for acquiring the average value of the head and tail position vectors of the text corresponding to the target medical attention item and the 0 th position vector of the image diagnosis text, and splicing the average value of the head and tail position vectors and the 0 th position vector to serve as the input of the sign anomaly evaluation model so as to acquire the anomaly evaluation conclusion of the sign corresponding to the target attention item from the sign anomaly evaluation model.

For example, after [ CLS ] pancreatic cancer operation, drainage tube retention shadow can be seen in the abdominal cavity, pancreas head defects such as pancreas tail shape, normal size and normal density are avoided, and abnormal strengthening stove is not seen after strengthening; the liver is normal in size and proportion of each leaf, the intrahepatic bile duct is not expanded, and the left and right liver leaves are provided with a plurality of low-density stoves … … with different sizes.

When identifying that liver lesions correspond to abnormal treatment of body signs of a plurality of low-density stoves with different sizes of left and right lobes of the liver, obtaining coding information by an image diagnosis text through a model, splicing a [ CLS ]' position vector with the average value of the liver and stove position vectors, and finally carrying out assessment of the abnormal body signs by the model through classification.

step 302: and obtaining an evaluation result of the medical image report according to the integrity evaluation result, the diagnosis consistency evaluation result and the abnormal evaluation conclusion of the image diagnosis text.

Finally, according to the integrity evaluation result of the image diagnosis text, the consistency evaluation result and the abnormal evaluation conclusion are diagnosed, and the evaluation result of the medical image report is obtained, so that the evaluation result of the medical image report can be realized.

Optionally, on the basis of fig. 1, the present application further provides a possible implementation manner of a medical image report evaluation method, and fig. 4 is a flowchart of a medical image report evaluation method provided in a further embodiment of the present application; as shown in fig. 4, according to the image diagnosis text and the pathological diagnosis report corresponding to the medical image report, consistency evaluation is performed on the medical image report to obtain a diagnosis consistency evaluation result of the medical image report, including:

Step 401: according to the image diagnosis text and the pathological diagnosis report, determining standard terms of a plurality of first candidate diseases corresponding to the image diagnosis text and standard terms of a plurality of second candidate diseases corresponding to the pathological diagnosis report from a preset disease knowledge base respectively;

it should be noted that, the preset disease knowledge base is a knowledge base built by related knowledge of pathological diagnosis, for example, related text information such as disease standard terms, disease attribute knowledge, disease imaging knowledge, disease pathology knowledge and the like can be obtained from an open source medical data center by combing related diseases and related systems of each image observation part, irrelevant information is extracted and filtered through the information, and then a disease specific knowledge base corresponding to each image observation part is built.

Determining standard terms of a plurality of first candidate diseases corresponding to the image diagnosis text from a preset disease knowledge base (or determining an image observation part through attribute text and determining a disease knowledge base corresponding to the image observation part) according to at least one diagnosis conclusion extracted from the image diagnosis text;

and according to the pathological diagnosis report, determining standard terms of a plurality of second candidate diseases corresponding to the pathological diagnosis report from a preset disease knowledge base (or determining an image observation part through relevant information of patient examination in the pathological diagnosis report and from a disease knowledge base corresponding to the image observation part) according to at least one extracted diagnosis conclusion.

It should be further noted that, the specific number of the determined first candidate disease or the determined second candidate disease is not limited in this application, and the number may be set according to actual use needs.

In a specific implementation, for example, a BM25 algorithm may be used to screen a preset number of first candidate diseases and a preset number of second candidate diseases from a preset disease knowledge base for the image diagnosis text and the pathological diagnosis report, respectively.

Step 402: and determining a first target disease corresponding to the image diagnosis text from the plurality of first candidate diseases according to the image diagnosis text and the standard terms of the plurality of first candidate diseases.

And determining the candidate disease with the highest matching degree from the plurality of first candidate diseases as the first target disease corresponding to the image diagnosis text according to the matching degree of the image diagnosis text and the standard term of each first candidate disease.

In one possible implementation, the image diagnosis text may be combined with standard terms of each first candidate disease into sentences (more specifically, the image visible text in the image diagnosis text may be combined with standard terms of each first candidate disease (i.e., disease attribute knowledge of the first candidate disease, disease imaging knowledge, etc.) to construct Prompt learning text (e.g., a Prompt learning text may be constructed, and since the Prompt model can utilize pre-training language model priori knowledge, the amount of labeling data may be reduced in use, and the labeling cost reduced), X1 and X2 are the same disease, X1 represents the image diagnosis conclusion in the image diagnosis text, and X2 represents the first candidate disease.

And (3) utilizing the BERT model to infer whether [ MASK ] represents 'phase' and 'not', and predicting the score, wherein the candidate disease with the maximum score is the first target disease corresponding to the image diagnosis text.

Step 403: and determining a second target disease corresponding to the pathological diagnosis text from the plurality of second candidate diseases according to the pathological diagnosis text and standard terms of the plurality of second candidate diseases.

And determining the candidate disease with the highest matching degree from the plurality of second candidate diseases as the second target disease corresponding to the pathological diagnosis text according to the matching degree of the pathological diagnosis text and the standard term of each second candidate disease. The specific matching manner is referred to step 402, and will not be described herein.

Step 404: and carrying out consistency assessment on the medical image report according to the first target disease and the second target disease to obtain a diagnosis consistency assessment result of the medical image report.

And carrying out consistency assessment on the medical image report according to the first target disease and the second target disease: if the first target disease and the second target disease are the same, the diagnosis consistency evaluation result of the medical image report indicates that: the consistency is achieved; if the first target disease and the second target disease are different, the diagnosis consistency evaluation result of the medical image report indicates that: there is no consistency.

For example, determining an image diagnosis text ' pancreatic tail occupation ' from a plurality of first candidate diseases according to the image diagnosis text, considering pancreatic cancer liver multiple blood vessel @ tumor, and the first target disease corresponding to liver multiple cyst ' is pancreatic duct adenocarcinoma;

determining a pathological diagnosis text from a plurality of second candidate diseases according to the pathological diagnosis text, wherein the pathological diagnosis text is combined with an immunohistochemical marker and an HE slice, and diagnosis (pancreas puncture tissue A) of a visible abnormal epithelial nest indicates malignancy, and the second target disease corresponding to the large possibility of adenosquamous carcinoma is pancreatic adenosquamous carcinoma;

it is apparent that the first target disease is inconsistent with the second target disease, and the diagnostic consistency assessment result of the medical image report indicates that: there is no consistency.

In summary, in the evaluation of the medical image report diagnosis consistency, the disease is standardized by constructing the prompt learning text, the information of the preset disease knowledge base is fully utilized, the processing speed is high, and the accuracy is high.

Optionally, on the basis of fig. 1, the present application further provides a possible implementation manner of a medical image report evaluation method, and fig. 5 is a flowchart of a medical image report evaluation method provided in still another two embodiments of the present application; as shown in fig. 5, the method further includes:

Step 501: and adopting a pre-trained diagnosis conclusion recognition model to recognize the diagnosis conclusion of the image diagnosis text, and obtaining a diagnosis conclusion recognition result corresponding to the image diagnosis text.

In one possible implementation manner, a pre-trained diagnosis conclusion identification model is adopted to identify diagnosis conclusion of the image diagnosis text, whether the diagnosis conclusion exists in the image diagnosis text is judged, and if the diagnosis conclusion exists, the diagnosis conclusion identification result corresponding to the image diagnosis text indicates that the diagnosis conclusion identification exists; and if the diagnosis result does not exist, the diagnosis result identification result corresponding to the image diagnosis text indicates that the diagnosis result identification does not exist.

In one specific implementation, the pre-trained diagnostic conclusion recognition model may be implemented as follows: firstly, adopting a BERT+ full-connection layer to carry out sequence labeling, obtaining a coding vector through BERT when a diagnosis conclusion is extracted, then obtaining a label set through full-connection layer mapping, and carrying out Softmax processing on the set vector, wherein the output value of each dimension represents the probability that a single token is of a certain class. And carrying out diagnosis conclusion identification through the scheme, and if the extraction result is null, judging that the diagnosis conclusion identification does not exist.

The foregoing is merely an example, and other ways of identifying a diagnosis conclusion may be used in practical implementations, which are not limited in this application.

step 502: and obtaining an evaluation result of the medical image report according to the integrity evaluation result, the diagnosis consistency evaluation result and the diagnosis conclusion identification result of the image diagnosis text.

Finally, according to the integrity evaluation result of the image diagnosis text, the diagnosis consistency evaluation result and the diagnosis conclusion identification result, the evaluation result of the medical image report is obtained, and the evaluation result of the medical image report can be realized.

In one possible implementation, before step 502, the method further includes: and adopting a pre-trained diagnosis conclusion recognition model to carry out diagnosis conclusion recognition on the pathological diagnosis conclusion text, and obtaining a diagnosis conclusion recognition result corresponding to the pathological diagnosis text. The method comprises the steps of adopting a pre-trained diagnosis conclusion identification model to identify a diagnosis conclusion on a pathological diagnosis conclusion text, judging whether the diagnosis conclusion exists in the pathological diagnosis conclusion text, and if the diagnosis conclusion exists, indicating that the diagnosis conclusion identification exists by a diagnosis conclusion identification result corresponding to the pathological diagnosis conclusion text; and if the diagnosis result does not exist, the diagnosis result identification result corresponding to the pathological diagnosis result text indicates that the diagnosis result identification does not exist.

The diagnosis conclusion identifying result may include: diagnosis result identification results corresponding to the pathological diagnosis text and diagnosis result identification results corresponding to the image diagnosis text.

Optionally, on the basis of fig. 1, the present application further provides a possible implementation manner of a medical image report evaluation method, and fig. 6 is a flowchart of a medical image report evaluation method provided in a further embodiment of the present application; as shown in fig. 6, the method further includes:

step 601: and determining the abnormal part corresponding to the medical image report according to the image diagnosis text.

In one possible implementation, the abnormal location in the medical image report may be determined from the medical image report by normalization (or may be determined according to the abnormal assessment conclusion described above), which is not limited in this application.

Step 602: and obtaining the matching result of the checking part of the medical image report according to the abnormal part and the diagnosis conclusion in the image diagnosis text.

In one possible implementation, the abnormal part and the diagnosis conclusion in the image diagnosis text can be matched, whether the abnormal part and the diagnosis conclusion in the image diagnosis text are matched or not is judged, if so, the matching result of the checking part of the medical image report indicates that the matching of the checking part is normal, and if not, the matching result of the checking part of the medical image report indicates that the matching of the checking part is abnormal.

step 603: and obtaining an evaluation result of the medical image report according to the integrity evaluation result, the diagnosis consistency evaluation result and the checking position matching result of the image diagnosis text.

Finally, according to the integrity evaluation result of the image diagnosis text, the consistency evaluation result and the checking part matching result are diagnosed, and the evaluation result of the medical image report is obtained, so that the evaluation result of the medical image report can be realized.

Optionally, on the basis of fig. 1, the present application further provides a possible implementation manner of a medical image report evaluation method, and fig. 7 is a flowchart of a medical image report evaluation method provided in a fourth embodiment of the present application; as shown in fig. 7, the method further includes:

step 701: and carrying out integrity assessment of patient information on the attribute text, and carrying out integrity assessment of examination information on the image diagnosis text to obtain an integrity conclusion of the patient information and an integrity conclusion of the examination information.

In one possible implementation manner, the patient information in the attribute text and the examination information in the image diagnosis text can be traversed respectively, so that whether the patient information is complete or not is determined, whether the examination information is complete or not is determined, and then the patient information integrity conclusion and the examination information integrity conclusion are obtained.

The foregoing is merely an example, and other ways of evaluating the integrity are possible in practical implementation, which are not limited in this application.

Step 702: and obtaining an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text, the diagnosis consistency evaluation result, the patient information integrity conclusion and the examination information integrity conclusion.

Finally, according to the integrity evaluation result of the image diagnosis text, the diagnosis consistency evaluation result, the patient information integrity conclusion and the examination information integrity conclusion, the evaluation result of the medical image report is obtained, and the evaluation result of the medical image report can be realized.

In a specific implementation, pancreatic cancer is exemplified. According to the medical image report quality evaluation method and device, the medical image report quality evaluation can be carried out according to the text content in the input medical image report and the corresponding pathological diagnosis report. The main evaluation content comprises the integrity of the image diagnosis text, the consistency of the image diagnosis text conclusion and the pathological diagnosis report conclusion and the like. A complete image diagnosis text generally includes descriptions of organ parenchyma, solid tumors, arterial evaluation, venous evaluation, etc. And a report in which the visual diagnosis text conclusion is consistent with the pathological diagnosis report conclusion requires that the standard terms of the visual diagnosis text conclusion be consistent with the standard terms of the pathological diagnosis report. The medical image report can be evaluated by the doctor or related staff, so that the writing quality of the image report is improved, the working intensity of the quality control staff of the hospital can be reduced, and the working efficiency is improved.

The present application is directed to form and content quality control for text content in medical image reports using a variety of angular aspects. Accuracy is considered while quality control efficiency is improved. The method not only carries out complete quality control on the image report, but also carries out diagnosis consistency judgment by combining the diagnosis conclusion of the pathological diagnosis report so as to evaluate the diagnosis accuracy of the image report.

The following is a description of a medical image report evaluation device, an electronic device, a storage medium, etc. for executing the medical image report evaluation device, the electronic device, the storage medium, etc. provided by the present application, and specific implementation processes and technical effects thereof are referred to above, and are not repeated herein.

The embodiment of the application provides a possible implementation example of the medical image report evaluation device, which can execute the medical image report evaluation method provided by the embodiment. Fig. 8 is a schematic diagram of a medical image report evaluation device according to an embodiment of the present application. As shown in fig. 8, the medical image report evaluating apparatus includes:

the obtaining module 81 is configured to obtain text content in the medical image report, where the text content includes: attribute text and image diagnosis text; the attribute text is used for representing the attribute of the medical image report;

The integrity evaluation module 83 is configured to perform integrity evaluation by using a pre-trained machine reading understanding model according to the attribute text and the image diagnosis text, so as to obtain an integrity evaluation result of the image diagnosis text, where the integrity evaluation result includes: the integrity evaluation value is used for representing the integrity of a plurality of medical attention items corresponding to the medical image report in the image diagnosis text;

the consistency evaluation module 85 is configured to perform consistency evaluation on the medical image report according to the image diagnosis text and the pathological diagnosis report corresponding to the medical image report, so as to obtain a diagnosis consistency evaluation result of the medical image report;

the evaluation result generating module 87 is configured to obtain an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text and the diagnosis consistency evaluation result.

Optionally, an integrity assessment module 83, configured to determine a plurality of medical attention items according to the attribute text; and according to the plurality of medical attention items and the image diagnosis text, evaluating by adopting a machine reading understanding model to obtain an integrity evaluation result.

Optionally, the integrity assessment result further includes: position information of a text corresponding to a target medical attention item in the medical attention items; the abnormality evaluation module is used for obtaining an abnormality evaluation conclusion of the sign corresponding to the target attention item by adopting a pre-trained sign abnormality evaluation model according to the position information;

The evaluation result generating module 87 is configured to obtain an evaluation result of the medical image report according to the integrity evaluation result, the diagnostic consistency evaluation result and the abnormal evaluation conclusion of the image diagnostic text.

Optionally, the consistency evaluation module 85 is configured to determine, according to the image diagnosis text and the pathological diagnosis report, standard terms of a plurality of first candidate diseases corresponding to the image diagnosis text and standard terms of a plurality of second candidate diseases corresponding to the pathological diagnosis report from a preset disease knowledge base respectively; determining a first target disease corresponding to the image diagnosis text from the plurality of first candidate diseases according to the image diagnosis text and standard terms of the plurality of first candidate diseases; determining a first target disease corresponding to the pathological diagnosis text from the plurality of second candidate diseases according to the pathological diagnosis text and standard terms of the plurality of second candidate diseases; and carrying out consistency assessment on the medical image report according to the first target disease and the second target disease to obtain a diagnosis consistency assessment result of the medical image report.

Optionally, the diagnostic conclusion identifying module is used for identifying the diagnostic conclusion of the image diagnostic text by adopting a pre-trained diagnostic conclusion identifying model to obtain a diagnostic conclusion identifying result corresponding to the image diagnostic text;

The evaluation result generating module 87 is configured to obtain an evaluation result of the medical image report according to the integrity evaluation result, the diagnostic consistency evaluation result and the diagnostic conclusion identification result of the image diagnostic text.

Optionally, the checking part matching module is used for determining an abnormal part corresponding to the medical image report according to the image diagnosis text; obtaining a checking part matching result of the medical image report according to the abnormal part and the diagnosis conclusion in the image diagnosis text;

the evaluation result generating module 87 is configured to obtain an evaluation result of the medical image report according to the integrity evaluation result, the diagnostic consistency evaluation result and the examination part matching result of the image diagnostic text.

Optionally, the information integrity evaluation module is used for respectively carrying out integrity evaluation of patient information on the attribute text, carrying out integrity evaluation of the inspection information on the image diagnosis text, and obtaining an integrity conclusion of the patient information and an integrity conclusion of the inspection information;

the evaluation result generating module 87 is configured to obtain an evaluation result of the medical image report according to the integrity evaluation result of the image diagnosis text, the diagnosis consistency evaluation result, the patient information integrity conclusion and the examination information integrity conclusion.

The foregoing apparatus is used for executing the method provided in the foregoing embodiment, and its implementation principle and technical effects are similar, and are not described herein again.

The above modules may be one or more integrated circuits configured to implement the above methods, for example: one or more application specific integrated circuits (Application Specific Integrated Circuit, abbreviated as ASIC), or one or more microprocessors (digital singnal processor, abbreviated as DSP), or one or more field programmable gate arrays (Field Programmable Gate Array, abbreviated as FPGA), or the like. For another example, when a module above is implemented in the form of a processing element scheduler code, the processing element may be a general-purpose processor, such as a central processing unit (Central Processing Unit, CPU) or other processor that may invoke the program code. For another example, the modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

The embodiment of the application provides a possible implementation example of an electronic device, which can execute the medical image report evaluation method provided by the embodiment. Fig. 9 is a schematic diagram of an electronic device provided in an embodiment of the present application, where the device may be integrated in a terminal device or a chip of the terminal device, and the terminal may be a computing device with a data processing function.

The electronic device includes: the medical image report evaluation method comprises a processor 901, a storage medium 902 and a bus, wherein the storage medium stores program instructions executable by the processor, when the control device runs, the processor communicates with the storage medium through the bus, and the processor executes the program instructions to execute the steps of the medical image report evaluation method. The specific implementation manner and the technical effect are similar, and are not repeated here.

The present application provides a possible implementation example of a computer readable storage medium, which can execute the medical image report evaluation method provided in the foregoing embodiment, and the storage medium stores a computer program, where the computer program is executed by a processor to execute the steps of the medical image report evaluation method.

A computer program stored on a storage medium may include instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (english: processor) to perform some of the steps of the methods of the various embodiments of the invention. And the aforementioned storage medium includes: u disk, mobile hard disk, read-Only Memory (ROM), random access Memory (Random Access Memory, RAM), magnetic disk or optical disk, etc.

In the several embodiments provided by the present invention, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of elements is merely a logical functional division, and there may be additional divisions of actual implementation, e.g., multiple elements or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of 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.

In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in hardware plus software functional units.

The integrated units implemented in the form of software functional units described above may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium, and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (english: processor) to perform part of the steps of the methods of the embodiments of the invention. And the aforementioned storage medium includes: u disk, mobile hard disk, read-Only Memory (ROM), random access Memory (Random Access Memory, RAM), magnetic disk or optical disk, etc.

The foregoing is merely a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think about changes or substitutions within the technical scope of the present application, and the changes or substitutions are covered in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A medical image report evaluation method, comprising:

2. The method of claim 1, wherein the performing the integrity assessment using a pre-trained machine-readable understanding model based on the attribute text and the visual diagnostic text to obtain the integrity assessment result of the visual diagnostic text comprises:

3. The method of claim 1, wherein the integrity assessment result further comprises: position information of a text corresponding to a target medical attention item in the medical attention items; the method further comprises the steps of:

4. The method of claim 1, wherein the performing consistency assessment on the medical image report according to the image diagnosis text and the pathological diagnosis report corresponding to the medical image report to obtain a diagnosis consistency assessment result of the medical image report comprises:

5. The method of claim 1, wherein the method further comprises:

6. The method of claim 1, wherein the method further comprises:

7. The method of claim 1, wherein the method further comprises:

8. A medical image report evaluating apparatus, comprising:

9. An electronic device, comprising: a processor, a storage medium and a bus, the storage medium storing program instructions executable by the processor, the processor and the storage medium communicating via the bus when the electronic device is running, the processor executing the program instructions to perform the steps of the medical image report evaluation method according to any one of claims 1 to 7 when executed.

10. A computer-readable storage medium, characterized in that the storage medium has stored thereon a computer program which, when executed by a processor, performs the steps of the medical image report evaluation method according to any one of claims 1 to 7.